AI Agents: The Next Frontier in Artificial Intelligence for 2025

AI agents might be the new workforce, but they still need a manager

AI agents keep getting smarter and more independent. But there is still work to be done before an agentic AI-driven workforce can truly assume a broad range of tasks. Increasingly, we hear about AI agents being the new "digital workers" -- a concept that arose before agentic or generative AI hit the mainstream in areas such as robotic process automation. Digital workers are designed to serve the discipline and obedience, but just like human workers, they, too, have their quirks. The movement toward a digital workforce has been taking big leaps lately, marked recently by Salesforce's unveiling of Agentforce 2.0, a digital labor platform for enterprises. The platform enables "a limitless workforce through AI agents for any department, assembled using a new library of pre-built skills, and that can take action across any system or workflow." The platform also takes steps well beyond RPA, featuring "enhanced reasoning and data retrieval to deliver precise answers and orchestrate actions in response to complex, multi-step questions," according to a statement from Salesforce. The agents even interact in Slack. Also: 15 ways AI saved me time at work in 2024 - and how I plan to use it in 2025 Major organizations are leveraging the platform to augment their teams with digital labor, the vendor added. Talent is scarce and expensive to train, so organizations are turning to AI to help with customer interactions and deal with workflow backlogs, but can no longer afford "inadequate solutions that provide generic responses," Salesforce stated. "Existing solutions such as copilots struggle to provide accurate, trusted responses to complex requests -- such as personalized guidance on a job application. They cannot take action on their own -- like nurturing a lead with product recommendations." Autonomous digital workers can now perform such work at many levels, industry leaders agree. "The convergence of skilled innovators, rapidly-deployable cloud tools, customer awareness and executive support has created an ideal environment for agentic AI to thrive in 2025," Chris Bennett, director of AI transparency and education at Motorola Solutions, told ZDNET. For example, Motorola Solutions has begun leveraging agentic AI "to improve public safety and enterprise security, with applications that analyze and surface data in real-time to provide crucial, immediate support to first responders and security personnel," Bennett stated. "AI agents never get bored, tired, or distracted, automating repetitive tasks and freeing responders for critical responsibilities and community engagement. AI agents can accelerate tasks like reviewing historical video footage, helping investigators quickly find missing persons through natural language search." This works via AI agents intuiting processes to "create a series of steps, or a recipe to solve a problem," said Viswesh Ananthakrishnan, co-founder and vice president of Aurascape. They can also "take actions to execute these steps and even collaborate with other agents to do so. When combined together, this data gives the agents a view of how the enterprise functions." Also: OpenAI's o3 isn't AGI yet but it just did something no other AI has done The AI agents then "develop and execute complex processes, like viewing demand forecasts and taking proactive action to generate and submit order forms for more inventory before supplies run low," he continued. "This type of automation saves workers significant time and frees them up from repetitive tasks." At the same time, AI agents need to be thoughtfully managed, just as is the case with human work, and there's work to be done before an agentic AI-driven workforce can truly assume a broad range of tasks. "While the promise of agentic AI is evident, we are several years away from widespread agentic AI adoption at the enterprise level," said Scott Beechuk, partner with Norwest Venture Partners. "Agents must be trustworthy given their potential role in automating mission-critical business processes." The traceability of AI agents' actions is one issue. "Many tools have a hard time explaining how they arrived at their responses from users' sensitive data and models struggle to generalize beyond what they have learned," said Ananthakrishnan. Unpredictability is a related challenge, as LLMs "operate like black boxes," said Beechuk. "It's hard for users and engineers to know if the AI has successfully completed its task and if it did so correctly." In addition, he cautions that there is still unreliability in AI agents. "In systems where AI creates its own steps to complete tasks, made-up details can lead to more errors as the task progresses, ultimately making the outputs unreliable." Also: Why ethics is becoming AI's biggest challenge Human workers also are capable of collaborating easily and on a regular basis. For AI workers, it's a different story. "Because agents will interact with multiple systems and data stores, achieving comprehensive visibility is no easy task," said Ananthakrishnan. It's important to have visibility to capture each action an agent takes. "This means deep visibility into activity on endpoint devices and the ability to process data in a vast variety of formats." Then, it's important to be able to "quickly combine this context from endpoints with network-level traffic to determine the data informing the agent's actions," as well as "recognize the type of AI agent interfacing with your data, whether it's a trusted entity, or a brand-new agent." This may boost an emerging human-centered role -- the AI systems engineer. "This new quality assurance and oversight role will become essential to enterprises as they manage and continuously optimize AI agents," Beechuk said. In multi-agent environments, "AI agents will be interacting and evolving constantly, consuming a steady diet of new data to perform their individual jobs," he explained. "When one of them gets bad data -- intentionally or unintentionally -- and changes its behavior, it can start performing its job incorrectly or with less precision, even if it was doing it perfectly well just one day before. An error in one agent can then have a cascading effect that degrades the whole system. Enterprises will hire as many AI systems engineers as it takes to keep that from happening." Companies and tech teams may be "well-positioned to support agentic AI, but we still need time and experience to strike the right balance between agentic and human workflows," Bennett advised. "Our advice is to view AI as an augmentation to human experts, not a replacement."

