Top CEOs Share Their Insights on Agentic AI and the Future of IT

Agentic AI is moving from buzzword to boardroom priority. Top tech CEOs now talk less about isolated AI features and more about autonomous agents that plan, act, and continuously support employees and customers. Their shared message is clear: IT will evolve from managing apps to orchestrating fleets of AI agents that sit on top of data, workflows, and infrastructure. This article pulls together public comments from leaders at Microsoft, Google, Salesforce, NVIDIA, and others, then translates their vision into practical implications for CIOs and IT leaders.

Introduction

If you listen carefully to how big-tech CEOs talk in 2025, the language has shifted. A few years ago, everything was about “AI features” and “chatbots”. Now you hear different terms: agents, copilots, digital workers, autonomous workflows, agentic operating models.

What has changed is not just technology but the entire mental model of people as a collective.

Instead of AI as a tool you occasionally query, leaders are increasingly picturing AI as software that can pursue goals, take actions across systems, and work alongside people like a digital colleague.

That, in simple terms, is what “agentic AI” usually means: AI that can understand a high-level objective, decide what needs to be done, and execute multi-step tasks across applications, with humans supervising rather than micromanaging.AP News+1

For IT leaders, this shift is as big as cloud was fifteen years ago. Below, we will look at how some of the most influential CEOs are framing the agentic era, then translate their perspectives into concrete guidance for the future of IT.

First, a quick grounding: why does “agentic AI” matter?

Consulting and research firms have started to quantify what CEOs are already feeling. IBM describes agentic AI as a new operating model where AI systems make more of the everyday decisions, while humans concentrate on the decisions that matter most.IBM Capgemini and Google Cloud highlight research suggesting that roughly one third of enterprise applications could include agentic capabilities by 2028, up from almost none in 2024.

At the same time, Gartner warns that more than 40 percent of current agentic AI projects may be scrapped by 2027 due to cost and unclear business value, and calls out “agent washing” where vendors label ordinary automation as agents. McKinsey’s technology trends reports note that boards are asking about agentic AI, but most companies are still at pilot stage, not scale.

So, CEOs are simultaneously excited and cautious. They see an architectural shift coming, but they also know much of today’s hype will not survive first contact with reality. That tension shows up clearly when you listen to individual leaders.

“Agentic AI isn’t about answering questions; it’s about executing tasks. Instead of telling you how to reset your password, an agent should trigger the reset, monitor completion and reinforce it in chat.” — Saurabh Kumar, CEO of Rezolve.ai

Marc Benioff: Agentic AI as the new customer and support engine

Salesforce CEO Marc Benioff has talked openly about what happens when you put AI agents at the center of customer operations instead of treating them as an add-on. In one widely cited conversation, he described “agentic AI” to thrill customers and transform business operations, not just reduce costs.

He has also gone further than many by revealing hard numbers. Salesforce has replaced about four thousand customer support roles with AI agents that now handle roughly half of all interactions, while maintaining customer satisfaction. Rather than simply deflecting tickets, these agents triage, respond, and escalate in ways that are starting to feel like digital colleagues.

For IT, Benioff’s stance signals a future where support platforms are built around agents as first-class citizens. Ticketing, CRM, knowledge, and workflow tools are increasingly there to feed and monitor agents, not the other way around. The IT challenge becomes integration and governance, not just tool selection.

Satya Nadella: From apps to agents and a different SaaS landscape

Microsoft CEO Satya Nadella has been explicit about where he thinks software is going. In interviews and keynotes, he has argued that we are moving from a world of “apps” to a world of “agents”, with AI collapsing the logic of business applications into goal-driven copilots that orchestrate work across traditional app boundaries.

In one analysis of his comments, Nadella is quoted as predicting that AI agents will transform SaaS as we know it, turning classic software products into platforms that expose capabilities for agents to compose, rather than screens for humans to click through. He also talks about AI as a way to introduce “lean for knowledge work”, where every department’s processes are continuously optimized by digital helpers.

For CIOs and IT teams, this is more than marketing language. If agents really become the main user interface to work, then integration, data contracts, identity, and security become the new UI stack. IT architecture has to assume there is a programmable layer on top of every business system, where AI agents operate on behalf of people. Governance and observability will matter as much as UX did in the old app-centric era.

Sundar Pichai: AI agents as a new “form factor” and platform shift

Google CEO Sundar Pichai has been consistent for years in calling AI a technology as profound as fire or electricity. In 2025, he sharpened that view by describing AI agents as a “powerful new form factor” that will change how humans interact with technology. He has suggested that businesses may adopt agents even faster than consumers, simply because CIOs already need deep integration across systems, and agents are a natural way to orchestrate that.

In his Google I/O 2025 keynote, Pichai framed AI agents as part of a broader platform shift where computing becomes more ambient, more context aware, and less tied to individual screens or websites. Agents watch activity, understand intent, and step in with actions rather than waiting to be asked direct questions.

