Major firms halt AI rollouts as hidden data and confusion expose implementation challenges

When revealed data brings AI rollouts to a screeching halt - and how to manage it

Follow ZDNET: Add us as a preferred source on Google. ZDNET's key takeaways * AI can boost productivity and improve data access. * Tech leaders have had to halt rollouts due to data concerns. * Long-fogotten insights emerge with AI prompts. Agentic and generative AI have opened up information and insights to professionals in enterprises. However, evidence suggests that trend could be too much of a good thing. At a recent conference, veterans of enterprise AI rollouts issued cautionary words to professionals considering diving headfirst into AI. The issues these professionals encountered even led to temporary halts in AI rollouts meant to boost employee productivity, as executives reassessed information that could be exposed internally. At the same time, the executives, who spoke on a panel at the recent Veeam conference in New York City, emphasized that AI wasn't the source of the challenge. Both panelists' organizations had accumulated vast stores of data, and one required a new governance structure. Also: 96% of IT pros use AI now: Their top 7 agentic applications and biggest implementation roadblocks Steve MacIntyre, senior vice president at Fidelity Investments, described how his 400,000-employee company saw data long tucked away in the recesses of its organization -- on SharePoint sites or in network-attached storage, for example -- suddenly surface via AI prompts. "It wasn't an AI problem," he said. "It was the productivity and the ability of AI to find things quickly." Wim Geurden, chief architect for enterprise tech at EY, described his company's challenge as pinning down data ownership across its global network of independent affiliates -- data that was also surfacing through its AI engine. "When big enterprise search was launched, all kinds of stuff started to surface in places that people went," he said. "EY Global doesn't own any of the data. Every member firm owns its data. That is where the first questions were raised. What's all this? How many SharePoint sites? We had multiple petabytes of data, and it was the Wild West. There was no lifecycle management on these SharePoint sites, and half of them had no owners. We didn't know when they were last accessed." Also: 51% of professionals say AI workslop lowers their productivity - stop it in 2 steps At Fidelity, information was emerging from a vast library of PowerPoint and PDF reports. "We have an entire history, decades of research notes at Fidelity, such as PDFs," said MacIntyre. "We gave out a few licenses for Copilot, and immediately, two days in, legal came to me and said we have an AI problem. One of my team did a search to find something and AI came back with all the PowerPoints that were on SharePoint from years ago." AI is a "tremendous search engine that runs at speed," MacIntyre continued. "Suddenly, it's searching everything that it has access to, and surfacing that to us in a meaningful way. Everybody thought we had an AI problem, but what it showed was a problem of securing data. This issue hit home when we immediately realized that we had all of this data that we didn't think we cared about -- unstructured data -- and along came LLMs, and suddenly all of that data becomes gold." Establishing guardrails At EY, as the gates of its vast data stores opened to AI, the priority was to "find who owns the data," said Geurden. "The second thing we did was we shut everything off." Users could only access the Copilot tool if they were licensed. Also: Building an agentic AI strategy that pays off - without risking business failure The data ownership verification process included identifying and labeling the data found across the EY enterprise, Geurden continued. For example, labels included "confidential" or "financial services." AI itself offered a means to help label the company's knowledge repositories of unstructured data, Geurden explained, noting the challenge of human labeling with a 25% annual turnover rate. However, labeling needs to go deeper than simple high-level tags. "The first thing is we have to know what was there when the AI ran," said Geurden. "We need to have the historical picture, the versions." Then, "we have to go way beyond the labeling of confidential information. We need to have geo-restrictions, geo-labeling, line-of-business labeling, linked to our contracts, because we get an enormous amount of client data specifying what we can do and what we cannot do." Also: Over 80% of US government agencies already use AI agents - and it's only the beginning All of this metadata has to be codified into contracts, he added: "That's the easy part. Then we have to codify it in some technological structure. That is, for now, still very, very cumbersome." Governance is the key to success across all aspects of these AI implementations, the executives emphasized. "We have to know what's being used," said MacIntyre. "That brings into play the idea of shadow AI, shadow IT, all those kinds of things -- and it goes back to the endpoint data. We have to know that the asset inventory is accurate. Are they aligned with the use cases that are registered and approved? That way, at least we know that if someone's working on something, they should be using Claude, because it's tied to a particular project that was approved for that." Also: These 4 critical AI vulnerabilities are being exploited faster than defenders can respond Next, "we have to think about what's the safe environment where we want these agents to run?" MacIntyre continued: "How do we want them to interact with the foundational models? What architecture do we put in place to funnel all that activity into a place that gives us the right visibility and telemetry so we can see that agents and applications using AI are behaving in the way that was intended? Or misbehaving?" An additional challenge -- perhaps the most vexing for all digital leaders at this time -- is establishing agent identity, said MacIntyre: "How do you give an agent identity? They then become an employee. But what if my agent only lives for seconds? It's a really interesting problem, and I don't know if anybody's solved it really well yet."

