AI adoption is stalling because businesses are optimizing tools instead of redesigning workflows

The real AI advantage begins when entrepreneurs stop limiting their own potential

Kirk Drake, founder of CU 2.0, says the main obstacle to AI adoption is not cost or complexity but the assumptions entrepreneurs bring to the technology. Most organizations already have the knowledge AI needs but have never organized it. Businesses that document their values, workflows, and brand voice before introducing AI get consistent results. Those that wait risk falling years behind competitors who started with small, incremental experiments. Artificial intelligence has rapidly become one of the defining business conversations of the decade. Yet despite growing investment and widespread discussion, many organizations continue to approach AI with hesitation. Research found that while organizations remain optimistic about AI's long-term value, turning experimentation into measurable business outcomes often depends less on the technology itself than on organizational readiness and leadership execution. Kirk Drake, founder of CU 2.0, believes many business leaders are focusing on the wrong challenge. Through CU 2.0, he works with organizations on AI strategy, digital transformation, and leadership development, helping executives understand how emerging technologies can be applied in practical business settings. From his perspective, the greatest obstacle to successful AI adoption is rarely technical capability. Instead, it stems from assumptions that AI is too expensive, too complicated, or too difficult for smaller organizations to implement. "The barriers entrepreneurs see are often barriers they've created themselves," Drake says. "Most businesses already have the knowledge AI needs. They simply haven't organized it in a way that allows the technology to understand it." According to Drake, one of the biggest misconceptions is that AI removes the human element from customer relationships. He argues the opposite can be true. Businesses have always wanted to deliver more personalized experiences, but most lacked the time and resources to do so consistently. AI, he explains, creates opportunities for personalization that would have been impractical even a few years ago. He points to organizations that build structured brand guidance, departmental communication styles, and individual workflows before introducing AI into daily operations. Those foundations, he says, allow technology to reinforce a company's personality rather than replace it. Drake believes this is where many entrepreneurs unintentionally limit themselves. They expect AI to define their business identity instead of recognizing that the technology reflects the quality of the information it receives. Businesses that have never clearly documented their values, workflows, or brand voice often mistake inconsistent AI output for technological weakness when, in reality, those inconsistencies already existed within the organization. "The technology is simply exposing gaps that were already there," Drake explains. "If you don't understand your business well enough to explain it, how can you expect AI to replicate it?" He also encourages leaders to rethink what successful AI adoption actually looks like. Many assume implementation requires major budgets, dedicated technical teams, or months of preparation. Drake argues that meaningful progress often begins with something much simpler: developing better prompts, documenting existing knowledge, and allowing AI to analyze work that already exists. He notes that entrepreneurs frequently underestimate how much information their organizations have already created. Existing emails, websites, presentations, procedures, and customer communications often contain enough context for AI to begin identifying patterns, generating documentation, and supporting daily work. Drake suggests that AI adoption should be viewed as a learning journey rather than a technology project. Teams that develop familiarity through everyday experimentation gradually build confidence before tackling more sophisticated implementations. In his experience, those incremental improvements compound over time, creating lasting operational advantages. That philosophy is rooted in a personal lesson. Reflecting on the early internet era, Drake recalls dismissing the significance of websites before eventually recognizing how transformative they would become for business. Looking back, he considers that hesitation one of the most valuable lessons of his career. "I promised myself I would never make that mistake again," he says. "Even if it means investing an extra hour every week to understand where technology is going, that small investment can shape the next twenty years of your business." The pace of AI development makes continuous learning increasingly important. Every new capability builds upon previous understanding, meaning businesses that begin developing practical experience today are often better positioned to adapt tomorrow. Organizations that delay entirely may eventually face the much more difficult challenge of catching up after competitors have accumulated months or years of experience. Drake believes entrepreneurs ultimately face a decision that extends beyond software selection or operational efficiency. AI can be viewed as another business expense or as an opportunity to expand knowledge, strengthen leadership, and unlock capabilities that were previously beyond the reach of smaller organizations. "The future belongs to the people who stay curious," Kirk Drake says. "AI is ultimately another skill you can learn. The decision to embrace that learning will shape not only your future, but the future of everyone your business serves."

The cost of being half-hearted in AI and how to avoid the Solow Paradox

Why a lot of projects don't deliver ROI, and how enterprises can tackle it Not too long ago, one of the world's most data-driven companies admitted that its AI spending was becoming "harder to justify". Uber's President and COO, Andrew Macdonald, told the Rapid Response podcast that the company had blown through its entire 2026 AI budget in roughly four months, with around 5,000 engineers leaning on Anthropic's Claude Code. Uber isn't alone. Forrester research found that enterprises are deferring around 25% of planned AI spend to 2027, as CFO scrutiny over ROI intensifies. And McKinsey's State of AI report summarized that while 62% of companies are experimenting with AI agents, only 23% have scaled them in even a single business function. While these may look like the statistics of a technology that isn't working, they're actually the statistics of a technology being used in the wrong way. If we rewind back to 1987, Nobel laureate economist, Robert Solow, observed something that many at the time really resonated with: "You can see the computer age everywhere but in the productivity statistics." Computers were everywhere across the innovative businesses that had invested heavily in them, but the productivity numbers didn't move. The returns only materialized years later, once organizations stopped bolting computers onto old processes and started fundamentally redesigning how they worked. Many executives, alongside Uber's COO, are at exactly that inflection point with AI - people are using it, running out of budgets to maintain usage, and at the same time, not really seeing the productivity boost they were hoping for. This is the Solow Paradox repeating its course, which begs the question: will enterprises learn from previous mistakes? The flatline behind the hype There's a critical distinction that most enterprise leaders are still failing to make: AI activity is not the same as AI maturity. You can run 40 pilots, adopt six platforms and report impressive usage statistics, and still be no closer to measurable business value. Uber found this out in painful, public fashion, and has since openly questioned whether the rising cost of AI token usage is translating into proportional productivity gains. What makes Uber's situation instructive is not just the financial exposure, but how the organization approached adoption. Internal leaderboards were introduced to rank teams by AI tool usage, with the incentive being to use more tools. The outcome was more usage, but the business impact slowly became harder to justify. This is what happens when you gamify adoption without redesigning the workflows underneath it. You optimize the tool usage metric, not the business outcome. Uber has now joined several other top organizations, including Microsoft, Meta and Amazon, in capping AI usage to tackle the issue. What's interesting here is that when AI token usage is unconstrained, activity becomes the proxy for progress, but when it's capped, organizations are forced to confront a harder question: what is each token actually producing? In that sense, token spend behaves like an economic mirror. It scales immediately with adoption, while productivity only improves when workflows are redesigned. The gap between the two is where most AI ROI disappears. The real culprit: Individual task optimization When AI tools are deployed at the individual level, they tend to optimize the task, not the workflow. A developer writes code faster, a marketer drafts copy in a fraction of the time, or a data analyst generates a summary report in minutes rather than hours. All of these examples are real gains, but if the code still sits in a review queue for four days, if the draft still passes through three rounds of manual approval, or if the report still requires someone to manually transfer it into a decision-making dashboard - the time saved will pool at the next bottleneck. Individual productivity gains that don't translate into workflow redesign don't compound. They stagnate, and this is the core of the maturity gap. AI maturity isn't about how many tools you've adopted, or how many pilots you've launched. It's about whether you've moved consistently from opportunity to outcome. That shift requires a fundamentally different way of working. Now, the term 'production-ready AI' gets used loosely. It's worth being precise about what it actually means in practice, because most enterprise AI deployments fall short of the bar. Production-ready AI has four characteristics. First, the output feeds directly into a downstream decision or action without manual transfer. It's embedded in the workflow, not adjacent to it. Second, the system has clearly defined failure modes, so the organization knows exactly what happens when the AI gets something wrong, and who's accountable for remedying it. Third, there's a named owner responsible for performance, adoption and iteration. Finally, and most critically, the surrounding process has been redesigned - not merely augmented. The framework that closes the gap So, how can businesses make this shift in practice? The answer is maintaining disciplined execution. One way to build that discipline is a structured cadence we call the 3-3-3 framework: three days to prioritize, three weeks to prove value, and three months to launch a first release. The logic is deceptively simple and deliberately structured. In the prioritization phase, the question is not "what can AI do?", it's "which specific opportunity, tied to a specific business outcome, has the right combination of value, feasibility, data readiness, and organizational sponsorship to pursue right now?" That focus alone eliminates a significant proportion of AI initiatives that consume resources without clear purpose. It's the antidote to the open-ended experimentation that left Uber burning through its budget before April was out. In the proof phase, the focus is validation - not in technical terms, but in commercial ones. Can this solution create measurable value for users and for the business? A proof of concept that demonstrates technical possibility without demonstrating business value isn't a proof of concept, it's a prototype without a destination. In the launch phase, the solution moves into a real environment. It integrates with existing systems, gets adopted by the people it was built for and gets measured against the outcome it was designed to improve. Not against token usage and not against adoption rates. This rhythm isn't a rigid formula, however. The shape of the work always depends on the business problem, the data environment, the technical complexity and the organization's appetite for change. What the framework provides is momentum and the discipline to keep that momentum anchored in value. Shifting from adoption metrics to outcome metrics The most consequential change enterprise leaders can make right now is a measurement decision. Enterprises that are serious about closing the gap between AI activity and business impact need to make three shifts. The first is from tool deployment to operating model redesign. Rolling out AI tools is table stakes but building a repeatable operating model - a structured path from idea to proof to scale - is the competitive differentiator. The second shift is from adoption metrics to outcome metrics. Usage rates, logins and token volumes tell you whether people are using the tools, but they don't tell you whether the tools are working. Define success in business terms from the outset: cost reduction, time-to-decision, revenue impact, customer satisfaction, and build your measurement framework around those outcomes. The third shift is from centralized experimentation to distributed accountability. The AI factory model - where a central operating model connects business priorities with delivery, adoption and value measurement - works precisely because it distributes accountability across the organization. Every initiative starts with a clear owner, a clear problem and a clear definition of what success looks like. Winning the AI day The productivity paradox Solow identified in 1987 eventually resolved itself. Not because computers got better - though they did - but because organizations learned to reorganize work around the technology, rather than fitting the technology around old ways of working. The same resolution is available to enterprises deploying AI today. But it requires leaders to make a deliberate choice - to stop measuring success in terms of how many tools have been adopted and start measuring it in terms of how many outcomes have been delivered. The organizations that win the AI era will not necessarily be those with the largest number of pilots, but rather the ones that build the maturity to turn the right ideas into value and then scale that value with confidence. That kind of maturity doesn't happen by accident. It requires structure, discipline and the willingness to ask harder questions about what AI is actually delivering, and what it's not. We've featured the best AI chatbot for business. This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Ask your employees one question about AI. The silence will tell you everything | Fortune