Three ways 2025 will be the year of agentic productivity

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More In the tech world, we like to label periods as the year of (insert milestone here). This past year (2024) was a year of broader experimentation in AI and, of course, agentic use cases. As 2025 opens, VentureBeat spoke to industry analysts and IT decision-makers to see what the year might bring. For many, 2025 will be the year of agents, when all the pilot programs, experiments and new AI use cases converge into something resembling a return on investment. In addition, the experts VentureBeat spoke to see 2025 as the year AI orchestration will play a bigger role in the enterprise. Organizations plan to make management of AI applications and agents much more straightforward. Swami Sivasubramanian, VP of AI and data at AWS, said 2025 will be the year of productivity, because executives will begin to care more about the costs of using AI. Proving productivity becomes essential, and this begins with understanding how multiple agents, both inside internal workflows and those that touch other services, can be made better. "In an agentic world, workflows are going to be reimagined, and you start asking about accuracy and how do you achieve five times productivity," he said. Palantir chief architect Akshay Krishnaswamy agreed that decision-makers, especially those outside of the technology cluster, are beginning to get antsy about seeing the impact these AI investments will have on their businesses. "People are rightfully fatigued about more sandboxing, because it's off the back of the whole data and analytics journey of the past 10 years, where people also did a ton of experimentation," said Krishnaswamy. "If you're an executive, you're like, 'this has to be the year I actually start to see some ROI, right?'" An explosion of orchestration frameworks Going into 2025, there is a greater need to create infrastructure to manage multiple AI agents and applications. Chris Jangareddy, a managing director at Deloitte, told VentureBeat that next year will be very exciting. Competitors will face LangChain and other AI companies looking to offer their own orchestration platforms. "A lot of tools are catching up to LangChain, and we're going to see more new players come up," Jangareddy said. "Even before organizations can think about multiagents, they're already thinking about orchestration so everyone is building that layer." Many AI developers turned to LangChain to start building out a traffic system for AI applications. But LangChain isn't always the best solution for some companies, which is where some new options includingMicrosoft's Magentic or LlamaIndex come in. But for 2025, expect to see an explosion of even more new options for enterprises. "Orchestration frameworks are still very experimental, with LangChain and Magentic, so you can't be heads down for just one," said PwC global commercial technology and innovation officer Matt Wood. "Tooling in this space is still early, and it's only going to grow." Better agents and more integrations AI agents became the biggest trend for enterprises in 2024. As organizations gear up to deploy multiple agents into their workflows, the possibility of agents crossing from one system to another becomes more apparent. This is particularly true when enterprises are looking to demonstrate their agents' full value to executives and employees. Platforms like AWS's Bedrock, and even Slack, offer connections to other agents from Salesforce's Agentforce or ServiceNow, making it easier to transfer context from one platform to another. However, understanding how to support these integrations and teaching orchestrator agents to identify internal and external agents will become an important task. When agentic workflows become more complex, the recent crop of more powerful reasoning models, like OpenAI's recently announced 03 or Google's Gemini 2.0, could make orchestrator agents more powerful. However, all of this will be in vain if enterprises do not get their employees to actually use new AI tools in 2025. Don Vu, chief data and analytics officer at New York Life, told VentureBeat that the last-mile problem of employees often choosing more manual methods over AI will continue for the next year. "The last mile problem is something that we've all stubbed our toe on in 2024, and understanding that change management, business process reengineering stuff that's not maybe as sexy as building an agent that can do all these incredible things," said Vu. "It's harder to change human behavior than deploy an app."

AI Agents Are Becoming More Humanlike -- and OpenAI Is Launching a New One in January. Are Entrepreneurs Ready to Embrace the Future?