'Confused' AI strategy hurts firms and baffles staff

When AI engineer Malcolm was working at a data analysis firm, executives wanted to use generative AI to categorise the customer database into a range of personas. "Don't use AI," was his advice. A traditional machine learning model would have been much more appropriate, he argued, producing consistent, repeatable results. And it would have been much cheaper. "They still went ahead with Gen AI," says Malcolm (we have not used his real name). That meant a process that was less accurate and much more expensive, but it also allowed the organisation to say they were embracing AI. Malcolm's experience will be familiar to staff at other companies. More bosses are embracing AI and insisting their staff use it. In February, global consultancy Accenture reportedly told staff that promotions to top roles would require "regular adoption of AI tooling" and it would be tracking their usage of the AI platform it has developed. And in May, rival firm KPMG said it had developed a dashboard to track whether its US employees' meet a 75% usage target for its AI tools. The company says this is part of "a holistic effort... to help people move up the AI maturity curve." Other organisations are taking a less targeted approach to implementing AI but nevertheless expect it to transform how their workforces spend their days. Governments are also hoping to tap into some AI magic. The UK government is banking on AI to help "rewire" the state and boost efficiency across Whitehall. However, research by the civil servant union, the FDA, shows that while civil servants were open to the idea of using AI to improve productivity, there's doubt that management can handle the transformation. Less than a third of civil servants had been consulted on how the technology could be rolled out, the union found, meaning "change is being done to workers, not with them". FDA general secretary Dave Penman said the rollout was "inconsistent across departments which limits the productivity gains". If organisations are quick to highlight AI adoption, says Dan Boyles, CEO of consultancy Hello AI Collective, they're not always clear on why they're adopting it and how they expect to benefit. "I was with an oil and gas company, and I sat with the C-suite, and I just went 'what's the reason for using AI?' And none of them could agree." The firm's CEO cited the need to keep up with competitors, Boyles continues, while the head of sales said they wanted to make more money, and the marketing team wanted to stop using outside contractors. This sort of confusion at the top can mean AI investments fail to deliver on expectations. "I think the wreckage is organisations not getting the ROI [return on investment] from it that they were expecting and not getting their people engaging with it," says a senior consultant at one large consulting firm, who did not want to be named. In his firm, everyone had access to two AI tools, but could request specialist tools for specific tasks, such as coding. If their job demands it, "some of our people will have access to four or five, AI, tools". Organisations needed to consider the people side of the equation, he continues. "There are generational differences in terms of confidence levels with regards to this. There are potentially gender differences." And before anyone in his organisation can have access to a tool, he says, they must take mandatory training covering AI ethics and risks such as bias. This training also makes clear that AI tools can be sycophantic and hallucinate, he adds. The pre-existing culture in an organisation can make or break an AI rollout, not least because AI tends to accelerate things for better or worse, says Caroline Rawlinson, CEO of Culture Amp, which tracks employee experiences and feedback. The firm says that while nine out of 10 HR professionals expect to increase their use of generative AI, a third said "say no one currently owns AI strategy at their companies". "If you're putting AI technology on top of a fragmented culture or a fear-based culture, it is not going to succeed," says Rawlinson. "At best, it becomes a very slow roll out as people don't understand what they're being asked to achieve or the tools that they're being provided with. At worst, it ends up as quite a big, wasted effort." In the case of the oil and gas company Boyles was helping, the president eventually said: "I want to increase my operating earnings because I want to sell [the company] in years." That motivation was the key bit of information for Boyles. His team could then go to each department, talk through their processes and technology, identify bottlenecks, and work out where AI could actually help.