I recently ran an AI strategy session for an organization's leadership team. They had everything the playbooks prescribe. Trainings. Access to multiple AI systems. A no-code platform. So I asked: How many have built something with AI that changed how work gets done? One hand raised. Most considered themselves users of AI, but not builders. No personal agents, no custom assistants or reusable workflows. That distinction matters. Assistance creates a one-time productivity gain. Building turns that gain into a reusable tool or workflow that can scale. While most leaders track how many employees use AI, the more revealing question is how many are building with it. Call it the builder activation gap: the distance between the many people who could build with AI and the few who do. Across executive education sessions and applied AI courses for working professionals, the pattern I see is the same. Nearly anyone who can describe what they want in plain English can now build a working assistant, app or automation without writing a line of code. Yet few are building anything useful. A couple of years ago, Caroline Davis, Chief of Staff at Capital Factory, saw herself as an AI user rather than a builder. When I asked her applied AI course who had ever built a working tool with the technology, her hand stayed down. Today, many recurring parts of her job run through tools she built. Chief among them is an agent called Sunny, named after her daughter. It connects to Davis's email, calendar, Airtable CRM, and Google Sheets and draws on roughly a dozen documented workflows that prepare briefs, track fundraising, onboard new investors, and more. Data pulls that once took hours now take 10 to 15 minutes. Several automations run on a schedule, completing work before she asks for it. The workflows are versioned, reused, and improved rather than disappearing after a single interaction, and Sunny regularly coordinates with other agents. She has even carried the same approach outside her day job, using AI to rebuild a photography business she had operated a decade earlier, including its website. The path began with something smaller. In the applied AI course, she used natural language to build her first working AI assistant. Davis describes what followed in terms of confidence rather than technical mastery or coding know-how. That experience, she says, "built up my acumen as a whole and gave me the confidence to test out stronger AI models." She stopped seeing AI as a reactive conversational helper and started seeing it as something she could leverage to run recurring work. The first build led to the next, and eventually to the library behind Sunny. Along the way, her self-perception shifted. She still resists the language of expertise. "I don't think I'm a power user by any means," she told me, "but most of my day is run through Claude at this point." Her trajectory is still rare. Roughly half of U.S. employees now use AI on the job at least occasionally, but only 15% are daily users, says Gallup. Another recent study in HBR, based on an analysis of 1.4 million AI interactions among more than 2,500 KPMG employees, found only about 5% of employees qualified as sophisticated users, the ones doing iterative, higher impact work beyond casual prompting. For most, using AI means assistance with the one-off task in front of them, like drafting emails or summarizing documents. Useful, but disposable. So where are all the builders? Some barriers to building are structural. Governance, access, time, and incentives all matter. But once those basics are in place, identity can be the hidden bottleneck. The problem is that enterprise work has long trained people into a division of labor. A few specialists build systems, everyone else operates inside them. That made sense when building took engineering and coding expertise. For many problems, it doesn't anymore. What hasn't changed is identity. Most employees simply see themselves as consumers of technology, not creators. And identity is stubborn. Herminia Ibarra's work on reinvention shows that people rarely think their way into a new self-image. They act their way into it, and identity catches up. Plenty of building still belongs to specialists, including complex systems, security-sensitive applications, and anything going to customers without supervision. But a huge share of everyday work problems live somewhere safer, where the person building the tool is the same one who can tell whether it works. What stops many is the perception that building is "not my lane." While AI has made building far more accessible, it hasn't yet made most people believe it's for them. Every organization is sitting on people who are where Davis was two years ago. The question is whether leaders leave them there. Three practices help leaders activate a builder identity across their workforce. Organizations searching for more value from AI won't find it in adoption numbers alone. They'll need to narrow the builder activation gap, turning more employees who see themselves as users into builders. So go back to the opening question. Ask your people what they've actually built and see whether the room goes quiet. The work is making sure it doesn't. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune. Will Drover is Professor of Entrepreneurship & Innovation and Department Chair at the Neeley School of Business, Texas Christian University, where he serves as Founding Director of Neeley AI Forward, a school-wide initiative, and as the Dean's Advisor on AI. His work on AI adoption has appeared in MIT Sloan Management Review with coverage in the Wall Street Journal and Los Angeles Times. Drover teaches graduate courses on applied AI, runs executive education programs on AI strategy and leadership, and in practice holds ownership stakes in early-stage AI and robotics ventures.

Why AI is making work faster, not better.

Faster tools, slower progress: the productivity paradox of AI We've been sold a comforting idea about artificial intelligence: that it's making us dramatically more productive. Faster outputs, smarter tools, less effort. A quiet revolution in how we work. But step back for a moment and ask yourself a simple question. Do you actually feel more productive? Not faster. Not busier. Productive. Because for most professionals I speak to, the answer is no. Work feels quicker, yes. But also more fragmented, more reactive, and oddly more exhausting. The promise of efficiency is there on paper, but the true experience tells a different story. That disconnect is worth paying attention to. A typical working day Look at how most of us spend a typical working day. We move between email, calendar, tasks, notes, messaging platforms, documents. Each tool holds a piece of the puzzle, none of them are the full picture. So, we become the system that stitches it together. We check an email, then jump to our calendar to understand the context. We open a task list, then search our notes to remember why that task exists. We respond to a message, then dig through previous threads to find what was agreed. This is not the work itself. It's the management of work. Now add AI into the mix. We have tools that can summarize emails, draft responses, transcribe meetings, generate notes, and even suggest tasks. Each of these capabilities is impressive in isolation. They save minutes here, seconds there. But they don't remove the fundamental problem. In many cases, they amplify it. Instead of switching between tools, we now switch between tools and their respective AI layers. An assistant in your inbox. Another in your document editor. Another in your meeting tool. Each one helpful, but none aware of the others. So, we're still managing everything ourselves. We're just doing it faster. This is where the narrative around AI productivity starts to unravel. Defining success We've defined success as speed. How quickly can a tool help you write, summarize, respond, or organize? And to be fair, AI has delivered on that front. But speed without context is a blunt instrument. If you're responding faster but to the wrong priorities, you're not more productive. If you're generating more output but not moving meaningful work forward, you're simply accelerating noise. The real friction in modern work isn't the execution of tasks. It's the constant need to decide what matters, to reconstruct context, to align fragmented information across multiple systems. AI, as it stands today, rarely addresses that layer. At warpSpeed, we've approached this from a slightly different angle. We didn't start by asking how to make tasks faster. We started by asking why work feels so disjointed in the first place. The answer was fairly obvious: everything is scattered. Email lives in one place, calendar in another, tasks somewhere else, notes somewhere else again. Every decision requires jumping between them. So, we focused on bringing those elements together into a single, connected environment. a system where context flows naturally between them, facilitated by AI. Not as a feature bolted onto individual tools, but as something that can see across them. Something that understands not just a single email or a single note, but the relationship between your communications, your commitments, and your priorities. The difference, while subtle, is meaningful. This is how I like to think a successful assistant would function. For example, when someone asks, "What should I focus on today?", the answer isn't generated in isolation. It draws on overdue tasks, unread emails that require responses, upcoming meetings, and previous commitments. It reflects the reality of that person's day. Small changes Similarly, we've seen how small changes in interaction design can shift behavior. One example is email. By rethinking how users move through their inbox, we've seen people process large volumes of emails in a fraction of the time they previously spent. Not because they're working harder, but because the system reduces friction and surfaces what matters. These are not dramatic, headline-grabbing transformations. They're incremental improvements grounded in real workflows. And importantly, they're imperfect. We're still learning, still refining, still discovering where the real value lies. But they point to something broader. If AI is to genuinely deliver on its promise of productivity, we need to rethink what we're asking it to do. Right now, most tools are designed to assist with tasks. Write this. Summarize that. Suggest a response. Create a list. What's missing is a deeper understanding of context and personalization. Who is this for? Why does it matter? What else is happening around it? What should take priority? Without that layer, AI remains reactive. It responds to prompts, but it doesn't help you navigate your day. Moving forward To move forward, the industry needs to shift in three ways. First, from isolated tools to connected systems. The value of AI increases exponentially when it can operate across your entire workflow, not just within a single tool. Second, from generic intelligence to personal context. The most useful AI will be shaped by how you work, what you care about, and how you make decisions. Third, from output to outcome. It's not enough to generate content or complete tasks. The goal should be to move work forward in a meaningful way. None of this is easy. It requires rethinking product design, data architecture, and user experience at a fundamental level. It also requires a degree of restraint. Not every problem needs another feature. Sometimes it needs fewer moving parts. Productivity at scale The irony is that the more powerful AI becomes, the more important simplicity becomes. Productivity at scale depends on removing the need to think through complexity, making intuition more valuable than ever. Because ultimately, productivity isn't about doing more things. It's about doing the right things with less friction. AI isn't broken. But the way we're using it might be. If we continue to layer intelligence on top of fragmented systems, we'll keep getting the same result: faster work, but not better work. The real opportunity lies in something quieter, but far more impactful. Using AI to remove the need to manage work in the first place. Not to help you keep up. But to help you stay focused on what actually matters. We've reviewed, rated, and ranked the best business laptops. This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