Bloomberg recently reported the anticipated launch of OpenAI's "Operator" agent, slated for January 2025. This development marks a significant milestone in the evolution of generative AI, a field that has rapidly advanced since OpenAI introduced ChatGPT in November 2022. Within just two years, the competitive landscape has expanded, with numerous players joining the race. As competition intensifies, the focus has shifted from basic chatbots to sophisticated AI agents capable of autonomously executing complex, multi-step tasks. Companies like Anthropic and Google have introduced their own AI agents, designed to handle diverse workflows with seamless integration into daily operations. Meanwhile, Microsoft's autonomous agents in Copilot studio are pushing the boundaries for enterprises by allowing customers to build their own AI agents. According to Microsoft, McKinsey & Company is working on an AI agent designed to streamline client onboarding, with early trials indicating it can cut administrative tasks by 30%. This shift signals a new chapter in AI innovation, where intelligent agents are poised to redefine productivity and transform the way businesses operate. Related: Microsoft Strikes Back at Salesforce, Announces New AI Agents That Can Take Over Finance, Sales, and Service Tasks To understand the role of AI agents, it's essential to grasp their capabilities and how they differ from traditional chatbots. AI agents offer advanced reasoning, adaptability and the ability to perform human-like tasks. Unlike basic chatbots, which are often limited to predefined tasks, AI agents can autonomously execute complex workflows and perform a wide range of tasks within a structured framework. Powered by machine learning, natural language processing and automation technologies, AI agents adapt to diverse scenarios and improve through continuous learning. For instance, while a chatbot might help find flights and hotels using platforms like Kayak, an AI agent can book the trip, process the payment, organize the itinerary and even reschedule plans if conflicts arise. They can also draft professional emails, manage calendars and integrate seamlessly with other tools to streamline workflows. In business contexts, AI agents take on roles traditionally requiring multiple human resources. Startups, for example, can leverage them to act as data scientists by collecting and analyzing datasets to derive actionable insights. They can also provide predictive analytics, assist with marketing automation, enhance customer relationship management, optimize supply chains and handle financial planning. This versatility makes AI agents indispensable collaborators for businesses seeking efficiency and innovation in a digital-first world. Related: What You Need to Know About 'AI Agents' and Why We Are One Step Closer to The Jetsons In an era dominated by AI agents, their capabilities extend far beyond simple automation. These agents have the potential to serve as invaluable co-founders for startups by taking on critical responsibilities across the business lifecycle. They can brainstorm innovative ideas, conduct market research, develop strategies, handle complex coding tasks, build and maintain websites and create compelling content for digital platforms. Operationally, AI agents can manage client communications, oversee customer satisfaction surveys, analyze feedback, calculate ROI and optimize supply chain processes in real time. With predictive capabilities, they can ensure inventory and logistics are managed efficiently. When designed and trained effectively, AI agents can perform virtually any task within a company, provided clear frameworks and objectives are established. Entrepreneurs can leverage these advanced tools as collaborative partners, enabling startups to scale rapidly, innovate effectively and focus human resources on high-value, strategic initiatives. In this way, AI agents can truly act as co-founders, driving businesses toward success. Related: You Have 2 Months to Prepare Your Business for AI Agents. Here's Why! New technologies inherently bring both opportunities and challenges, and AI is no exception. For entrepreneurs, the rewards are potentially transformative, but the risks should not be overlooked. For instance, when MIT students tasked ChatGPT-4 with designing a hypothetical pandemic, the AI not only conceptualized the scenario but also outlined steps to execute it. This underscores the double-edged nature of AI -- similar to how the internet revolutionized access to information while also introducing risks like harmful content. AI's risks, including spreading misinformation or generating harmful outputs, highlight the need for caution. However, these risks are not reasons to reject innovation, but rather to approach it responsibly. Entrepreneurs, by nature, are calculated risk-takers. By equipping themselves with the tools, knowledge and strategies to manage potential downsides, they can embrace AI agents as transformative assets. Preparedness, vigilance and a commitment to ethical use are key to leveraging AI's full potential while safeguarding against its pitfalls. One popular piece of advice by venture capitalists for entrepreneurs solving problems with AI is to ensure that technology updates every three to four months to help them advance their products easily. If the product gets better with each update, it indicates that a specific problem is being effectively addressed within a specific niche, leveraging unique data for model training. On the flip side, constant concerns about the next big thing might suggest that the solution is not unique enough and needs reevaluation. The same applies to building AI agents; the focus should be on solving a specific problem that can be tackled using AI agents. Related: 'Human-Capable' AI Agents Will Change the Workforce Within 3 Years, According to a CEO Currently Creating the 'Perfect' AI Employee We stand on the threshold of a transformative era where work and innovation are being redefined. AI agents are evolving from tools to collaborators, taking on roles akin to task force members, coworkers or even digital co-founders. While these advancements promise unparalleled efficiency and innovation, they also demand a balance between automation and the human touch. By fostering collaboration between humans and AI, businesses can unlock extraordinary potential, scaling rapidly and achieving breakthroughs once deemed impossible. Whether accelerating success or expediting valuable lessons from failure, this partnership is set to redefine productivity, creativity and the way businesses grow in the future.