Should you treat AI agents as colleagues? Fortune 500 executives can't settle the debate | Fortune

The debate over how to integrate AI agents into the workplace has produced no shortage of frameworks, mandates, and org-chart overhauls. And this week at Fortune's COO Summit, it produced something rarer: complete, 180-degree disagreement between two executives who have thought about this longer than almost anyone, and still left with no clean resolution. Eric Kelleher, President and COO of Okta, has named the agents on his team Leo, Sloan, Hank, and Walker (among others). They show up in business reviews alongside his human staff. The turning point, he said, came during a standup when he asked staff to give names to their own agents. "In that exercise, AI became a colleague as opposed to a tool," he told Fortune on the sidelines of the panel, "and that catalyst is valuable." Francine Katsoudas, the Executive Vice President and Chief People, Policy & Purpose Officer at Cisco, heard something like that and pushed back hard. "I would not look at AI as a colleague," she told a separate audience at the COO Summit just hours later. "I think we should look at AI and agents as part of the workflow, but not a colleague. And I think the sooner we land that, the more confident our people will be." Both executives are operating at scale and are navigating the same underlying crisis: companies have largely figured out how to experiment with AI, but remain in experiment phase, if not in collective denial about how to actually redesign work around it. Cognizant, whose research team presented new data at the COO Summit, found that 93% of jobs are already being disrupted by AI -- six years ahead of their own 2023 projections. But the productivity gains that were supposed to follow haven't materialized. Their researchers called it an "activation gap." The debate over what to call agents might sound is not just semantics. Katsoudas also talked to Fortune Editorial Director Kristin Stoller about how her firm handled 4,000 announced layoffs as part of an AI restructuring -- noting that on the teams using AI most effectively, trust within those teams actually began to drop about nine months in. "We just have to invest so much more," she said. "We have to share with our people what we know, what we don't know." The mechanism she's betting on: investing in skills, not just severance. In previous Cisco restructurings, pairing training with internal redeployment allowed the company to place 75% of impacted employees. "Just imagine if that became 85 or 90 percent," Katsoudas said. "It would make people feel a lot less worried because they know they're going to land. She said it's what Cisco is "working through today. It's tough." A randomized experiment published by Harvard Business Review in May reached a similar conclusion from a different direction: humanizing AI can shift accountability away from individuals, increases escalation, and reduces the quality of human review -- the opposite of what most companies deploying agents are hoping for. A separate experiment by Boston Consulting Group found that human workers responded to their AI colleagues by scapegoating them and getting more careless with their own work. Research from the University of Arizona adds another wrinkle: disclosing AI use at work makes colleagues trust you less in the short term, but staying silent and getting caught later is worse. Companies are, in effect, caught in a transparency trap -- honesty carries a social penalty, but concealment carries a steeper one. Franklin's answer to that trap is blunt governance. "We don't just let any person into your home to talk to your children, eat your food, sleep in your bed," she said. "You ask them who they are, why they're there." The same logic, she argued, applies to AI. "We don't just let any AI in. We need to have clear guidelines and clear guardrails around what happens when you bring AI in." It's a frame that treats trust not as a feeling to be managed but as a system to be designed -- before the agents arrive, not after. Kelleher's concern runs the opposite direction. The problem, in his diagnosis, isn't that workers will feel displaced by agents with names -- it's that managers still aren't taking agents seriously enough as a category of labor. "We have trained every manager in the world to think about one thing," he said, "and that is: what's their headcount? What's the org