How Epignosis learned to embed AI into its employee workflows

Although all too few organizations appear to be listening, it has become a well quoted fact that AI implementations are unlikely to succeed unless the technology is embedded into day-to-day employee workflows. That MIT study last summer was among the first to make the point. It found that 95% of enterprise AI pilots fail due to a lack of effective integration. As a result, they provide little or no measurable benefit to the bottom line. But one company that has worked hard to break this pattern is Epignosis. A workplace learning technology supplier, it employs 300 people who service the needs of 70,000 customers worldwide. Last year, its Chief Executive Nikhil Arora mandated that each employee should use AI daily to help augment their work. Everyone was given a subscription to OpenAI's ChatGPT Large Language Model (LLM) as a result. But a key problem came to light: there was no means of understanding how much the tool was being used. There was also no apparent improvement in company key performance indicators either. On recognizing that the LLM was mainly being employed as a glorified search engine, the decision was made to bring in AI learning platform provider, Mindstone. The aim here was to offer AI competency training to the entire workforce. Tackling employee resistance But a process that should have taken three weeks ended up taking about three months due to employee resistance. Excuses ranged from a lack of time to people not seeing the value, or a belief they were already proficient. Dimitris Damaskos, Epignosis' Head of Business Systems and AI-Enablement and a team of three, explains: We realized that just having top management evangelizing AI wasn't enough. We needed to energize the top and bottom at the same time, and the rest would follow. So, my team was created because AI is nothing without context. We also took AI Champions who already existed, those who already had successful GPTs and we knew were good at AI. Then we said we're opening up the Champions program to the whole company, so 'come and help us write the organization's AI strategy'. The point was to make everyone feel included and part of the decision-making process as "people resist change when it comes from someone else", Damaskos says. His goal was to: Make everyone a designer on the change management initiative we were trying to do. So, we got volunteers and promised them extra training, more hours, a change to their job title to include 'AI Champion' so everyone knew. Most importantly, I'd written a big AI strategy document for 2026 and beyond, which I threw away. Instead, we started from scratch and said 'alright, what do we need to do?' Every department of the company now has at least one champion, of which there were 15 in total. Their initial aim was to work out how AI could effectively be deployed across the business in a repeatable way to ensure it was embedded into ways of working. A personal AI assistant On the tools side, meanwhile, Mindstone released an alpha version of its Rebel AI agent orchestration platform in January this year, which it asked if Epignosis would like to test. The system, which was rolled out and onboarded between February and June, provides each employee with a personal AI assistant. Each personal assistant is connected to a team- or organization-wide file-based memory bank. This means a new workflow can be developed once, usually by an AI Champion, stored in either a team or company folder depending on its function, and shared with others. The assistant runs on each individual's local machine or device to ensure control over data privacy and storage. But it also integrates with third party systems, including LLMs, using application programming interfaces or Model Context Protocols. The upshot is that Rebel learns how each employee - and the wider company - operates. It connects to the applications they use, such as emails, team files, and meeting notes, and undertakes repetitive tasks on their behalf. For instance, the system will read incoming messages, highlight what is important, and draft replies for their approval. It will give them a quick meeting summary before they join and afterwards write down action points. When starting a new task, it will also provide the background information required to move forward, although training it to do so inevitably takes time due to initial hallucinations. Identifying possible use cases To understand how best to deploy the platform, the AI Champions were also asked to identify possible use cases for their specific departments. In a process that took six months, the use cases were then evaluated, the required data sources recorded, and users trained up. As Damaskos says: People are using the system in widely different ways: some for automation, some for meeting preparation. Sales and customer success are taking call transcripts and getting Rebel to update the CRM system, so they don't have to. We've also had 20% more demos per account exec since we started using it this year as the AI frees up more time. It's doing the prep and other boring work, so they have more time to book demos - although we can't know if it's 100% due to AI or other initiatives too. Nonetheless, it did take some time for employees to accept the system, Damaskos indicates: We installed the application for everyone, but initially no one dared use it as they didn't know what it was. So, we had meetings that were at least 30 minutes long to explain what this new technology was and why it was better than the tools they already had. Because we were doing it with the AI Champions, part of the onboarding was to give examples of what other people in their department were using it for. We also had bi-weekly showcases, where we demonstrated at least one cool AI use case to the whole company. One example here was automating childcare benefits. In the past, parents used to manually send their daycare receipts to payroll, which paid out a maximum of E300 per month. But the workflow automation took the requisite PDFs, compiled them into an expense report, and automatically sent them on, saving parents' time in the process. Higher quality work and better employee satisfaction For the five percent of employees who are still resisting the platform though, Damaskos recommends simply sitting down on a one-to-one basis and clarifying its benefits. As he points out: FOMO [fear of missing out] is a strong motivator. But he is also keen to point out that the key focus of using such technology is not about boosting productivity: It's a trap for you to count hours saved. People at the beginning were resistant. They were anxious AI was going to take their jobs. But when they saw the system working, they realized that, while it could take over some part of their role, it would never be able to take 100%. So, once people understood it was there to augment their work, they were happy and were proud to say this only took them five minutes as they knew how to use the AI correctly. From February to June, the system's return on investment dashboard estimates it has saved Epignosis 12,477 working hours. It has also created nearly 2,000 different documents, which include presentations and Excel spreadsheets. But Damaskos believes that a better measure of success than productivity is actually higher quality work and employee satisfaction. To work out such metrics, he tracks adoption levels, outputs, and how many additional systems users have connected to their personal assistant. This is because "the more tools that are connected, the better the results", he says. Users are also asked to record valuable use cases, which "aren't necessarily the highest impact but do illustrate what AI is best at, such as summarizing transcripts". Making work easier and more satisfying The next step, meanwhile, will be to migrate a selection of the organizations' departmental AI models to the cloud. Two out of a proposed 30 models have already been transferred. The idea is that users will not need to ask their own chatbot to prepare for a customer event, for instance. Instead, the information will already be held in the CRM system and "magically appear half an hour before the meeting", Damaskos says: We're trying to standardize things as change management is a long process that takes a lot of time to do. As for key considerations when introducing this kind of technology, there are several. Firstly, Damaskos warns that Rebel is memory-intensive if local processing is required for data privacy purposes. As a result, the company had to replace a lot of computers that were only three to four years old. Secondly, commercial LLMs are token-heavy, which is expensive for small companies. As a result, Epignosis switched to using open-source models, GLM, Kimi, and MiniMax. Finally, Damaskos says: People need to be encouraged to use AI. They need executive evangelism to know 'it's OK to use AI, we like AI, you should be proud if you're using AI'. At the same time, you also need a team that will give you ideas and help you resolve errors in the day-to-day things. So, you need permission from the top and support from the bottom. Our point is to try and automate as much as we can. It's part of our arsenal to make everyone's work easier and more satisfying. This is because when people are motivated, they perform better.

Your AI agent can be a teammate. But it still needs a boss | Fortune

Companies have spent the past two years teaching employees how to use AI. Most of that training has focused on the basics, such as prompting skills. It has helped people become comfortable with AI and begin to see its value. But that level of interaction is only the beginning, and it will not hold up as AI systems become more autonomous. The newest generation of AI systems is not simply waiting for instructions. Agents now pursue goals, call tools, make recommendations, trigger actions, and hand work from one system to another. The human role changes with them: the employee is no longer simply a user but the person responsible for directing and overseeing the agent. As Gianpaolo Barozzi, Cisco's 3P CTO, explains, "Agentic AI changes the relationship between people and technology. It requires new ways of setting boundaries, calibrating trust, and maintaining human accountability." Most companies are not training for that. They are teaching people how to get more out of AI. They are not teaching people how to lead it. This is becoming one of the most important capability gaps in the next phase of enterprise AI. Without that management discipline, employees may grant agents more trust and autonomy than the work or the technology warrants. The risk is not that agents will be obviously bad. The greater risk is that they will be useful, fast, fluent, and confident enough that people begin to relax their judgment at exactly the moment they need to sharpen it. In a 2025 MIT Sloan Management Review and BCG study, 76% of executives said they viewed agentic AI more as a coworker than as a tool. That language is directionally right. Agents will increasingly function like teammates. They will participate in work, shape decisions, coordinate tasks, and take on pieces of execution that used to belong only to people. But "teammate" is not a single relationship. Once an agent begins participating in the work, people need to understand what role it is playing and manage it accordingly. In our research, people frequently approached agents through one of five mental models: tool, intern, service provider, teammate, or expert. Each model creates different expectations about competence, autonomy, trust, supervision, and accountability. When an agent is treated like a tool, the human expects it to perform a bounded task on command. The person operates the system, evaluates the output, and remains responsible for the result. This works for narrow, repeatable tasks but becomes insufficient as the agent gains autonomy. An example of AI being used as a tool would include an employee using an agent to summarize a meeting transcript or reformat data into a standard report. When an agent is treated like an intern, the human provides context, inspects the work closely, corrects mistakes, and gradually expands its responsibilities. The agent may be capable, but it is still learning the organization, the work, and the standards required to perform well. For example, a manager might ask an agent to draft a client briefing, then review the work closely and coach it on what the organization considers important. Wharton professor Ethan Mollick has used the "AI intern" analogy to describe this relationship. Like a new employee, the AI needs a defined role, sufficient context, clear assignments, and ongoing evaluation of where it is, and is not, reliable. It can produce first drafts, conduct initial research, analyze data, or prepare a briefing, but the human must inspect the work and provide the judgment the agent lacks. When an agent is treated like a service provider, the human defines the desired outcome, scope of work, deliverables, performance standards, decision rights, constraints, and escalation requirements. The agent is given discretion over how to execute the work within those agreed parameters, while the organization maintains appropriate visibility, review points, and control. For example, a procurement agent might manage an end-to-end sourcing process against defined cost, quality, risk, and compliance targets, while escalating exceptions or consequential decisions for human approval. As Hala Jalwan, co-founder and CEO of Rivio.ai, a start-up that builds such procurement agents designed to operate as service providers by taking on work traditionally handled by outsourced and offshore teams, explains, "When an agent takes responsibility for a body of work, the human role becomes more managerial, not less important. Someone still has to set the objective, define the boundaries, determine when the agent should escalate, and remain accountable for the outcome." When an agent operates like a teammate or peer, the relationship is collaborative and ongoing. The agent helps develop ideas, coordinates work, responds to feedback, and participates in shared problem-solving. Its value comes not only from what it knows, but from how it works alongside others. Even so, the human remains accountable for the outcome. For example, an agent might participate throughout a product launch by developing options, tracking decisions, coordinating follow-ups, and adapting as the team's direction changes. BNY offers an enterprise example of what it takes to support this kind of human-to-agent relationship. As BNY CIO and Global Head of Engineering Leigh-Ann Russell explains, "We launched our first digital employee in 2025 with the same approach as our enterprise-wide systems -- governance was built in from the start. We understood that, over time, digital agents and people would work side by side. Digital employees are onboarded, governed, and continuously monitored with the same rigor we expect across our technology estate,, while accountability always remains with our people." The example illustrates that treating an agent as a colleague does not mean treating it as an equal bearer of responsibility. It requires clear oversight, performance expectations, and human ownership of the outcome. When an agent is treated like an expert, the relationship is more consultative. The human turns to it for specialized knowledge, analysis, or recommendations that may exceed their own capabilities. The primary risk is deference: because the agent appears authoritative, people may be less likely to question its conclusions. Its expertise should inform the decision, not own it. For example, a leader might ask an agent to analyze a complex dataset and recommend where the business faces the greatest operational risk. Research at Procter & Gamble illustrates how AI can play this expert role within a broader collaborative relationship. In a field experiment involving professionals working on product innovation challenges, AI helped employees generate solutions that extended beyond their own areas of specialization. Commercial professionals incorporated more technical thinking, while research and development professionals incorporated more commercial thinking. Less experienced participants were also able to perform more like teammates with deeper product-development expertise. In this setting, AI acted as a source of specialized knowledge that broadened the human's perspective but the employee still had to assess whether its recommendations fit the business context. Each model can be useful, and each creates different risks when applied carelessly. A person who sees an agent as an expert may fail to question it. Someone who views it as a teammate may assume it shares organizational context or responsibility. Someone who treats it as a service provider may delegate an outcome without maintaining sufficient visibility. And someone who treats it like an intern may waste time micromanaging work it can already perform reliably. Nor should an agent be placed permanently into one category. The same system may function like an expert when analyzing a large dataset, a service provider when executing a defined procurement workflow, a teammate when helping a group solve a problem, and an intern when navigating a new or ambiguous situation. The appropriate model depends on the task, context, risk, and the agent's demonstrated reliability. Calling an agent a teammate may recognize that it is participating in the work rather than merely producing outputs, but it does not tell us how much authority to grant it, how closely to supervise it, or how much confidence to place in its judgment. This is why companies need to build human-agent fluency: the ability to recognize which relationship the work requires and manage the agent accordingly. That is very different from prompt engineering. Prompt engineering asks, "How do I get a better answer from the machine?" Human-agent fluency asks, "What role should this agent play, and what does that role require of me?" As Tom Lamberty, Senior Consultant in Cisco's 3P Tech Office, puts it, "How we define an agent's role shapes how we work with it: how much authority we give it, how closely we question it, and where we retain judgment. We shape our agents, and then our agents shape us." That is also the better lesson from recent research on agent framing. Research by Boston University's Emma Wiles found that people caught 18% fewer errors when work was described as coming from an agentic "AI employee" rather than a chatbot. The finding should not lead companies to abandon teammate language altogether. It should lead them to take that language more seriously. When agents function like teammates, they require clear standards, challenge, escalation, and ownership. A teammate relationship without appropriate oversight is not teaming. It is overtrust. Organizations should therefore make the expectations associated with each model explicit: what the agent may decide, when it may act independently, how its work will be reviewed, and what would justify changing the relationship. Employees must also learn to recognize when they are granting an agent too much, or too little, trust. That is the difference between AI usage and agentic readiness. A workforce is not ready simply because employees use AI frequently. It is ready when people can identify the relationship the work requires, manage it appropriately, and recalibrate it as the task, risk, and agent capability evolve. Across every model, human accountability for the outcome remains constant. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Embedding intelligence at the coalface of work