Unlocking value from data: How AI agents conquered 2024

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More If 2023 was the year of generative AI-powered chatbots and search, 2024 was all about AI agents. What started from Devin earlier this year grew into a full-blown phenomenon, offering enterprises and individuals a way to transform how they work at different levels, from programming and development to personal tasks such as planning and booking tickets for a holiday. Among these wide-ranging applications, we also saw the rise of data agents this year -- AI-powered agents that handle different types of tasks across the data infrastructure stack. Some did basic data integration work while others handled downstream tasks, such as analysis and management in the pipeline, making things simpler and easier for enterprise users. The benefits were improved efficiency and cost savings, leading many to wonder: How will things change for data teams in the years to come? Gen AI Agents took over data tasks While agentic capabilities have been around for some time, allowing enterprises to automate certain basic tasks, the rise of generative AI has taken things entirely to the next level. With gen AI's natural language processing and tool use capabilities, agents can go beyond simple reasoning and answering to actually planning multi-step actions, independently interacting with digital systems to complete actions while collaborating with other agents and people at the same time. They also learn to improve their performance over time. Cognition AI's Devin was the first major agentic offering, enabling engineering operations at scale. Then, bigger players began providing more targeted enterprise and personal agents powered by their models. In a conversation with VentureBeat earlier this year, Google Cloud's Gerrit Kazmaier said he heard from customers that their data practitioners constantly faced challenges including automating manual work for data teams, reducing the cycle time of data pipelines and analysis and simplifying data management. Essentially, the teams were not short on ideas on how they could create value from their data, but they lacked the time to execute those ideas. To fix this, Kazmaier explained, Google revamped BigQuery, its core data infrastructure offering, with Gemini AI. The resulting agentic capabilities not only provide enterprises the ability to discover, cleanse and prepare data for downstream applications -- breaking down data silos and ensuring quality and consistency -- but also support pipeline management and analysis, freeing up teams to focus on higher-value tasks. Multiple enterprises today use Gemini's agentic capabilities in BigQuery, including fintech company Julo, which tapped Gemini's ability to understand complex data structures to automate its query generation process. Japanese IT firm Unerry also uses Gemini SQL generation capabilities in BigQuery to help its data teams deliver insight more quickly. But, discovering, preparing and assisting with analysis was just the beginning. As the underlying models evolved, even granular data operations -- pioneered by startups specializing in their respective domains -- were targeted with deeper agent-driven automation. For instance, AirByte and Fastn made headlines in the data integration category. The former launched an assistant that created data connectors from an API documentation link in seconds. Meanwhile, the latter enhanced its broader application development offering with agents that generated enterprise-grade APIs -- whether it's for reading or writing information on any topic -- using just a natural language description. San Francisco-based Altimate AI, for its part, targeted different data operations including documentation, testing and transformations, with a new DataMates tech, which used agentic AI to pull context from the entire data stack. Multiple other startups, including Redbird and RapidCanvas, also worked in the same direction, claiming to offer AI agents that can handle up to 90% of data tasks required in AI and analytics pipelines. Agents powering RAG and more Beyond wide-ranging data operations, agentic capabilities have also been explored in areas such as retrieval-augmented generation (RAG) and downstream workflow automation. For instance, the team behind vector database Weaviate recently discussed the idea of agentic RAG, a process allowing AI agents to access a wide range of tools -- like web search, calculator or a software API (like Slack/Gmail/CRM) -- to retrieve and validate data from multiple sources to enhance the accuracy of answers. Further, towards the end of the year, Snowflake Intelligence appeared, giving enterprises the option to set up data agents that could tap not only business intelligence data stored in their Snowflake instance, but also structured and unstructured data across siloed third-party tools -- such as sales transactions in a database, documents in knowledge bases like SharePoint and information in productivity tools like Slack, Salesforce and Google Workspace. With this additional context, the agents surface relevant insights in response to natural language questions and take specific actions around the generated insights. For instance, a user could ask their data agent to enter the surfaced insights into an editable form and upload the file to their Google Drive. They could even be prompted to write to Snowflake tables and make data modifications as needed. Much more to come While we may not have covered every application of data agents seen or announced this year, one thing is pretty clear: The technology is here to stay. As gen AI models continue to evolve, the adoption of AI agents will move at full steam, with most organizations, regardless of their sector or size, choosing to delegate repetitive tasks to specialized agents. This will directly translate into efficiencies. As evidence of this, in a recent survey of 1,100 tech executives conducted by Capgemini, 82% of the respondents said they intend to integrate AI-based agents across their stacks within the next 3 years -- up from a current 10%. More importantly, as many as 70 to 75% of the respondents said they would trust an AI agent to analyze and synthesize data on their behalf, as well as handle tasks such as generating and iteratively improving code. This agent-driven shift would also mean significant changes to how data teams function. Currently, agents' outcomes are not production-grade, which means a human has to take over at some point to fine-tune the work for their needs. However, with a few more advancements over the coming years, this gap will most likely go away -- giving teams AI agents that would be faster, more accurate and less prone to the errors usually made by humans. So, to sum up, the roles of data scientists and analysts that we see today are likely to change, with users possibly moving to the AI oversight domain (where they could keep an eye on AI's actions) or higher-value tasks that the system could struggle to perform.