chart look like? Who reports to who?" That thinking, he argued, doesn't fit this moment. His proposed fix: push token budgets down to people managers, forcing a concrete reckoning with a workforce that now includes AI agents operating alongside humans -- and making that trade-off visible in the budget itself. Sarah Franklin, CEO of Lattice -- whose entire business is built around helping companies manage and develop their people -- made the same diagnosis from the other direction. The performance management process, she argued, is "deeply broken" -- cyclical, once or twice a year, disconnected from how businesses actually move. AI has exposed that, rather than fixing it. "You set up your OKRs at the beginning of the year," she said, "then six months in, priorities have changed, focus has changed. Not that that's bad. It's that the performance process hasn't kept up with the business." What Kelleher and Franklin actually agree on, underneath the framing fight, is more important than the disagreement: the bottleneck is at the managerial level. Org charts, budget cycles, performance processes -- these were all built for a workforce of humans and not yet rebuilt for one that isn't. Cognizant's analysis of 80,000 tasks found that in 90% of them, a human still needs to be involved in some way. But whether they call the AI agents that they work alongside colleagues is the question. "We evolve from workforce planning to work planning," Kelleher said. "What I'm finding right now is that's a really big leap for people to make." Whether the agents helping bridge that gap are colleagues or tools may matter less than whether the humans managing them are finally forced to reckon with what work actually looks like now. For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.

Agentic AI - an infinite regress of cost/benefit challenges?

A computer can never be held accountable. Therefore, a computer must never make a management decision. Those were the words famously used by IBM in a 1979 training manual and slide presentation. But according to a new Harvard Business Review report, Solving Agentic AI's Data Infrastructure and Telemetry Needs, published this month, that axiom has been thrown out completely by business leaders. It says: Agentic AI is moving from experimental to operational inside most businesses. Companies aren't just autocompleting emails or summarizing documents anymore, they're asking software to decide what to do next. [...] Such uses of technology represent an epochal shift in how work gets done. In other words, management decisions are being ceded to AI agents. But is that wise? The report adds: There's the need for good governance to dictate how agents are allowed to act. Agents' speed and autonomy mean that companies must have strict guardrails around what they can access and do - and the ability to stop them when necessary. Because agents can potentially make changes to business records and data sources, these concerns go hand in hand with considering liability and compliance. Consider that statement for a moment - "Agents can potentially make changes to business records and data sources". In essence, that is an admission that an agentic AI can re-write history, in effect, and leave your organization de-coupled from verifiable fact, evidence, and a reliable audit trail. Separate research from Freshworks, The Global Cost of Complexity Report: The Midmarket AI Complexity Trap, published this week, surveyed over 12,000 IT professionals. Among its many findings, the report says: Managing AI is now adding to the workload it was meant to reduce, with teams fixing flawed outputs and governing tool sprawl across a growing stack of AI products. More than eight in ten (86%) of mid-market IT leaders say managing AI complexity has actually increased their team's workload, and 80% report that AI outputs are introducing noise, errors, or rework, a phenomenon the report terms 'AI slop'. This re-defines AI slop from meaning lazy, ersatz, AI-generated art to what others have called "vibe citing": hallucinations presented as business insight. And according to the Harvard report, that is happening in an overwhelming majority of organizations. The Freshworks report adds: AI is generating work faster than it is eliminating it, and IT teams are absorbing the difference. So much for the infinite productivity uptick predicted by some AI acolytes... Inevitably, this frustrating new burden