And while companies are still facing ongoing economic uncertainty, rising customer expectations and rapid technological change, the question boards are asking of the business isn't "How many hours AI can save?" it's "How much better the business can perform?" Angela Colantuono, president and managing director, SAP Australia and New Zealand (ANZ), says creating value has become the defining business issue. "Six months ago, everyone wanted to know how many hours AI could save. Today, the question is how much value it can create," says Colantuono. "The conversations everyone's having around AI, the most important thing is, how do you do it really well? It's not just a chatbot that you're putting in front of your existing systems. It's a much more strategic, holistic view." For many organisations, that means moving beyond experimentation and embedding AI across end-to-end business processes rather than introducing another application for employees to learn. The gap between organisations that have done this and those still running pilots is widening and leaders know it, she says. Simplify before you automate While AI attracts much of the attention, Colantuono says data quality and process simplification remain among the biggest barriers to improving business performance. "What we know is organisations really struggle with getting access to data," says Colantuono. "When data is not accurate, the results are not where they need to be, from AI or any other way of doing business." She says many organisations have accumulated increasingly complex technology environments over time, making it harder to connect information across finance, human resources, procurement and operations. Rather than adding another layer of technology, businesses are simplifying end-to-end processes before embedding AI within them. "When you simplify the process, then technology can really deliver and the organisations getting this right aren't just becoming more efficient, they're becoming more capable of delivering outcomes that matter," says Colantuono. The same thinking is reshaping technology investment. Colantuono says many organisations are trying to invert what she describes as the 80/20 technology burden, where most technology effort and funding is spent maintaining fragmented systems and integrations rather than improving business outcomes. "What we've seen over many years is organisations go with a best-of-breed approach in solving technology decisions," she says. "What that's meant is there's been a considerable amount of money in integration." Simplifying technology environments creates a stronger foundation for trusted data and allows AI to operate within existing workflows instead of becoming another disconnected tool. "When you use AI in the flow of work, it becomes part of how the business operates." AI in practice The shift from experimentation to embedded AI is already becoming visible in customer-facing organisations. Furniture retailer Freedom has significantly expanded its online product range in recent years, increasing its catalogue from about 15,000 products to more than 70,000 as part of its strategy to offer customers an "endless aisle" experience. The expanded range created a new challenge. "As the catalogue rapidly scaled from 15,000 to over 70,000 products, identifying the right item became increasingly complex for customers," says Federico Jalil, digital product manager at Freedom. Rather than treating AI as a standalone initiative, Freedom integrated AI-driven search and personalisation directly into the customer journey. By analysing customer intent through search queries, browsing behaviour, product interactions, add-to-cart activity and previous purchases, the retailer sought to make product discovery more relevant. The rollout produced immediate results. "We observed an immediate and measurable shift in customer behaviour on our website. This included a notable increase in search bar utilisation, alongside significantly higher conversion rates among customers who incorporated search into their browsing journey." The experience also reinforced another lesson emerging across Australian organisations: AI depends on reliable data. "Trusted customer and product data is fundamental to delivering an effective AI experience," Jalil says. "As the well-known principle states, 'garbage in, garbage out', regardless of how advanced the AI model may be, poor-quality data will inevitably lead to poor-quality outcomes." He says success depends less on collecting more information than ensuring existing information is accurate, properly indexed and structured appropriately. "We have also learned that success is not driven purely by data volume, but by ensuring the right data is properly indexed and structured, minimising noise and enabling more accurate and meaningful outputs from LLMs." AI becomes a boardroom issue AI is becoming a boardroom issue rather than simply an IT initiative. Colantuono says conversations with directors at the Australian Institute of Company Directors highlighted that business leaders are asking how AI can create competitive advantage while ensuring organisations maintain appropriate levels of trust, ethics, compliance and security. "They have to balance opportunity with accountability." "They have to ensure that investment decisions, risk frameworks and governance structures keep pace with the speed of change, but also enables organisations to scale AI confidently." She says SAP approaches AI through three principles. "We want to make sure that it's responsible, we want to make sure it's reliable and we want to make sure it's relevant." Those discussions mirror conversations taking place with chief executives and chief information officers, where attention is shifting from deploying AI quickly to embedding it responsibly within core business operations. Colantuono says many of the same productivity questions are emerging across government agencies, where organisations are seeking productivity improvements that can enhance frontline service delivery while maintaining governance and accountability. From experimentation to execution As AI capabilities continue to evolve, both Colantuono and Jalil point to a similar conclusion. Experimentation remains important, but lasting value comes from integrating AI into core business activities rather than treating it as a separate project. "At both a personal and professional level, I'm currently exploring the evolving landscape of AI platforms," Jalil says. "Early interactions with these tools quickly highlight their capability, however, their true value becomes evident when they are embedded into core workflows or business functionality." For organisations still early in their AI journey, he says the first step is identifying a genuine business problem. "Start small and scale from there, identify a clear challenge or friction point, and focus on opportunities that can be automated and delivered at scale," Jalil says. Across government, retail and enterprise, Australian organisations are confronting many of the same questions around value, service delivery and responsible AI adoption, while being asked to achieve more with less. The next frontier is not simply embedding AI into the way we work but enabling organisations to redesign work so people, AI and business processes operate together to create entirely new levels of speed, agility and value. Organisations that combine trusted data, simplified processes and governed AI won't simply automate existing work. They'll build the foundations of the autonomous enterprise, where trusted data, intelligent applications and AI agents work together to help organisations anticipate, decide and increasingly act with confidence under human oversight.

Most AI Fails Without Human Workflow Design

Every week, I talk to business leaders frustrated with AI. They bought tools, ran pilots, hired consultants, and still aren't seeing the promised results. My response typically surprises them. Most companies don't have an AI problem. They have a workflow design problem. The real failure mode When AI underperforms, the instinct is to blame the model, the vendor, or the implementation team. What I've observed across hundreds of client engagements is that failure lives in the handoffs. Somewhere between AI generating an output and a human acting on it, accountability disappears. The system produces a document or a customer response, and the assumption is that production equals completion. That assumption is where things break down. Most organizations layer AI onto fragile processes, but AI doesn't fix fragility. It only augments what's already there. So, you get more output, faster, and with it, more errors moving through the system, which increases risk. What good workflow design looks like Redesigning for AI should be viewed as an organizational design project that happens to involve technology, as opposed to a purely IT project. Those who view it as the latter typically start by asking, "What can we automate?" To succeed, you need to ask, "Where does human judgment create irreplaceable value, and how do we structure the work around that?" Those are very different questions. In customer experience workflows, AI handles volume effectively. Tier-one inquiries, ticket routing, sentiment flagging, and knowledge base retrieval are well-suited to automation. Exceptions are where AI consistently falls short. These are the customers whose situations don't match templates or the complaints with regulatory risk. This is where judgment matters, and that's the domain of human teams. Workflow design must be built around that reality. The need for oversight should change how you staff Earlier this year, Connext published its 2026 AI Oversight Report. One data point that has stayed with me is that 28 percent of users report that AI still requires active supervision to produce reliable outputs. While some may read that figure as a limitation, I see it as a job description. Oversight is no longer a remediation task you assign when something goes wrong. For a growing share of AI-dependent workflows, supervision is the work. The role is to continuously guide, correct, and improve what AI produces. Many organizations are already developing the human capability that makes AI trustworthy. In the long run, their efforts will result in durable operational advantages. A case that illustrates the difference One of the clearest examples is our work supporting a fast-scaling AI-driven digital health startup. The company built strong AI capabilities for patient engagement and administrative workflows, and they were growing quickly. The challenge was that growth was outpacing their ability to maintain quality and compliance as volumes increased. AI alone couldn't solve the problem. They needed a hybrid model where trained human teams who understood the AI's logic and its failure modes operated as embedded extensions of the business. That's often what most outsourcing relationships miss. Traditional task-based outsourcing hands off defined, repeatable work to an external team. That model has value in stable environments, but it breaks down quickly in AI-augmented operations, because the work isn't stable. The oversight team doing the work must understand the system they're overseeing, not just the task. What made the digital health example work was that the teams were trained to correct, guide, and improve AI outputs with shared accountability for outcomes. The new roles AI is creating One of the prevailing narratives around AI is that it eliminates jobs. The operational reality is more nuanced, and more interesting. AI is eliminating certain categories of repetitive, rule-based work. At the same time, it's creating demand for roles that didn't exist five years ago. Today's work needs AI trainers who improve model performance through structured feedback, QA reviewers who specialize in identifying where AI outputs drift from acceptable standards, and workflow orchestrators who manage the handoffs between automated systems and human decision-makers. These roles are becoming permanent features of high-performing operations. The competitive edge in AI-augmented operations comes down to how well your team knows when the model is wrong, and what to do about it. Why embedded teams outperform vendor models I've spent over a decade building Connext around the belief that offshore teams perform at their best when they operate as genuine extensions of the client's business, with shared context and accountability, as well as continuous feedback loops embedded into how the work gets done. That belief has become more consequential in an AI-augmented environment. The feedback loop is the mechanism by which AI improves. If the team doing oversight is operating in a silo, the feedback never reaches the system. They catch errors, but they don't correct them at the source. High-performing global teams succeed because they're close to the work and understand why a workflow is designed the way it is. They can identify when AI outputs don't match business intent, not just when they fall outside a defined rule. That kind of judgment must be cultivated through genuine integration. The traditional outsourcing model, with its emphasis on task completion and cost-per-transaction pricing, was built for a different era. In an AI-driven operation, it actively undermines the feedback mechanisms that make AI reliable. The leadership question worth asking If I were advising a CEO preparing to scale AI, I'd start by asking if they've designed the human layer of this system with the same rigor they applied to the technology layer. Most haven't taken that step. The technology selection process is thorough, deliberate, and well-resourced. But the workflow design process often receives a fraction of that attention. The real value is in doing the hard work of designing workflows where humans and AI each do what they're best at, with clear handoffs, accountability, and organizational muscle to keep improving over time. Get 1 Smart Business Story delivered straight to your inbox when you subscribe to Inc.'s free daily newsletter.