For IT, this implies a future where “the system of engagement” is no longer a single portal or app, but a mesh of agents that can sit in chat, browser, mobile, and line-of-business tools. IT will need to manage agent permissions, contexts, and data scopes with the same care they once applied to user roles and single sign-on.

Jensen Huang: Digital workers that turn data into action

NVIDIA CEO Jensen Huang often talks about AI in very pragmatic terms: digital workers, not just models. In a 2024 discussion about agentic AI, he emphasized that the next wave of enterprise transformation will be driven by AI agents that can take data, turn it into knowledge, and then turn that knowledge into action across sales, service, and back-office workflows.

The framing is important. Rather than positioning agents as “smarter chatbots”, Huang describes them as software employees that can run continuously, learn from logs, and collaborate with humans. NVIDIA’s strategy, unsurprisingly, ties this to the computing infrastructure and GPU stacks required to support huge numbers of such agents.

From an IT perspective, this draws attention to scale and performance. If every major business function has thousands of micro-agents running in the background, IT must think about scheduling, observability, and resource management in new ways. Monitoring will not just be about servers and APIs, but also about the behavior of agent populations.

Industry CEOs: From pilots to scaled agentic operations

Beyond big-tech, many enterprise CEOs are wrestling with a different question: how to move beyond pilots. For example, Chakri Gottemukkala, co-founder and CEO of o9 Solutions, has spoken about using agentic AI to transform enterprise planning and execution, and has stressed the need to move from slideware and small experiments to live demos that show real operational value.

McKinsey’s articles aimed at CEOs talk about “growing pains” in the agentic age, noting that boards are asking for agentic strategies while many organizations still struggle with data silos, unclear goals, and change management (Source: McKinsey & Company+2McKinsey & Company+2). IBM’s research on agentic operating models makes a similar point: the real shift is not just in tools but in decision rights, accountability, and process design.

The common theme: CEOs increasingly see agentic AI not as a feature but as a structural change in how work is organized and how decisions are made.

What all these CEO perspectives mean for the future of IT

If you step back and look at these leaders together, a pattern emerges. They are describing slightly different futures, but they rhyme.

You can summarize their combined message to IT in a few key ideas:

• Agents become the new front end. Instead of employees navigating dozens of interfaces, they talk to agents in chat, voice, or ambient interfaces that take action on their behalf.

• Apps become capabilities. SaaS products and internal systems are increasingly treated as collections of APIs, events, and workflows for agents to orchestrate.

• IT becomes an orchestrator and governor. The job shifts from shipping monolithic apps to exposing secure capabilities, curating data, and setting guardrails for agent behavior.

• Infrastructure becomes “agent aware”. Monitoring, security, and capacity planning all need to consider the behavior of fleets of agents, not just human users.

This is not a small change. It is a reframing of what it means to “run IT”.

How CIOs and IT leaders can respond in the next 12 to 24 months

Given the hype cycle, it is tempting either to ignore agentic AI or to jump in too fast. The CEOs above, plus the research from Gartner, IBM, and McKinsey, all point toward a more balanced approach.

In the near term, IT leaders can focus on a few pragmatic moves:

First, clean up the foundations. Agents are only as good as the data and workflows they sit on. This is the time to improve your APIs, standardize event streams, rationalize identity, and get your knowledge base into a structure that AI can reliably consume.

Second, experiment where the stakes are reasonable. Customer support, IT helpdesk, HR queries, and internal operations are good sandbox areas. This is where Benioff’s customer agents, Microsoft’s copilots, and many enterprise pilots started. You get fast feedback and measurable impact with relatively contained risk.

Third, design for supervision and transparency. Every CEO who is serious about this topic emphasizes human oversight. Whether it is Benioff comparing AI agents to self-driving cars that still need a human in the loop, or IBM talking about keeping “the decisions that matter” with people, the pattern is clear. IT should build logs, dashboards, and approval flows into any agentic deployment from day one.

Fourth, partner with the business on operating model changes. Agentic AI will fail if it is treated as “just another tool”. It touches roles, incentives, KPIs, and risk appetite. The CEOs quoted here are not just shipping features; they are re-framing how their companies work. CIOs should be at that same table, not just in the engine room.

Finally, keep an eye on the cost and hype. Gartner’s warning about a large share of agentic projects being scrapped by 2027 is a real risk signal. Many early efforts will be experiments that do not make it to scale. IT leaders will need to be ruthless about measuring value and shutting down projects that are more slideware than substance.

In Closing

When top CEOs talk about agentic AI, they tend to use big language: platform shifts, digital workers, new operating models. Underneath that language is a very concrete implication for IT.

If agents become the primary way employees and customers interact with systems, then IT is no longer just building and maintaining applications. It is building an environment where fleets of autonomous or semi-autonomous digital teammates can safely, reliably, and productively work.

That means architecture, integration, data quality, security, and governance will matter more than ever. It also means that the IT function is moving closer to the center of business strategy, because whoever controls the agents, controls the way work actually gets done.

The IT team’s job is to decide what those actions should be, how far they go, and how to make sure humans remain firmly in charge of the future they are building.