AI Agents in 2025: How Can They Transform Industries Worldwide

These systems can make decisions without human intervention After the novelty of generative artificial intelligence (AI) wore off, many raised an important question -- Yes it is cool, but how can it make an impact in the real world? It was a valid question. While AI chatbots can be seen as a one-stop-shop for quickly looking up information, having an impromptu conversation, making it write essays, generate images or videos, their role is largely limited to a system where a human user will have to constantly command it to get an output and oversee the result. Even if its capabilities cannot be dismissed, and it did make a significant impact in improving workers' productivity in certain areas, it lacked one critical element that stopped it from becoming a faithful assistant that could handle and truly automate tasks -- decision-making. Generative AI today can help with certain aspects of a person's work, but it cannot execute a task. For instance, you can ask it to write an email to a client letting them know about an unexpected delay, but it cannot send that message or deal with the angry reply they send. Similarly, you can use Gemini or ChatGPT to ask for "the best smartphone for shooting videos", and it can recommend the latest iPhone 16 Pro Max or the Samsung Galaxy S24 Ultra. But it will not be able to scour the web to find you the best deal and make a purchase. Realising this gap, tech companies working on large language models (LLMs) began using the word AI agent. Researchers believe AI agents can take a knowledge-based AI system and transform it into an action-taking system that can perform end-to-end tasks without human intervention. The term gained prominence during the second half of 2024, and currently, it is being treated as a panacea for all work-related problems. And while there is some truth to it, is it really a transformative technology of that potential? The answer might be a bit complex, but we will do our best to break it down and highlight all the different aspects that you should know about. Let us dive into it. Since this technology is still in its nascent stage, there is no unified definition of what exactly constitutes an AI agent. IBM defines it as a system "that is capable of autonomously performing tasks on behalf of a user" by designing a workflow and using tools. Similarly, Google, which announced its first AI agent dubbed Project Mariner last year, calls it a system that acts like an assistant to humans and helps them complete tasks. A more comprehensive definition is given by Amazon, which describes it as "a software programme that can interact with its environment, collect data, and use the data to perform self-determined tasks to meet predetermined goals. Humans set goals, but an AI agent independently chooses the best actions it needs to perform to achieve those goals." Put simply, an AI agent can be understood as an AI system that can take action instead of just telling the user about the action. A typical AI agent will have a large language model (LLM) as its brain. But it will also include other elements that enable it to use that intelligence in actions. Most commonly, these extra parts are different sensors, mechanical parts, encoders, or integration into other software. The sensors enable an AI agent to collect data across different formats. These can be visual, sound, temperature, or electronic signals. Mechanical parts are typically used for embodied AI or robots which need to execute real-world actions such as lifting an object or moving from one spot to another. Encoders are used to convert different types of signals into information that can be processed by LLMs. Finally, software integration enables the ability to execute tasks. It is also important to highlight another crucial difference between AI models and AI agents at this point. AI models contain a pre-training database which forms the basis of their knowledge. Anything that is not part of the database will not generate an output. A good example of this was the early version of ChatGPT which was not connected to the Internet and had a knowledge cut-off date. If it was prompted to answer a question about current affairs, it would not be able to answer that. On the contrary, AI agents, when integrated with relevant systems, can independently collect new data to solve problems that would not be possible based on their existing database. For instance, Google's Project Mariner can interact with the browser to find the best deal on a smartwatch. Another aspect of AI agents is the capability to handle complex tasks. AI agents are capable of advanced reasoning and as such can break down a complex task into multiple easier tasks and then complete them one after another. This contextual understanding of the problem and the ability to know how to break it down is a fundamental function of AI agents. A good example of this is Gemini's recently added Deep Research tool. Users can ask it to explain a technical or niche topic. The AI would then create a multi-step research plan, break down the topic into smaller parts, find relevant research papers and articles on the topic, execute the plan, conduct research, and analyse the gathered data to create a detailed report. AI firms have been touting AI agents as a tool that can be used across industries and in different scenarios. It can be used as a voice assistant for devices that can perform device-specific tasks (such as taking a picture or playing music). It can be added to an app or software and carry out tasks within that (purchasing a product via a browser-based agent). It can also be added to enterprise systems and it can detect fraud or find ways to optimise different processes. Apart from this, AI agents are also said to perform transformative tasks in certain industries. In healthcare, it can be used for diagnosis, treatment recommendation, and drug discovery. In the automotive sector, it can be used to create self-driving cars. AI agents are also said to be able to pilot drones in disaster areas to gather and analyse data and offer actionable insights for rescue operations. It also has applications in manufacturing industries via AI-powered robots, in the gaming industry as a game developer or as a non-playing character (NPC) inside games, and in the education sector to create personalised study plans and to grade test papers in a human-like fashion. However, it is important to note that while tech companies are marketing AI agents as a catch-all for all kinds of end-to-end intelligent automation, the current technology limits its use case to largely specific task-based roles instead of a general-purpose tool. With that being said, it is important to ground our expectations and understand what we can realistically expect from AI agents in the current year. It is unlikely that AI agents will enter the workforce in any of the critical sectors such as manufacturing, automobiles, healthcare, or education. However, this year should mark the entry of AI agents in consumer electronics, mobile and desktop applications, as well as websites and platforms. Google's Project Mariner, for example, could be integrated with Google Chrome and assist users in making purchases and finding files from the web by the end of this year. OpenAI is also rumoured to launch its AI agent this year which could further enhance ChatGPT's capabilities and allow it to perform certain actions on a user's device and the Internet. Anthropic's Computer Use tool is also expected to make a global release and assist users in their day-to-day tasks on the device. Eventually, we should also see a shift where AI agents can mimic keystrokes, mouse movements and clicks, and do much more on devices. For example, by the end of the year, more agentic tools such as the coding agent Devin could be writing code end-to-end, testing them, finding and fixing glitches, and deploying them without human intervention. But, it would be highly optimistic to include this in the 2025 itinerary. On the enterprise side, AI agents could take a larger role in completing some organisational tasks such as monitoring a large volume of data, preparing analytical reports, and offering recommendations and course corrections. It could also be used in some cybersecurity roles. Notably, Meta has stated that it already uses AI to ensure guidelines are being followed. YouTube also uses AI to monitor copyright violations. However, we do not expect AI agents to enter any of the critical work functions this year because the technology is largely untested and its reliability will be questionable. Businesses, specifically public enterprises or those backed by large investors, are generally risk-aversive and are unlikely to provide access to sensitive data. With AI being the current trend in the tech space and the potential to disrupt a large number of industries, it is understandable why there is so much excitement about AI agents. However, beyond the rose-tinted glasses exist several issues with AI agents that need to be addressed before the technology can witness large-scale adoption. On the other hand, if it goes unchecked, the technology can pose several risks. One of the main issues with AI agents is bias and discrimination which comes from their training data and can lead to discriminatory outcomes. This also highlights another issue of transparency in AI agents. With complex algorithms and architecture, most AI agents are complicated and opaque systems where it is difficult to understand how and why they make certain decisions. There are security and privacy issues as well. From a security perspective, AI agents can be vulnerable to adversarial attacks, where malicious actors manipulate input data to deceive the system. Additionally, since AI agents need to be connected with multiple systems and collect a large amount of data to carry out tasks, they also pose privacy risks. With so many challenges, AI firms will have a tough job ahead to convince enterprises and individuals of the upside of the technology while reassuring them of the downside. Regardless, it cannot be denied that AI agents will constitute a large part of AI announcements in 2025.