How AI turned your best work into the bare minimum | Fortune

A project that used to take three weeks now takes one. A report that required a full day gets done before lunch. And employees are being evaluated against that accelerated standard before anyone has agreed it is sustainable, accurate, or fair. The time savings went to the company, but the pressure went to the employee. Here is the dynamic playing out inside nearly every organization adopting AI right now: efficiency gains from AI are not being returned to employees as breathing room. They are being immediately converted into higher output expectations. The assumption is simple and rarely stated out loud: if AI freed up your afternoon, that afternoon now belongs to the next assignment. Workers are not getting time back. They are getting more work, on a faster clock, with the same number of hours in the day. The Research Is Consistent: AI Is Intensifying Work, Not Reducing It New research from GoTo and Workplace Intelligence makes the paradox explicit. The Pulse of Work in 2026 study, which surveyed 2,500 employees and IT decision-makers across ten countries, found that employees save more than two hours per day using AI tools. But the same study found that 60% of employees feel pressured to use AI to boost productivity, 50% say they rely on it too much, and 39% say that reliance is making them less intelligent. The productivity gain and the human performance cost are arriving simultaneously. Most organizations are only tracking one of them. An ActivTrak's analysis of 443 million hours of work activity across more than 1,100 organizations found that AI doubled time spent on email and messaging while focused deep work fell by 9%. A Harvard Business Review study published earlier this year found that after AI adoption, workers operated at a faster pace, took on a broader scope of tasks, and extended work into more hours of the day, often without being asked. The tools are generating more activity while depleting the capacity for the high-quality thinking that makes that activity valuable. Evaluated by Speed, Not by Judgment When AI compresses timelines, the most visible change is speed, and speed quickly becomes the proxy for performance because it is the most legible output of AI adoption. Gallup's research finds that 65% of employees say AI has improved their productivity, and frequent AI use among managers has doubled from 15% to 30% since 2023. But that growing adoption is producing a widening gap between who benefits and who absorbs the pressure: leaders report the strongest gains, while individual contributors remain the least likely to receive guidance on how to use AI effectively. And 54% of managers say workplace expectations have directly increased due to AI. The bar is rising and the measurement framework is not keeping pace. What speed-based evaluation misses is the cognitive work AI cannot do: evaluating outputs for accuracy, catching hallucinations before they become decisions, and applying contextual judgment the model lacks. Reviewing AI output for errors and making the 80-percent-done draft actually good is not drudgery you can do on autopilot. That is executive-level judgment running in the background all day. When employees are evaluated primarily on how fast they produce, that invisible cognitive labor goes unrecognized and eventually exhausted. The downstream consequence is what researchers are calling workslop: fast-output, low-value work that floods organizations when speed is the only thing being measured. The GoTo and Workplace Intelligence study found that 43% of employees have used AI-generated content despite suspecting it was low quality or contained errors, and 77% say AI-generated work takes more time to review than human work. As Built In reported, when managers reward accelerated output above all else, employees default to quantity over quality, and those on the receiving end spend extra time fixing what AI produced. Faster output, in practice, often means slower net progress. The Cycle Nobody Is Interrupting What makes this dynamic so difficult to address is that it compounds quietly. The GoTo and Workplace Intelligence study found that 65% of employees say employers are failing to equip them with the skills they need as AI takes over more work, and 80% say most workers are not being trained properly to use AI tools. Yet nearly one in four IT leaders say AI mistakes have already affected customers or their company's bottom line. Organizations are accelerating adoption while underinvesting in the human infrastructure required to sustain it. Leaders reporting AI-enabled productivity gains to boards are not typically reporting the simultaneous increase in cognitive load, the decline in focused work time, or the erosion of judgment quality those gains depend on. The metrics that look good in a presentation get elevated. The metrics that would reveal the human cost have not been built yet. What Leaders Need to Do Differently The first intervention is the simplest and rarest: actually, returning time to employees. When AI compresses a three-week project to one week, the default response is to assign two more. The alternative is to use recaptured capacity for the work AI cannot do: deeper relationships, more rigorous quality review, skill development, and the kind of unhurried judgment that produces durable rather than fast results. Organizations that treat every efficiency gain as an invitation to pile on more work will find they have built a faster hamster wheel, not a more capable workforce. The second is redesigning how performance is measured. The employees generating the most output are not necessarily producing the most value. The ones catching AI mistakes before they become decisions, applying contextual judgment that makes outputs usable, and maintaining quality while others optimize for volume are the ones organizations most need to identify and retain. That requires measuring impact rather than throughput, a harder problem with significantly more valuable answers. The third is honest expectation-setting at every level. Most employees did not sign up to manage an AI system on top of their existing job. They signed up to do their job better. When organizations add AI tools without adjusting workloads, timelines, or success metrics, they are not empowering their workforce. They are quietly redefining what enough looks like, without asking whether anyone can sustain it. That conversation belongs in the open, not buried inside a productivity dashboard. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

The Truth About AI That Every Business Leader Needs to Hear Right Now

You should not fear AI, nor should you blindly trust it. You must manage it with a strict framework. Treat every conversation as highly disposable, give it permission to fail, and isolate your projects. For the past couple of years, the business community has been trapped in a state of whipped-up hysteria. Every headline screams a variation of the same extreme narrative: Either artificial intelligence is an existential threat poised to wipe out entire industries overnight, or it is a god-like entity that will double your corporate productivity by next quarter. Like most business owners, I got tired of the theoretical debates. I bypassed the standard consumer chat apps and logged directly into the advanced, backend developer playgrounds where these systems are built. I didn't just ask them to write a generic email; I fed them complex corporate files, intricate programming code and heavy datasets. I wanted to see the future of business management. Instead, I found a deeply flawed, incredibly erratic reality. What I discovered through weeks of hands-on testing exposes a massive gap between the marketing promises of Silicon Valley and the actual behavior of the hardware. If you are currently allocating capital, shifting budgets or restructuring teams based on the assumption that AI is an independent, thinking brain, you are walking into an expensive trap. The Great Illusion: Sorting vs. thinking When you first interact with a top-tier language model, the experience is intentionally designed to mimic human thought. It responds politely, adapts its tone to your personality and formats data beautifully. But as I pushed these systems with real-world business management and analysis tasks, the illusion shattered. I realized that what we are calling "intelligence" is actually a highly polished, superpowered indexing and categorization tool. AI does not possess logic, strategy or a mental model of cause and effect. It cannot look at a failing department and understand the human tension in the room, shifting market sentiment or hidden corporate priorities. What it can do, with incredible speed, is match patterns. This is incredibly useful, but it is not a thinker. It is an intern. When you ask it to behave like a corporate strategist -- such as designing a five-year growth plan -- it breaks down. Because it cannot say "I don't know," it relies on mathematical probabilities to generate what looks like a plausible answer. It blindly guesses what you want to hear, smoothing over contradictions and inventing data just to be agreeable. The "exhaustion" phenomenon The most shocking revelation from my testing occurred during prolonged, complex sessions. When you start a brand-new conversation with a clean slate, the AI acts sharp, accurate and incredibly capable. But as the conversation goes on -- after several prompts -- something bizarre happens. The AI begins to lose the plot. It confuses different files you uploaded, forgets instructions you gave it ten minutes prior and begins generating repetitive, low-quality nonsense. In the tech world, this is a known phenomenon, but business owners are completely blind to it. Every time you send a message, the system has to reread the entire conversation history from scratch. As that history grows, the machine's focus dilutes. It gets distracted by its own previous words. Mistakes snowball, and the output rapidly degrades into what can only be described as complete garbage. Worse, I discovered that simply hitting "delete chat" or opening a new tab doesn't fix it. The backend infrastructure tracking your account remembers your heavy usage. When you strain the system, tech companies quietly put your account on a low-priority, compressed tier to save on their massive server electricity costs. The system essentially handicaps itself without warning you. The only way to get the "smart" version back is to literally close the laptop, walk away and wait hours for the server clocks to reset your account. The valuations shield: Why the panic is fostered This hands-on reality forced me to look at the macroeconomic picture. Right now, tech giants are spending hundreds of billions of dollars building massive data centers. Yet, financial reports show that the vast majority of companies investing in these tools have seen zero measurable return on investment. The business model is fundamentally broken because running these massive mathematical computations costs an astronomical amount of electricity and hardware power. So, why the constant public panic about AI taking over the world? When you look closely, the extreme hype and fear start to look less like genuine concern and more like a coordinated corporate survival strategy. Tech executives go to Congress and beg for regulation because "AI is too dangerous and powerful." They tell investors that artificial general intelligence is just around the corner. Why? Because no venture capitalist or Wall Street investor is going to pump trillions of dollars into a tool that is merely an expensive, power-hungry data-sorting machine. But they will pour trillions into a digital god. The manufactured panic keeps valuations inflated. It buys tech companies a three-to-five-year runway, hoping their hardware engineers can figure out how to make chips cheaper before the investor capital completely dries up. It is a massive smoke-and-mirrors show designed to postpone a market correction. The playbook for entrepreneurs As an entrepreneur, you should not fear this technology, nor should you blindly trust it. You must manage it with a strict, sober framework: The prompt limit: Treat every conversation as highly disposable. Never let a chat drag on indefinitely. Run a few prompts, extract the clean data, wipe the memory completely, and start fresh. Give permission to fail: Stop letting the machine guess. Explicitly write into your instructions: "If the data is missing from the uploaded file, do not infer or guess. Stop immediately and state that you do not know." Isolate your projects: If the system starts giving you low-quality outputs, do not waste time arguing with it. Create an entirely separate workspace project on the developer platform to break the server-side memory lock and force a true cold start. The future does not belong to the companies that replace their humans with AI. It belongs to the practical entrepreneurs who see through the Silicon Valley mythology, treat AI as a superpowered filing clerk and ruthlessly audit every single line it outputs.

