Agentic AI has taken middle stage within the worlds of AI, tech, and enterprise, dominating the discourse and furthering the strain for firms to swiftly combine the tech or fall behind their opponents. Greater than anybody, it’s chief AI officers (CAIOs) who’re charged with untangling the guarantees and realities of AI’s newest buzzword.
As they oversee experimentation with and rollouts of AI brokers and information different leaders on the journey, CAIOs are additionally navigating by way of the hype, considerations round safety and belief, and interconnectedness (or lack thereof) of those techniques. To not point out having to grapple with the query: What even is an AI agent?
Hype-chasing causes firms to lose focus
Nobody can appear to agree on what, precisely, the time period “AI agent” actually means, as Fortune and others have reported. Corporations are defining the time period in another way and sometimes utilizing it to explain assorted options and capabilities, together with many who have been beforehand described with different phrases reminiscent of “AI assistants.” For Accenture chief AI officer Lan Guan, who led the construct of an AI agent answer referred to as Refinery AI for purchasers and in addition works straight with them on their very own AI and AI agent deployments, this has prompted her to commit quite a lot of time to only serving to purchasers type by way of the contradictions.
“A year ago, everyone was saying, ‘I need to do gen AI.’ Now everyone is saying, ‘I need to do agentic AI or AI agents.’ And it’s like, at the end of the day, a lot of these things are still the same thing. They’re just getting called different things depending on who you’re talking to,” she mentioned. “And so there’s a ton of confusion in the marketplace with our clients on, ‘What is an AI agent? What am I deploying?’ And so we spend a lot of time on education.”
A runaway impact of this has been firms shortly spinning up so-called AI brokers “just for the press release,” says Michelle Bonat, chief AI officer of AI Squared, who additionally works with firms throughout regulated industries on their AI growth. The strain to have a solution for the agentic AI second is inflicting some firms to rename options or chase AI brokers to remain on pattern, typically merely creating skinny layers of brokers on high of basis fashions.
“I’m totally seeing that. I’m seeing that every day,” Bonat says. “That’s why this space is full of noise.”
Safety, errors, and belief dominate the chance evaluation
Regardless of the hype and muddled terminology, the core thought of AI brokers—techniques designed to autonomously take motion to hold out particular duties—remains to be producing plenty of justifiable pleasure. It’s additionally key to creating the forms of techniques technologists and science fiction lovers have at all times dreamed of, able to executing sequences of complicated duties throughout a number of platforms on our behalf. However there are actual roadblocks.
Uri Yerushalmi, cofounder and chief AI officer at Fetcherr, which makes use of AI for predictive pricing within the airline trade, believes the alternatives round AI brokers are “enormous” however that unlocking that worth relies on addressing actual challenges round belief and integration and avoiding failure factors. For instance, brokers should combine with legacy techniques and align with real-world constraints with out disrupting present workflows. And as we give brokers extra autonomy, we have to construct guardrails, monitoring, override techniques, and mechanisms for human alignment, he mentioned.
“Businesses need to trust the agent’s decisions,” he added. “That requires transparency, consistency, and demonstrable ROI.”
Some of the regarding failure factors is compounding errors. Google DeepMind CEO Demis Hassabis has in contrast this concern to compound curiosity in funds, explaining that even when an agentic mannequin has solely a 1% error, it will trigger a series response of errors that will, after a couple of thousand steps, in the end make the chance of an accurate outcome utterly “random.” Bonat factors to this drawback of compounding errors as a extreme problem by way of trusting AI brokers, saying this potential to compound one misstep with out people even being conscious of it might “create havoc.”
That is very true for the type of multi-agent techniques many companies are considering, which Guan mentioned could cause blind spots and get you into bother in a short time.
“It may not work for you, and may actually introduce a lot of risk,” she mentioned. “Think about it—a lot of the business workflows and transactions or interactions are high stakes. You don’t want agents to just issue a refund for every customer, right?” she mentioned, including that whereas her purchasers have a powerful urge for food to see affect from AI brokers, they’re additionally cautious of shock excessive cloud payments and safety dangers.
Safety considerations are definitely high of thoughts within the AI agent panorama. By 2028, Gartner predicts, 25% of enterprise breaches might be traced again to AI brokers, together with abuse from each inside and exterior malicious actors. The dominating issue contributing to safety dangers is the mix of autonomy and supposed interoperability of agent techniques, which might have them connect with, change knowledge with, and autonomously act throughout a large swath of platforms and techniques. Put in another way, the precise nature of how these techniques perform and what they’re supposed to do is what makes them so dangerous.
Interoperability goals wrestle to interrupt free from walled gardens
Like all CAIOs, Ali Alkhafaji, chief AI and know-how officer at Omnicom Precision Advertising Group, is anxious about knowledge leakage and different safety dangers. He’s additionally involved that lots of the firms commercializing agent techniques are utilizing safety as a handy excuse to additional lock their prospects inside their ecosystems, going in opposition to the collaborative and decentralized imaginative and prescient many see as intrinsic to an agentic future: “Not because it can’t be solved, but because it’s not in the commercial interest of the vendor to solve it.”
“Every vendor is building their own ‘agent framework,’ but no one is solving for enterprise-level interoperability. Without open frameworks and semantic standards, we’re just building smarter silos,” he mentioned, including that agent collaboration protocols stay immature and that it’s irritating to see main distributors and hyperscalers proceed to bolster walled gardens.
Deloitte U.S. head of AI Jim Rowan is seeing this play out amongst his purchasers, noting that they’re principally sticking with their present suppliers and utilizing their agent capabilities as they’re launched. It’s one other iteration of the platform benefit that’s driving development for suppliers like OpenAI, Google, and Microsoft as they onboard their present prospects into their new AI pipelines and merchandise.
“There is a definite tension in the marketplace between who wants to own the agent system of record. Like, who’s gonna own the registry, who’s going to orchestrate all the orchestration that’s happening around agents,” mentioned Rowan. “We see that with the hyperscalers and the SaaS providers and the third-party-tool startups that are in the space as well. I think the jury’s still out on who’s owning that.”
Correction, Sept. 3, 2025: An earlier model of this story misstated the identify of Accenture’s AI agent answer.
Fortune World Discussion board returns Oct. 26–27, 2025 in Riyadh. CEOs and world leaders will collect for a dynamic, invitation-only occasion shaping the way forward for enterprise. Apply for an invite.