Solution Providers Prepare For 2025's AI Agent Era

'We're learning how to capitalize on taking that human time, going down, and leverage agents to do as much as possible,' says Corey Kirkendoll, CEO of 5K Technical Services. Retrieving company policies, persuading customers to upgrade their purchases and interviewing job candidates are some of the use cases for artificial intelligence agents solution providers are leveraging internally and exploring for customers, partners tell CRN. Executives with AI leaders such as Microsoft, Google and Salesforce have started talking about this next part of the AI journey-which is already having an influence on enterprise technology by some measures. Salesforce reported in December that AI and agents influenced $60 billion in sales during the 2024 Cyber Week shopping season. Deloitte said in a November report that a quarter of companies using GenAI will launch agentic AI pilots or proofs of concept (PoCs) in 2025, with the share reaching 50 percent in 2027. Gartner in October said that 15 percent of day-to-day work decisions will be made autonomously through agentic AI by 2028, with 33 percent of enterprise software applications including agentic AI. [RELATED: The 10 Biggest AI News Stories Of 2024: Nvidia, GenAI And Security] Corey Kirkendoll, CEO of Plano, Texas-based 5K Technical Services-a member of CRN's 2024 Managed Service Provider 500-told CRN in an interview that he has leveraged agents for employees seeking policies around time off and procedures plus speeding up the hiring process by pre-qualifying candidates for technology specialist jobs. AI copilots are still an important part of the AI adoption journey, with customers needing to master the more general AI use cases before adding in autonomous AI agents to perform a more complicated task, he said. In the future, Kirkendoll hopes to use agents to create marketing material, onboard new 5K employees and for some self-help customer service. "We're learning how to capitalize on taking that human time, going down, and leverage agents to do as much as possible," he said. Here's what six solution providers told CRN about how agents mark a new evolution in the AI era.