We Didn't Use AI to Replace Our Team

Artificial intelligence (AI) is becoming one of the most overhyped topics in business. Every platform promises transformation, every tool claims to save time, and every company seems to be racing to integrate AI. In my experience, the biggest gains come from solving small, repeatable problems that slow teams down. We've intentionally kept our team lean, which has made AI especially valuable. Instead of using AI to reduce headcount, we use it to remove friction, eliminate repetitive work, and free up our team to focus on work that requires human judgment, creativity, and expertise. AI helps our specialized team spend less time on tedious processes and more time on meaningful work that moves the business and customers forward. This is the real opportunity for small businesses. AI doesn't replace strong teams; it helps already strong teams operate faster, smarter, and more efficiently. These are four ways we've seen AI accomplish those outcomes. 1. Expand capability The biggest gains we've seen from AI have come from applying it to time-intensive, repetitive processes that don't require deep creativity or strategic thinking. By removing these bottlenecks, AI allows our team to prioritize high impact work that benefits most from human expertise. 2. Translate complexity In SaaS, one of my favorite uses of AI has nothing to do with writing code -- it's understanding it. As a leader, I often need to understand exactly what a piece of code does, but interrupting developers to explain technical details pulls them away from deep focus work. AI can translate complex code into plain English faster and often more clearly, helping me make faster, more informed decisions while protecting my developers' productivity. AI doesn't always need to be the builder. Sometimes its greatest value is making specialized or technical information more accessible across teams, departments, and leadership levels. 3. Turn data into customer-facing content We work with rental property owners, and it used to take customers 30 minutes to write compelling property listings, sometimes with help from our support team. We built an AI tool that compiles existing property data to generate a professional listing in about 30 seconds. Customers move faster, support volume decreases, and our team can focus on higher-value customer needs. This is where AI works best: structuring and refining existing information into a format that's faster, clearer, and more useful. This same approach applies anywhere businesses turn data into customer-facing communications -- product descriptions, reports, proposals, and more. 4. Automate repetitive work Testing and quality assurance are essential parts of software development, but they're also some of the most repetitive and time-consuming tasks. We've started using AI for initial code testing before human review, providing instant feedback and identifying potential issues earlier in the development process. The result is faster development, fewer repetitive tasks, and more time for developers to focus on the creative problem-solving they enjoy. In any industry, the best AI applications often remove the repetitive, time-consuming, but low-value tasks that prevent employees from focusing on more meaningful contributions. AI helps small teams do more without adding complexity One of the biggest advantages of AI for small businesses is the ability to expand capacity and scale output without adding unnecessary layers of complexity. Work that once required additional management, larger teams, or significant manual effort can now be streamlined with the right tools and workflows. We've seen the results firsthand by using AI to improve how customers access information. Our AI-powered knowledge base tool instantly summarizes information from our internal resources, resolving 82 percent of customer lookups that previously would have become support requests. Customers get faster answers, while our support team is free to handle the complex cases that require human judgment and problem-solving -- expanding our capacity without adding more support staff. AI doesn't eliminate the need for talented people -- it increases what talented people can accomplish. A smaller team no longer automatically means limited capability or slower execution. AI gives us the ability to grow capabilities and maintain the output of a much larger organization without creating unnecessary operational overhead. AI requires human judgment AI has tremendous potential but isn't plug-and-play or one-size-fits-all. Many companies overestimate AI's capabilities or underestimate the tradeoffs. These factors are critical to consider. This is particularly important in industries like property management, fintech, healthcare, or any business handling sensitive data, where small mistakes can have major consequences. Maintaining a security-first mindset with clear permissions, oversight, and guardrails is critical. The future belongs to businesses that use AI intentionally The best AI strategies start with a problem. Find the repetitive tasks slowing your team down. Identify where employees are spending the most time on work that doesn't require their highest-value skills. Then use AI intentionally to remove those barriers. For small businesses, this represents a rare opportunity. You don't need the largest team or biggest budget to compete. You need to understand where your people spend their time and use technology to help them spend more of it on work that moves the needle. Get 1 Smart Business Story delivered straight to your inbox when you subscribe to Inc.'s free daily newsletter.

Stop Measuring AI Adoption. The Capability Gap Inside Your Team Is the Real Reason You Are Falling Behind.

The instinct to standardize AI enablement through committees and best practices breaks down at this pace of change -- what actually works is protected, facilitated time (three hours minimum, no multitasking) where every operational group experiments in the context of their own work and shares what they learned. AI is not rolling out in a smooth curve. It is moving in step changes, where what felt like strong performance a few months ago quietly becomes the baseline, often without any clear signal that the bar has moved. And it has moved again since I started writing this article. The pace of this shift is catching most teams off guard. Just in the last few weeks, we've gone from AI tools that help with individual tasks to systems that can operate autonomously across your entire computer. Agentic platforms like OpenAI's Operator, Perplexity's computer use, OpenClaw and Anthropic's Claude Cowork are now executing complex workflows from start to finish. The implications for how we think about work are immediate and hard to overstate. Many tasks that used to take hours can now be done in minutes. A workflow that requires coordination across teams can be handled by one person. The volume of AI innovation shipped in March alone outpaced anything we've seen before, and Anthropic is scaling revenue at a pace no company has hit before. Today, a lot of the conversation still focuses on adoption: who is using AI, who is not and how quickly teams are rolling it out. But that framing misses what is actually happening. Capability is no longer evenly distributed Some people are fully engaged with these tools. They are testing ideas, building custom GPTs, creating skills in Claude and developing real instinct for where AI actually helps. Others use it more cautiously, keeping it at the edges of their work. There is also a third group emerging. These are the people going much deeper, using open-source tools, pushing on the edges of what is possible and running "scary" experiments -- Zuckerberg building an AI version of himself, for example -- sometimes faster than the organization can keep up with. All three groups are using AI, but they are not operating at the same level. That difference compounds quickly by changing how work gets done, how fast it moves and how people experience their roles day to day. The work itself is changing For years, most roles were built around execution. The focus was on delivering the work -- following the process and getting the output across the line. AI is now compressing much of that layer. The work is shifting toward evaluating, connecting and deciding what matters. You can get to a first version quickly now, which changes where people spend their time and how they approach the work. In some cases, it goes further. You can begin to build systems that reflect how you think, how you write and how you approach problems. That starts to change how individuals operate, not just how fast they move. But most organizations are still structured around the previous version of the job. That raises broader questions about how organizations are structured and what roles are actually needed, which I'll come back to in a future article. What actually helps teams keep up When companies think about AI enablement, the instinct is to standardize quickly -- creating rules, forming committees, defining best practices and rolling out a consistent approach. That approach breaks down. It assumes you can define how new capabilities should be used and then push them into the organization. In practice, that model no longer holds up. A more effective approach is to build structured time into how the company operates so every operational group can engage with these tools directly and regularly, within the context of their own work. The premise is simple: people know their own jobs best. At our company, we started with "AI Days" -- a small group would step away from their day-to-day to focus on their own work and explore how AI could drive efficiency or create new value for the business. Over time, we realized everyone should be able to participate, so we introduced more frequent sessions we call "AI Fridays." These are small, facilitated groups, usually no more than ten people, with an AI implementation expert in the room the entire time. People are expected to commit fully. No multitasking, no checking email. We block off a minimum of three hours because anything less isn't enough to engage meaningfully. We close with a short, structured share-out: What did you try? What worked? What didn't? During these share-outs, you start to see how differently people approach the same tools. People from different operational groups are working toward the same goal, but with very different backgrounds and ways of thinking, and that tends to produce outcomes you would not get otherwise. Simply seeing how others approach their work can be surprisingly inspiring and starts to shift how you think about your own. It also helps normalize failure in a very real way. This work takes time, and people do get stuck or go down paths that do not lead anywhere useful right away. When that happens in a group setting, it becomes part of the process rather than something to avoid. Seeing others push through it, or even laugh it off, makes it easier to stay engaged instead of stepping back too early. There is also a broader shift happening. Most people are used to tools like Google, where every interaction starts fresh. AI builds context over time and responds differently as you continue working with it. Getting comfortable with that iterative loop -- where you refine, adjust and build on what came before -- takes practice. Over time, these sessions create a rhythm where people update their sense of what is possible and get comfortable with how quickly things change. It's common to spend hours building something and then see a new capability replace part of it not long after, and that becomes part of how the work evolves. Where leadership makes the difference Leading in this environment means paying attention to how quickly expectations are shifting, including your own. I've seen this happen inside our own business. You spend time with the tools and start to see the work differently. The issue is when that shift stays in your head. The team ends up trying to catch up to a standard they cannot see, and that is where the disconnect starts. There is still a baseline expectation that does not change. If something is being shared or used to make a decision, it needs to be understood and owned. That applies regardless of how it was produced. There are a few behaviors that tend to make the difference: * Own it. If expectations are shifting, leadership needs to take responsibility for driving that change. * Create real space for experimentation. This work requires time and focus, not something squeezed in between other priorities. * Set clear, meaningful milestones. The goals should be ambitious enough to push teams to actually change how they work. * Tie outcomes to incentives. People need to see that leaning into this shift leads to real upside, not just more work. There's a shift that comes with this. Something you spent years learning can now be done differently, and sometimes faster, than you would have done it yourself. That takes a bit of getting used to. You work through it, figure out what works and what does not, then push that back into the team. That's when it starts to change. People pick it up, build on it and take it further than you would have on your own. That's where it starts to get interesting, and where it actually becomes pretty exciting.

Your AI strategy isn't failing because of bad design. It's because your team doesn't believe in you | Fortune