also impacts on organizations' ability to demonstrate ROI from AI, especially given the many reports in recent years that have identified cost and time savings as the dominant reasons for enterprise adoption of the technology. More, it has worrying implications: what of the errors that have never been spotted, and which have yet to work through the system? How many management decisions - human or agentic - have been based on hallucinated data? And, to echo that fifty-year-old IBM observation, who is responsible for those failings? (AI vendors are working to ensure it isn't them, of that you can be sure.) Factor in the 1,494 known examples worldwide of fake AI caselaw being presented in completed court cases, (over two-thirds of them in the US), plus EY Canada retracting a report on loyalty scheme fraud this month due to its countless AI-generated citations, and Deloitte part-refunding the Australian government last year for a report that was also full of hallucinations, and there is ample evidence that even expert service providers are not just trusting AI too quickly, but also abandoning their professional responsibilities in the process. Indeed, it is remarkable just how quickly due diligence appears to have been set aside by experienced professionals. So, how can we expect less experienced junior employees to do better? The Harvard Business Review report is far from evangelical in its own summary of agentic AI's impact on business.Sponsored by AI telemetry platform Cribl, it notes: We see the same pattern over and over: forward-looking executives have big ambitions for agentic AI, but they're trying to run those agents on top of fragmented, expensive, and opaque telemetry. Most businesses don't have an artificial intelligence (AI) problem in the abstract; they have a telemetry and platform problem in practice." Ouch. And yet other research clearly does find that organizations have an AI problem too, in the sense of seeking deterministic answers from systems that are probabilistic in nature. Bear in mind, Anthropic CEO Dario Amodei recently wrote on his own website: Models are trained on vast amounts of literature that include many science-fiction stories involving AIs rebelling against humanity. This could inadvertently shape their priors or expectations about their own behavior in a way that causes them to rebel against humanity. That is surely a simple admission that AIs can't tell truth from fiction, so perhaps we should not be surprised when they generate fiction in place of evidenced, verified, peer-reviewed fact? (And who trained the system?) The Harvard Business Review report goes on to say that, while 96% of senior leaders say agentic AI will be "critical to business strategy within two years", only 23% believe they currently have the strategy and infrastructure in place to support it. Other findings include the 47% of organizations deploying AI agents who say that infrastructure costs have far exceeded their expectations, while almost as many (46%) cite unclear ROI and performance metrics resulting from that. In short, the ambition is enormous but the infrastructure supporting it isn't, notes the report: The problem isn't a lack of vision. It's that most enterprises are trying to run a new generation of AI on legacy observability and security stacks that were never designed for it. [...] For organizations at the leading edge of agent deployment, those legacy systems are not just struggling, they are already failing under the load. Cribl co-founder and CEO Clint Sharp adds: Data is growing at a 30% CAGR, but budgets are not, and now AI agents are multiplying that problem by an order of magnitude. The infrastructure most enterprises built for the last decade simply wasn't designed for the agentic workloads of the next one. But this leaves many enterprises with an infinite regress of challenges: they are being exhorted to invest more in a supporting infrastructure that is fit for the agentic age, and yet they are already struggling with mounting costs. Meanwhile, the promised ROI is absent to marginal in many cases - see diginomica, passim - and significant amounts of saved time are taken up with tracing and fixing errors and hallucinations. The point is this: it is a huge ask to expect enterprises to spend even more money in the hope that ROI will finally appear as promised - infinite speculation with no guarantee of accumulation, perhaps. With early sight of the