AI Solution Providers Prepare For The Agent Era

"It was a bit of an epiphany for him," Grant Davies, principal of digital strategy at Perficient, said about showing a customer how AI agents talk to each other. Grant Davies, principal of digital strategy at Perficient, told CRN in an interview that customers can hardly believe it when he demonstrates artificial intelligence agents "talking to each other"--that is, one agent generates content and one or more agents help improve the content. "It was a bit of an epiphany for him," Davies recalled of one customer seeing the demo. "A bit of a mind blow." Perficient-based in St. Louis and No. 56 on CRN's 2024 Solution Provider 500- is among the camp of solution providers ready to bring customers into the agentic phase of generative AI while exploring ways to leverage the technology internally to improve operations. [RELATED: The 10 Biggest AI News Stories Of 2024: Nvidia, GenAI And Security] Executives with AI leaders such as Microsoft, Google and Salesforce have already started talking about this next part of the AI journey-which is already having an influence on enterprise technology by some measures. Salesforce reported in December that AI and agents influenced $60 billion in sales during the 2024 Cyber Week shopping season. Deloitte said in a November report that a quarter of companies using GenAI will launch agentic AI pilots or proofs of concept (PoCs) in 2025, with the share reaching 50 percent in 2027. Gartner in October said that 15 percent of day-to-day work decisions will be made autonomously through agentic AI by 2028, with 33 percent of enterprise software applications including agentic AI. Carlos Marques, technology director at Tallahassee, Fla.-based Mainline Information Systems-a member of CRN's The 2024 MSP 500-told CRN his company is looking at AI agents-a more autonomous version of GenAI that completes directed tasks-as a way to lower the cost of delivery for managed services, increase productivity without needing to hire more people and to take over repetitive mundane tasks from employees. The technology is still early, however, Marques said. Integration work can take months and some customers aren't ready to trust AI agents with sensitive information. "Everybody understands the potential," Marques said. "But there's also going to be a fight of, is it more or less risky to let the agent do privileged things?" Amas Tenumah, global leader of service transformation and AI innovation at Seattle-based Slalom-No. 27 on CRN's 2024 Solution Provider 500-told CRN in an interview that he likens agents to an intern given tasks to complete by the user. The solution provider has dozens of agents in production, in industries ranging from health care to retail. Use cases have included customer service, even issuing customers credit after an issue, Tenumah said. Agents have also shown better ability at upselling customers, proving more persuasive than human counterparts on occasion. Agents hold the promise of a interface for subject matter experts (SMEs) with no coding experience to now execute complicated tasks once reserved for programmers. "People who have humanities degrees, this is your time now to shine," he said. "You will be the next developers who are going to be making the big bucks. Even my kids, (I tell them) don't bother coding anymore." Mike Strohl, CEO of Concord, Calif.-based e360-No. 128 on CRN's 2024 Solution Provider 500-told CRN in an interview that he is seeing success leveraging agents with human resources (HR) use cases, such as allowing workers to search for company policies. Where solution providers remain key to introducing AI agents and any new technology to the world is translating that technology to the industries and customer problems partners are more intimately familiar with compared to vendors, Strohl said. "You really have to be able to speak the language of the client, which means you have to understand their business-their applications, their people, their data, their customers, their outcomes-if you're going to provide value from an AI standpoint," Strohl said. "Otherwise you're not going to get very far." Corey Kirkendoll, CEO of Plano, Texas-based 5K Technical Services-a member of CRN's 2024 Managed Service Provider 500-told CRN in an interview that he has leveraged agents for employees seeking policies around time off and procedures plus speeding up the hiring process by pre-qualifying candidates for technology specialist jobs. AI copilots are still an important part of the AI adoption journey, with customers needing to master the more general AI use cases before adding in autonomous AI agents to perform a more complicated task, he said. In the future, Kirkendoll hopes to use agents to create marketing material, onboard new 5K employees and for some self-help customer service. "We're learning how to capitalize on taking that human time, going down, and leverage agents to do as much as possible," he said. Eric Walk, principal for enterprise data strategy at St. Louis-based Perficient, told CRN in an interview that the solution provider is leveraging agents for use cases such as orders status checks. The agents interpret the context of a user's request and which application programming interfaces (APIs) to call to perform the task. Within Perficient, AI agents are writing the first draft of user stories for consultants to leverage, and others are using paired programming AI to code. Agents are writing the first draft of request for proposal (RFP) responses and automating the production of case studies after project completions, Walk said. AI copilots will continue on for more general use cases, Walk said. Customers adopting AI copilots will still look to solution providers for change management and achieving more value from the technology. Agentic frameworks take humans out of the process more so compared to copilots. He said he hopes that AI vendors don't over promise on what agents can deliver out of the box and that they remember the importance of cross-platform integrations and interactions for enterprises with a hodgepodge of vendors and data sources. But complicated technology is where the channel thrives. "The No. 1 challenge to this is adoption and ROI realization," Walk said. "They're cool tech, but they've got to do something meaningful." Perficient's Davies said that agentic AI still has to cut down on the level of incorrect answers and hallucinations to become production ready, but he expects improvements in the future. For now, he sees an important part of success in keeping customers' agent expectations realistic. The agents aren't cure-alls, but the savings they help find even in incremental use are appetizing to customers. "If you can use individual agents or agentic frameworks to pull an hour out of somebody's day, and you do that three times a week or five times a week, and you do it for 50 people, that's real money," he said.

Rise of AI agents: powerful assistant or Pandora's box?