Most AI strategies fail in the same place. Not in the design phase, not in the technology selection, and not in the rollout plan. They fail the moment a leader stands in front of their team and realizes -- sometimes immediately, sometimes months later -- that the people nodding in the room have already decided not to follow. That is the belief gap. And until you close it, no strategy survives contact with your organization. John Chen understood this the hard way. When he walked into BlackBerry as CEO in November 2013, the company had gone from controlling roughly half the U.S. smartphone market to low single digits in just a few years. He moved quickly on the obvious stuff: write-downs, workforce cuts, a Foxconn manufacturing partnership, and a decision to focus on enterprise customers rather than chase the mass market. Nobody accused him of moving slowly. But the problem that consumed him had nothing to do with product or technology. People had stopped believing it was safe to tell the truth. Staff were burying bad news for fear of getting blamed. They weren't challenging decisions. He understood that when people stop speaking honestly, every strategy built on top of that becomes a house of cards. In a 2025 Deep Purpose podcast interview with Ranjay Gulati, Chen explained: "Your people could handle a lot of things if they believe that you treat them fairly." He didn't choose between strategy and belief. He insisted that one was the foundation of the other. Most leaders get that sequencing backward. They build the strategy first and assume belief will follow. It won't. The Real Reason Your AI Rollout Is Stalling Here is what an AI strategy failure actually looks like from the inside. The meeting goes well. You get plenty of nods. Everyone understands the plan. Yet nothing changes. The temptation is to diagnose this as a communication problem -- more data, clearer messaging, a better slide deck. That diagnosis is almost always wrong. After twenty-five years working with leaders at companies like Google, Pfizer, JPMorgan Chase, and Morgan Stanley, and after studying belief in places like North Korea and Rwanda -- where what people believe is literally a matter of survival -- I've arrived at a different conclusion. The greatest AI strategy in the world will not save you when your people don't believe in it. In its State of the Global Workplace 2026 report, Gallup found that nearly one in five U.S. employees said it's at least somewhat likely AI will eliminate their job in the next five years. You are not asking your team to adopt a productivity tool. You are asking them to trust you with their sense of security, their professional identity, and their sense of what the future holds for them. Compliance, under those conditions, is the floor -- not the goal. The most important skill a leader can develop is not communication or motivation, but understanding how conviction actually forms in another person's mind. Ted Lasso, the fictional AFC Richmond soccer manager from the Apple TV+ series, described belief in terms that would embarrass most corporate training programs: "Belief doesn't just happen because you hang something up on a wall . . . It comes from in here [heart] ... And up here [brain]. Down here [gut] ... Only problem is we all got so much junk floating through us, a lot of times we end up getting in our own way." What Ted calls "junk," I call Inner Propaganda -- the accumulated assumptions, shortcuts, and self-protective conclusions your brain has already made before you walk into any meeting. As a leader, your job is to understand how that works in yourself and in the people you're trying to move. Feeling Comes Before Logic -- Every Time The instinct is to use data to change minds. But that isn't how we're convinced. Motivated Reasoning -- a term coined by psychologist Ziva Kunda -- describes our tendency to use reason to either find what is true or to justify what we want to be true. Most of the time, people have already decided what they want to believe, and they use their reasoning to build the case for it. Logic arrives after Feeling and Identity have already made dinner, eaten, and done the dishes. This is why your AI business case isn't moving your team. It isn't that the case is weak. It's that you're presenting it to people whose emotional conclusion -- this is a threat or this is hype -- was formed long before you started talking. A 2025 study by Yang Woon Chung and colleagues, published in the Australian Journal of Psychology, found that employees resist AI adoption because they fear job loss, mistrust AI-driven decisions, and feel excluded from the change. Not one of those barriers is solved by a stronger ROI argument. Years ago, I was called in by a large financial institution in Dublin that was six months into an organization-wide transformation following the 2008 financial crisis. It wasn't working. My first question was simple: where is it actually working? Two departments, they said. Both had introduced a Friday lunchtime town hall -- a weekly hour where people could talk through changes coming down the pipeline before and during the transformation. Here's what was striking: employees couldn't actually influence the decisions. The change was fixed. And yet both departments were significantly outperforming the rest of the organization in adaptation. Why? One manager put it simply: it made people feel the change was up to them. They felt autonomy, even though they had little real opportunity to influence outcomes. When the practice spread across the institution, resistance dropped measurably within two months. Feeling preceded follow-through. Consider the difference between a leader who tells their team, "AI adoption is a strategic priority for Q3," and one who opens with: "I spoke to a financial analyst who spent four hours every Monday morning pulling data for a report nobody read. AI now does it in four minutes. She spends those four hours on the work that actually got her into finance in the first place." The second leader is not making a better argument. They are creating a different feeling first. Identity Is Why People Really Resist If feeling is the gateway, identity is the locked door behind it. When an AI rollout asks someone to become a different kind of professional, it isn't asking them to learn a tool. It's asking them to revise who they are. Most people won't do that without help. Years ago, I worked with a woman named Maria who had tried unsuccessfully many times to quit smoking. "I'm just a smoker," she said. I asked how many cigarettes she smoked a day. Twenty. About three to four minutes each. "You spend around eighty minutes a day smoking and twenty-two hours and forty minutes not smoking. If someone had two jobs, which job would you say was theirs -- the 80-minute one or the 22-hour one?" She paused. "The one they do for 22 hours and 40 minutes." "So you're actually a non-smoker who smokes for 80 minutes a day." "I never thought about it like that before." Years later, she told me she hadn't touched a cigarette since that conversation. The change in identity label freed her to make the adjustment. She already was the person she wanted to be. She just needed help seeing it. This is your job with AI adoption. Your team members who don't see themselves as technical need to understand that being skilled with AI in 2026 doesn't require technical skills -- it requires precision, clarity of intent, and the ability to iterate and refine. Those are skills they already have. The coder who understands what good output looks like is far more valuable than someone who can only write the code itself. Their identity doesn't need replacing. It needs updating. Reason Is the Last Step, Not the First Plato argued that reason was the key to getting out of the cave -- to seeing the world as it actually is rather than as shadows on a wall. He was partly right. But reason isn't just a tool for finding truth. It's also a tool for proving it. There are two modes: investigation, which uses logic to figure out what's actually true; and argumentation, which uses logic to prove a conclusion already reached. Most of the time, when you're presenting your AI strategy, your team is in argumentation mode -- using their reasoning to defend what they already feel and already believe about themselves. Ensure your message feels right and fits who they believe they are before you make the rational case. In that order. Every time. John Chen stepped down in 2023, with the BlackBerry board explicitly crediting him with "saving the company and repositioning it as a software company." He succeeded because he understood that conviction is built from the inside out -- feeling first, identity second, reason last. The next time you stand in front of your team to present your AI strategy, the goal is not a roomful of nods. It should be the excitement, the urgency, and the certainty that people carry out of the room with them. You'll know the difference not just in what they do next, but in how they do it. That's leading at the level of belief rather than information. Because your AI strategy doesn't have a design problem. It has a belief problem. And that problem starts -- and ends -- with you. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Of Course, Companies Are Investing in AI

Ask a business leader today whether their company is using artificial intelligence, and the answer is almost certainly yes. The budgets have been approved and the tools are already in use. Ask the harder question, whether that investment has changed how the company makes decisions, and the confidence tends to dissolve. That hesitation is the real state of AI in most companies right now. The debate over whether the technology was real or hype is finished, and adoption has raced ahead of understanding. In the rush from skepticism to deployment, few organizations stopped to define what they were buying. The question that dominated the last two years, whether to invest in AI, was the easy one. Asking about the return on investment is harder, and most companies skipped that. They did not ask what problem they were solving, what success looks like before they begin, and how they will measure whether it worked. Those omissions never appear on an invoice, and they are the reason so much AI spending produces motion without return. Activity is not impact The gap between motion and return is now visible in the data. Gartner, drawing on a survey of 204 finance leaders conducted in March 2026, found that adoption has become common while measurable value has not kept pace with expectations. Among finance organizations that have deployed AI, 66 percent reported efficiency and productivity gains, yet far fewer could point to AI changing the decisions that actually move the business. Marco Steecker, a director analyst in Gartner's finance practice, framed the problem plainly. "Counts of pilots, tools rolled out or use cases in production show that finance is moving, but they do not prove that AI is delivering the value boards now expect," he noted. Deploying AI and creating value with it are different achievements, and most companies have reached only the first. The pattern extends well beyond finance teams. The IBM Institute for Business Value, in a 2025 survey of 2,000 chief executives, found that 64 percent acknowledged investing in technologies before fully understanding their value. The investment came first. The clarity, if it arrived at all, came later. A tool in search of a problem In our work with mid-market companies, the AI decision almost always arrives already framed as a technology question. Is it the right platform, is the price reasonable, will it integrate cleanly? Those questions matter, though they sit downstream of the ones that determine the return. What is the business problem? What would a good outcome look like? How would anyone know if it arrived? Consider a composite drawn from situations we encounter regularly. A specialty distributor, profitable and growing, had invested in an AI-driven forecasting platform. The team was enthusiastic, the competitive pressure real, the vendor credible. Six months in, the CEO could not say whether it was working. The forecasts were more frequent and granular than before, but the company was not making better decisions because of them. Inventory stayed misaligned with demand in the same categories it always had. The platform had become another report, more sophisticated than the ones it replaced, and just as disconnected from the decisions that mattered. When our fractional CFO examined the situation, they determined that the platform was working as designed. However, the company had not done the harder work of defining what they were trying to fix before they bought it. The tool was purchased to answer a question the company never asked. Gartner's research describes this trap directly, noting that forecasting and insight generation are among the hardest AI use cases to convert into real impact, because too many initiatives focus on incremental improvement rather than the material business problems that resist other methods. The CFO redirected the conversation toward the decision the company had been struggling with for years: how to align purchasing with regional demand the team understood intuitively but had never modeled. Once that problem was named, success was measurable, and the platform finally had something useful to do. The company's next AI investment started with the problem rather than the product. Where the return comes from The second investment succeeded for an unglamorous reason. Someone insisted on naming the problem before approving the spend, and that single sequencing act turned a tool into a return. That sequence is available to any company, and it costs nothing beyond the discipline to slow down at the moment when everyone in the room wants to move. AI has made that moment harder to hold. The spending is larger, the pressure to act is greater, and the temptation to mistake activity for progress has never been stronger. Boards want to hear that something is underway. Vendors stand ready to get it underway quickly. Almost no one in the room has an incentive to ask what the company will be able to point to a year from now. That question belongs to financial leadership, the one function in the building that exists to connect what a company spends to what a company gets. It is why the AI decision should reach a CFO before it reaches a vendor. Before you spend another dollar on AI, be able to say what you are buying and why. Get 1 Smart Business Story delivered straight to your inbox when you subscribe to Inc.'s free daily newsletter.

AI changed what work looks like. Now the operating model must follow | Fortune

Over the past few years, organizations have begun aggressively driving AI adoption across their workforces, expanding access to tools, encouraging experimentation, and, in some cases, mandating use. However, redesigning operating models, ways of working, and the leadership needed to guide this shift has lagged behind the demands of this new era. That gap is becoming the new fault line in the next phase of AI. Across industries, leaders are discovering that deployment alone won't deliver results. It's one thing to roll out access to a new tool, but rewiring workflows, decision-making, accountability, and expectations is much harder, and far more consequential. A new IBM study, Where AI breaks -- or breaks through, shows that nearly two-thirds of executives say AI is reshaping roles and workflows. Yet many organizations have been slow to redesign the systems that support that shift, creating a widening disconnect between leaders' perception of AI progress and how employees actually experience it. While 78% of executives say employees are involved in designing AI-enabled workflows, only half of employees agree. At the same time, leaders report role transformation at twice the rate employees experience it, pointing to steady progress embedding AI into operations, yet just 17% of employees say AI is part of their daily work. More than a perception problem, this points to a deeper organizational reality: AI is changing work faster than companies are changing the way work is done. As AI takes on more routine and process-oriented tasks, such as synthesizing information and analyzing data, the human role increasingly centers on direction, judgment, and decision-making. In practice, this means that more employees are spending less time producing outputs themselves and more time identifying problems, guiding outputs and evaluating results. A data analyst no longer spends hours pulling reports but focuses instead on defining which questions matter. Developers start to ask, "What should this system accomplish?" instead of, "How do I code this?" Marketers spend less time drafting and more time refining strategy, and so forth. This change in the nature of work forces a broader operating model shift, as many organizations built their operating rhythms around individual contributors producing high-quality work. Now, teams must also assign work to AI agents and integrate them into daily workflows, which requires a level of clarity, communication, and oversight many enterprises have never had to operationalize before. This also places new expectations on managers: they must coach judgment, oversee human-AI collaboration, and help teams adapt to new ways of working, even as many organizations have yet to evolve their management practices accordingly. This new friction is becoming increasingly visible across the enterprise. Our study shows that many organizations still lack clear rules for when AI outputs should be challenged or overridden. And even where guidelines do exist, most employees say they already feel outdated or disconnected from how work actually happens. These are not technology gaps. They're operating model gaps, and the organizations pulling ahead are addressing them directly. Providence Health offers a glimpse of what this looks like in practice. The health system deployed an AI-powered HR agent to streamline its hiring process and improve caregiver experience. Managers now spend 90% less time on administrative hiring steps, freeing them to focus on the judgment calls the system can't make, evaluating which candidates are the right fit, deciding where caregivers are needed most, and overseeing decisions that affect patient care. What many organizations are still missing is that AI doesn't create value simply by accelerating existing tasks. It creates value when companies redefine who is responsible for what, where human judgment is required, and how decisions move through the organization. In Providence's case, the technology matters but so does the accompanying shift in focus, expectations, and accountability. Organizations that redesign decision-making, workflows, and incentives alongside AI adoption are achieving up to 73% higher revenue growth and an 11% operating margin advantage, along with stronger trust between leadership and employees. Getting there comes down to a few adjustments: clarifying who makes decisions in AI-enabled work, embedding checkpoints where human judgment is expected, and measuring performance based on outcomes, not just output. Just as important, employees need to understand how expectations are changing and feel confident applying AI in their roles. AI is already reshaping how work gets done. The organizations that break through will align their operating models with the realities of AI-enabled work, turning a fragmented experience into a consistent, trusted, and ultimately more valuable way of working. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