above research, I put some of these findings to Jakob Freund, co-founder and CEO of enterprise orchestration platform Camunda last week at the company's agentic-focused CamundaCon event in Amsterdam. After all, the core theme of that event was the need to rein in and manage agentic AIs, and not let them manage you and risk damaging your business. What did he make of the Harvard Business Review finding that 96% of organizations see agentic AI as critical, but only 23% have the infrastructure in place to capitalize on the technology? And how could those businesses be persuaded to invest yet more money, given the soaring costs already reported by many? Freund said: I'm not surprised at all to hear these two numbers. It's exactly the problem that we're tackling. So, in that sense, that is another confirmation of why there's a need for [orchestration]. But in a way, that is also a self-selected filter. I'm not so concerned that leaders wouldn't see the need for that investment, because they do need to put together an AI-first operating infrastructure. The challenge now is how to get there. That said, I am sure there are still a lot of leaders and organizations out there that have not fully realised that yet. OK, but many of those leaders report that costs are already soaring, and yet they are not seeing the anticipated payback. Freund replied: Yeah, but it depends on how we define an infrastructure cost. I think many are negatively surprised [sic] by the cost of tokens in LLMs, right, and token consumption efficiency is an imperative. Deterministic orchestration based on if-this-then-that [ITTT] statements is very cheap, with regards to resource consumption, so you probably want to do deterministic orchestration as much as possible and dynamic, LLM-driven orchestration only as much as necessary, because it is by definition way more expensive. And far less reliable, perhaps? He concurred: And less reliable, exactly. It is less efficient and more risky. So, what's the takeaway from all this? Essentially it is this: things have not changed that much from the days of IBM's famous axiom. You are still in charge, and you should not let your technology manage you or take the decisions for you. And beyond that, you should not assume that it has presented you with the verifiable facts. Take the time and the responsibility to check, because you won't be able to blame the AI.

The automation illusion: Why AI is making COOs' jobs harder, not easier | Fortune

When the COO of Nike, the chief of operations at an $84 billion food distributor, and the CEO of a major tech media company walked into the same room at the Fortune COO Summit, they came ready to talk about what AI was doing for them. Speed. Scale. Revenue unlocked. The future arriving ahead of schedule. What they described instead, during a lunch roundtable hosted by Thomson Reuters, was something closer to organized chaos. "The biggest challenge I could see is speed without clarity," said Venkatesh Alagirisamy, EVP and COO of Nike. "I see a lot of hype around AI that drives a lot of energy within organizations in wanting to adopt AI, but without that clarity, without that sense of purpose, that speed could get us in the wrong direction." Welcome to what panelists called the "automation illusion" -- the dangerous gap between what AI promises operations leaders and what it actually delivers. The promise was simple The way the COOs described it to Fortune Editorial Director Diane Brady the AI pitch was almost too good. Automate the routine. Free up the workforce. Let the machines handle forecasting, logistics, compliance, customer service. Let humans handle strategy. Aayush Bhatnagar, global head of customer service at Sysco -- which moves food to restaurants across North America, generating nearly $84 billion in annual revenue -- put it plainly: the goal was to take tribal knowledge baked into decades of human relationships and institutionalize it at scale. "Every piece of broccoli you're eating has moved an average of 2,000 miles," he said. The supply chain that makes that happen runs on judgment calls made by people who've been doing it for years. AI was supposed to absorb that expertise and multiply it. And in some ways, it has. Nike launched an internal learning platform 12 months ago -- peer-curated, bottoms-up, not mandated from above -- and logged 20,000 digital courses taken, with 3,000 live training sessions conducted. Sysco is