Technology companies are actively seeking new ways to leverage artificial intelligence (AI) technologies, leading to the emergence of AI agents, which are not only able to generate content but are also able to take action based on the information they gain from their environment. According to the Financial Times (FT), OpenAI CFO Sarah Friar predicts that AI agents will become the buzzword in 2025. The first batch will be deployed as researchers or assistants to help people manage daily tasks. "I think 2025 is going to be the year that agentic systems finally hit the mainstream," OpenAI's new chief product officer Kevin Weil, said at a press event ahead of the company's annual Dev Day, according to The Verge. What are AI agents? AI agents take generative AI (GenAI) to a new level, can independently perform tasks, make decisions, and learn from their experiences. According to Adnan Ijaz, director of product management for Amazon Q Developer, and Yoon Kim, an assistant professor at MIT's Computer Science and Artificial Intelligence Laboratory. AI agents usually follow a three-part workflow: 1. The user sets a goal and provides a prompt. 2. AI agents approach the task by breaking it down into smaller, simpler subtasks and collecting the needed data. 3. AI agents execute tasks using what's contained in their knowledge base plus the data they've amassed, making use of any functions they can call or tools they have at their disposal. For instance, if the user requests to book a flight, the AI agent will first search for all flights that meet the specified criteria, then select the most affordable option, and proceed to make the booking through the airline's application programming interface (API). However, humans can still guide the process and intervene when required. For example, if the cheapest flight has no available seats, the AI agent will notify the user, allowing them to decide on the next step. AI Agents are poised to become the next golden goose Jared Friedman, a partner at the Silicon Valley startup accelerator Y Combinator, believes that AI agents have the potential to surpass SaaS, with significant advantages in reducing costs while boosting efficiency. Friedman advises entrepreneurs to identify highly repetitive and tedious administrative tasks and leverage AI agents to automate these processes. AI agent tool use In fact, several tech companies have already shifted their business focus and conducting preliminary tests of related functionalities in 2024. Google Google made a surprise move just before the year-end holidays by releasing the first model of its Gemini 2.0 series, Gemini 2.0 Flash, and plans to expand the availability of Gemini 2.0 into Google's products by 2025. In addition, Google showcased several AI agent prototypes built on Gemini 2.0, including Project Astra, Project Mariner, and Jules. Project Astra has the ability to converse in multiple languages and execute multi-step online tasks; Jules is designed to assist developers with coding tasks. Microsoft Microsoft announced the launch of a series of pre-built AI agents within Microsoft 365 at Ignite 2024. For instance, the project manager agent in Planner automatically assigns tasks, tracks progress, and sends notifications, while the coordinator agent in Teams provides real-time meeting notes and shares key information summaries. Users can also create custom agents using the low-code platform, Copilot Studio. Salesforce SaaS company Salesforce launched its AI agent platform, Agentforce, in October. During its most recent earnings call, Salesforce revealed that it had secured multiple orders and plans to hire over 1,000 new employees to drive sales of the platform. Startups are also pouring resources into R&D to capture market share. OpenAI In October, OpenAI introduced Swarm, an experimental multi-agent coordination framework, with its design centered around two core concepts: "routine" and "handoff." The former refers to a group of agents following instructions to complete specific tasks, while the latter allows for seamless transitions between agents specialized in distinct functions. For example, in a customer service system, a triage agent conducts an initial assessment and forwards specific queries to agents with expertise in sales, support services, or refunds. Furthermore, OpenAI is preparing to launch an AI agent called 'Operator' in January 2025, granting researchers access to its API, Bloomberg reported. Perplexity In early December 2024, Perplexity launched the "Buy With Pro" AI shopping tool, designed to assist users in searching for products online, comparing prices, and completing purchases directly. Perplexity has also partnered with digital payment provider Stripe, which offers the AI agent a one-time charge card for online payments. This approach allows the AI agent to complete transactions without needing access to the user's bank account, thereby reducing potential risks. However, the service is still facing challenges, including slow processing speeds and occasional transaction failures, requiring human intervention to ensure accuracy. In October 2024, Anthropic introduced a "computer use" feature, based on the upgraded Claude 3.5 Sonnet, enabling AI to interpret screen displays and control computers like a human. Challenges to adoption David Singleton, former VP of Engineering at Google Android, established a company called /dev/agents, focused on creating operating systems for AI agents. Singleton pointed out that the current difficulty in developing AI agents is too high, and he aims to build an AI agent development platform similar to Android, simplifying the development process for developers. Beyond technical challenges, data privacy and security represent major obstacles for AI agents. These agents will access, analyze, and collect vast amounts of personal data. If targeted by malicious actors, personal information could be exposed and compromised. Moreover, since AI agents are based on large language models (LLMs), hallucinations are a concern. Additionally, with AI agents handling tasks directly for consumers and bypassing the need to visit websites, this could disrupt the revenue streams and business models of retailers, publishers, and advertisers. As AI agents interact more frequently with one another and humans are increasingly excluded, ensuring trust and accountability will be top of the concerns.