Align AI Agents with Organizational Ethics and Compliance

Over the past year, there hasn't been a board meeting, customer discussion, or leadership conversation I've participated in that hasn't included artificial intelligence. That tells us something important: AI has moved well beyond being an emerging technology. Indeed, it is rapidly becoming part of how organizations operate daily. Like most business leaders, I'm excited about what AI can make possible. It has the potential to improve productivity, accelerate innovation, uncover new insights, and help organizations serve customers in ways we couldn't have imagined just a few years ago. But over the course of these conversations, I've also become convinced that we're spending too much time asking the wrong first question. Most discussions begin with, What can AI do? I believe they should begin with a different question: How do we ensure AI reflects our organization's values? While technology changes how work gets done, values determine how decisions get made. That distinction will become increasingly important as AI becomes embedded in the everyday work of all organizations. AI is changing more than how we work Unlike previous generations of enterprise technology, AI hasn't been introduced through carefully managed implementation plans. Employees have embraced it quickly, experimenting with new tools, and discovering practical ways to improve their work, often well before organizations establish clear policies or governance. That isn't surprising. AI is remarkably accessible, and its capabilities continue evolving at an extraordinary pace. What is surprising is how quickly AI has become part of everyday decision-making. Employees are no longer using it simply to summarize documents or draft emails. They're using it to analyze information, develop recommendations, solve customer problems, and support decisions that carry real business consequences. That changes the leadership challenge. AI is a human behavior issue, not a technology issue. Every interaction with AI involves judgment. Every decision raises questions about transparency, accountability, fairness, privacy, and trust. Those have always been at the heart of ethics and compliance. AI simply brings them into sharper focus. AI amplifies culture AI amplifies the existing culture. Organizations with strong values, thoughtful leadership, and a culture where people exercise sound judgment are far more likely to use AI responsibly. Organizations without those foundations face a different reality. AI can accelerate inconsistent decision-making just as easily as it accelerates productivity. Technology reflects the environment in which it operates. That's why culture becomes even more important in the age of AI. At LRN, we've spent decades helping organizations build ethical cultures because we've consistently seen the connection between integrity and business performance. Our research continues to show that organizations with stronger ethical cultures outperform their peers in innovation, adaptability, customer satisfaction, and employee engagement. AI can strengthen that relationship. Look beyond productivity Today's AI conversations understandably focus on capability: Can AI improve productivity? Can it reduce costs? Can it help employees accomplish more? Those are all important questions. But they're incomplete. An AI system can produce technically accurate work while overlooking context, introducing bias, mishandling sensitive information, or recommending actions that are inconsistent with an organization's values. Being technically correct is not always the same as exercising good judgment. LRN's 2026 Ethics & Compliance Program Effectiveness Report illustrates how early we still are in this journey. Only one-third of organizations currently reference AI ethics in their codes of conduct. At the same time, AI adoption across ethics and compliance programs continues to accelerate. The gap between adoption and governance is growing. Closing that gap will require leadership. Leadership sets the standard Technology teams play a critical role in helping organizations deploy AI responsibly. But they cannot define an organization's values. That responsibility belongs to leadership. Boards and executive teams need to establish expectations about where AI should be used, where human judgment must remain central, and how accountability will be maintained. Organizations need practical policies, effective governance, and ongoing training. But those efforts are most effective when employees see leaders consistently reinforcing the behaviors and values they expect. Employees pay far more attention to what leaders reward than to what policies say. AI will increasingly operate within that same culture. The organizations that succeed will be the ones integrating AI thoughtfully into cultures built on integrity, accountability, and trust. The next frontier One area we're exploring at LRN is the distinction between compliance accuracy and ethical reasoning. Our early observations suggest they are not the same. An AI agent may correctly apply a policy while failing to recognize the human realities that often shape ethical decisions: fear of retaliation, unequal power dynamics, cultural differences, or competing stakeholder interests. Those situations require judgment, and judgment remains fundamentally human. As AI evolves, organizations will need to evaluate more than whether AI can complete a task. They'll need to ask whether it completes that task in a way that reflects who they are as an organization and what they stand for. That is a much higher standard, one that will distinguish good organizations from great ones. A leadership opportunity Every generation of business leaders is shaped by transformational technologies. For previous generations it was the internet, mobile computing, or the cloud. For ours, it is artificial intelligence. AI will undoubtedly help organizations become more innovative, more productive, and more responsive than ever before. But technology alone will not determine who succeeds. Leadership will. The organizations that earn the greatest trust -- and ultimately deliver the strongest long-term performance -- will be those that combine the power of AI with something technology can never replace. That is, human judgment, integrity, empathy, and accountability. At LRN, we call this "principled performance": the belief that sustainable business success is built on ethical leadership as much as operational excellence. As AI becomes part of every organization, I believe that principle will become more important than ever. Get 1 Smart Business Story delivered straight to your inbox when you subscribe to Inc.'s free daily newsletter.

Bausch & Lomb CEO: the AI hysteria is nothing new | Fortune

AI is being discussed as though business has never encountered a technological shift before. It has. Every generation has its breakthrough technology - the innovation that promises to change how companies operate, compete and grow. The technology changes, but the corporate response is remarkably familiar. Companies rush to buy new systems, hire specialists and announce ambitious transformation plans. They focus on which platform to choose, how quickly they can deploy it and whether they are moving as fast as their competitors. In the excitement, it's easy to mistake adopting the technology for having a strategy. We've seen versions of this before. New technologies arrive with predictions that they will redraw the competitive landscape. Companies invest heavily, reorganize around them and assume that early adoption will translate into lasting advantage. In time, however, access broadens. Competitors acquire many of the same capabilities, and features that once seemed revolutionary become standard. That's when the real difference between companies becomes clear. Acquiring a tool and building an organization capable of using it are two very different things. AI may be more powerful than the technologies that preceded it. It may move faster and reach further into every function of a company. I don't discount the magnitude of the change. But the underlying business lesson is not new: Technology does not create lasting advantage on its own. People do. More specifically, organizations that learn faster than everyone else do. That's the part of the AI conversation that businesses risk overlooking. Companies are racing to evaluate models, compare vendors and deploy new tools. Those decisions matter, but they are unlikely to determine which organizations ultimately succeed. Access will continue to broaden, capabilities will spread and today's breakthrough will become tomorrow's expected feature. The more important question is whether an organization's people are prepared to keep learning as the technology changes. That's why I believe the most important AI investment a company can make is not simply in technology. It's in building a workforce that is curious, adaptable and willing to challenge how work has always been done. That kind of culture cannot be purchased from a vendor. It has to be built over time. At Bausch + Lomb, we've tried to make learning part of how we operate rather than something reserved for an occasional training event. Last year, we partnered with Coursera to launch an enterprise-wide AI learning program because we believe AI literacy should become a core business skill, not a niche technical capability. We also made the courses mandatory for our knowledge workers. Some people questioned that decision, which was understandable. Mandatory training is not always welcomed, and completing a course does not make someone an AI expert. But if AI is going to affect nearly every business function, giving people a foundation for understanding and using it should not be optional. Training, of course, is only a starting point. Learning creates value when people put it to work. We launched our VisionAI Challenge and invited colleagues across the company - not just AI experts - to identify practical ways AI could improve how we operate. Ideas came from manufacturing, R&D, commercial operations, finance, HR and other parts of the business. What stood out to me was the range. Some ideas were ambitious. Others addressed small, persistent problems that consume time and make work more complicated than it needs to be. Those ideas may not make headlines, but collectively they can make a company faster and more effective. The experience reinforced something I've seen throughout my career: The people closest to the work often have the clearest view of how it can be improved. The challenge for leaders is to give them the knowledge, permission and opportunity to do something about it. It's equally important to help good ideas travel. That's the purpose of AI in Action, a platform where colleagues share practical examples of how they are using AI to solve problems, eliminate repetitive work and improve how they serve customers. Some examples save dozens of hours each month; others save only one or two. They all have value - especially when one person's solution gives someone elsewhere in the company a better way to approach a similar problem. Over time, those improvements add up. More importantly, the people who develop them become teachers as well as problem-solvers. That's another lesson that predates AI. Companies often treat transformation as something directed from the center: A small group selects the technology, defines the process and tells the rest of the organization how to use it. Central coordination has a role, particularly when it comes to standards, security and responsible use. But lasting change rarely happens by memo. The companies that benefit most from AI will be the ones that equip people throughout the organization to experiment responsibly, share what they learn and help others improve. That also requires leaders to reconsider our own role. For years, leaders were expected to have the answers. Increasingly, our responsibility is to create an environment where people ask better questions. We need to set clear expectations and guardrails, but we also need to be comfortable acknowledging that no one yet knows exactly where this technology will lead. AI will continue to evolve. Today's leading model will eventually be replaced, and capabilities that seem extraordinary now will become familiar. No company can lock in a lasting advantage simply by choosing the right tool at one moment in time. What can endure is curiosity, adaptability and a culture where learning is understood to be part of the job. People often talk about AI as though it is replacing human potential. I think it can amplify that potential, but only when companies spend as much time preparing their people as they do evaluating their platforms. The current frenzy will eventually subside, as it has with other waves of technology. AI will become more embedded, more familiar and more widely available. When it does, the advantage will not belong to the company that was first to buy the latest technology. It will belong to the company whose people kept learning what to do with it. The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.