using AI to rethink how it forecasts and buys. Thomson Reuters is deploying it to help lawyers, tax accountants, and trade professionals work faster. But this has all come with a big reality check. The illusion kicks in Laura Clayton McDonnell, president of corporates at Thomson Reuters, expanded on the automation illusion. "We're going to move fast, we're going to get these answers really quickly," she said. "But what about making sure that output is reliable, it's accurate, it's something that you can drive your business on?" That, she added, is where companies really need to pause instead of give in to the need for speed. For the professionals Thomson Reuters serves -- lawyers walking into courtrooms, accountants navigating tariffs, trade teams dealing with sanctions -- there is no margin for error. "You cannot be wrong," McDonald said. "You just can't be wrong." A large language model that confidently produces a plausible-but-wrong answer isn't a productivity tool in that context, but a liability. The illusion runs deeper than accuracy, though. The bigger problem is that AI has made the operating environment fundamentally less predictable -- precisely the environment COOs are paid to manage. Olivia Nottebohm, COO of Box, said she has watched it play out inside her own company. Box sells AI products. It runs Box AI internally. It talks about AI constantly. And when Nottebohm looked at the adoption numbers, they were low. "Here we are, an AI company selling AI," she said, "and I wasn't even seeing the adoption I was expecting." When she dug in, she found the answer wasn't resistance -- it was confusion. People didn't know how. The tools were available. The skills weren't. She shared that the company impemented a program called "No Boxer Left Behind." It worked, but it also revealed a harder truth: even at a tech-forward company, the gap between deploying AI and operationalizing it is enormous. "Really making sure that people don't feel disenfranchised, I think that has been the thing that took me the longest to figure out," she shared, adding that she "should have figured it out sooner." The company's mandatory trainings are clear about what Boxers have to learn, "and if you choose to opt out of being on the AI transformation, that's up to you. But we, as an employer, are not going to let you do that." The management problem no one has solved Nothing illustrated that gap more starkly than Bhatnagar's admission about his team. Four weeks ago, he told the room, he added seven AI agents to his direct reports. They have names. They have defined roles -- an escalation agent, a delivery agent, a communications agent. Their performance is reviewed alongside the humans at his weekly business review. "I lost some sleep that night," he said, "thinking that our traditional laws of leadership, principles of leadership, do not apply to these agentic agents." To his point, there is no management literature for that, no HR policy or performance improvement plan you can put an agent on. And yet COOs like him are already accountable for their output -- output that can scale instantly and go wrong just as fast. "How do I train my managers now?" he asked the room. It may have been the most honest summary of where enterprise AI actually stands. The deeper stakes Near the end of the discussion, the question hanging over the room became explicit: what happens to the entry-level workers who traditionally built their judgment doing the exact tasks AI is now absorbing? McDonnell kept returning to the same guardrail: the human in the loop isn't optional, it's structural. "I don't think we've found a tool yet that actually can exercise business judgment," she said. "That's the difference maker." Alagirisamy framed it as the central leadership capability of the moment: learning agility. Not AI fluency, not technical depth, but the organizational muscle to keep adapting as the ground keeps shifting. "Does your team have the learning agility to adapt to this new environment?" he said. For COOs, the automation illusion isn't just about bad AI outputs. It's about the widening gap between the speed at which the technology is moving and illusion of how much work can be automated, and the reality that it's much easier said than done. They came in talking about what AI was doing for them. They left still trying to figure out what to do about it. For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing.

Okta's COO says companies are in denial about the hardest part of the AI revolution: redesigning work itself | Fortune

Eric Kelleher has a problem that no amount of AI can solve for him. The President and COO of Okta has agents on his team. He's named them -- Leo, Sloan, Hank, Walker -- and they show up in business reviews alongside his human staff. He's personally booked a flight to Bangalore and spent the entire trip standing up an open-source agent on a separate machine, a deliberate act of immersion he then assigned to every member of his leadership team. "That flight to me was transformative in how I recognized what the capabilities of this technology are," he told a roomful of top operations executives at the COO Summit this week. And yet, he added, the hardest part isn't the technology. It's the managers. "We have trained every manager in the world to think about one thing and that is: what's their headcount," Kelleher said. "Our managers have spent decades learning how to think about headcounts and payroll." The shift he's advocating for at Okta -- getting managers to budget explicitly for both human labor and digital labor, to think about work charts that include AI agents as genuine colleagues -- is, he said, "a much harder problem than getting people to experiment with Claude Code." "One of the things I'm really advocating for within Okta is to get our managers thinking about how to design work to include human workers and digital workers," Kelleher told a room of top operations executives. "Everyone has the mandate [to adopt AI]," he said, but people are not really thinking through what it means to tackle that mandate. "One added piece that's very top of mind for me right now is: when we go into budget planning, when we go into cycles, we have trained every manager in the world to think about one thing and that is, what's their headcount? What's the org chart look like? Who reports to who? How many layers do we have? How does this span of control?" That thinking doesn't fit this moment, he added. It was appropriate for the session hosted by Cognizant: New Work, New World: How AI is reshaping your org chart, with Head of Research Ollie O'Donoghue and Chief Business Officer, AI, Sushant Warikoo, digging into the topic. Kelleher's remarks crystallized a growing frustration among executives: companies have largely figured out how to experiment with AI, but remain in collective denial about how to actually redesign work around it. From headcount to 'work planning' Kelleher's proposed solution is deceptively simple: stop thinking about labor purely in terms of people. His fix? Push token budgets down to people managers. The idea is to force a concrete reckoning with a workforce that now includes AI agents operating alongside human employees -- and to make that trade-off visible in the budget itself. "What we want to start seeing is how do work charts evolve where we have digital workers working alongside human colleagues," he said. The current conversation is focused too much on AI displacing jobs, he said, "not changing the nature of work itself." Kelleher's remarks came as Cognizant released new research showing that the AI transformation is happening far faster than anyone predicted -- and yet its value is failing to materialize. In 2023, the firm projected 90% of jobs would be disrupted by AI by 2032. Today, that figure is already 93%, six years ahead of schedule. But the productivity gains that were supposed to follow haven't. O'Donoghue described this as an "activation gap," or a chasm between what AI can theoretically do and what companies are actually achieving. "There's a bit of a disconnect between theory and reality," O'Donoghue said, citing analysis of 80,000 different tasks, conducted each of the last three years. "Ninety percent of the tasks that we analyze ... the human still needs to be involved in some way." That makes the organizational redesign problem more urgent, not less. If humans are still in the loop, the question isn't whether to replace them -- it's how to restructure their roles around machines that are increasingly capable of doing the transactional parts of their jobs. The harder management problem Several executives in attendance described trying to crack this problem from different angles. Jon Blotner, President of Wayfair, said the company had reversed course on a top-down AI mandate and instead gave every employee access to Claude, Gemini, and ChatGPT -- then watched teams start reinventing their own roles. "We see people reinvent their jobs and say, okay, look, I basically automated my work," he said. "That person's incredibly valuable." Cognizant's Warikoo agreed that is the unsexy core of the problem. "Humans and agents have equal privilege," he said. "But the entire architecture for enterprises was built on the notion of humans working on business workflows with static application architectures." AI agents require persistent context and operate continuously, a fundamentally different model than the episodic, batch-driven systems enterprises were built around. "It's not about the AI," Warikoo said. "At the end of it, it's about the humans. It's about amplifying human potential, where humans get to do higher-value work." Kelleher's diagnosis is that most organizations aren't there yet. The instinct, still, is to think about digital workers the way companies once thought about software: as a tool employees use, not as a category of labor to be managed, budgeted for, and integrated into the org chart alongside people. "I see the future now," Kelleher told Fortune on the sidelines of the panel, "and it's clear to me, we're not going back." He said a turning point for him was a standup with staff when he asked staff to give names to thieir OpenClaw agents. "In that exercise, AI became a colleague as opposed to a tool and that catalyst is valuable." He agreed that it is similar to the adoption of electricity, when whole factories were slow to realize they didn't need their old steam engines anymore. He said it's similar to how current AI adoption is "just, like, asking people to add chatbots." Later that afternoon, Kelleher told other executives that his team has started realizing that digital agents are colleagues, of sorts. "It's really uncomfortable, but it's very transformative." "We evolve from workforce planning to work planning," Kelleher told the room. "What I'm finding right now is that's a really big leap for people to make."