Something shifted in the room during Tom Zehren's keynote at the Bellagio on Monday morning. The audience was full of CIOs who had spent the past two years managing expectations downward, defending AI budgets they could not yet justify with outcomes, and absorbing board pressure from both directions. Zehren did not validate the frustration. He made the case that this moment, right now, is the one IT leaders have been waiting for.

The opening keynote of Info-Tech Research Group's LIVE 2026 conference was titled "Agentic IT: From Hype to Value." The session covered five widely repeated claims about AI, cleared away the ones that do not hold up under research scrutiny, and then gave the room a four-pillar operating framework for getting to real enterprise value. The stat everyone has been quoting about AI pilot failure rates got one sentence of context and a quick dismissal. The rest of the hour was about what to do.

The provocative setup came from a claim Zehren addressed head-on: that IT departments are on their way to becoming little more than agent-management functions, a kind of human resources operation for software that runs itself. He disagreed with it, clearly and on the record.

His counter is grounded in where the real constraint on AI value lives. Model quality is not the problem. Infrastructure is not the problem. The binding constraint, in Info-Tech's research across roughly 550 IT executives, is organizational readiness: the capacity to translate AI capability into business logic, workflow redesign, and stakeholder alignment. That work requires people who understand technology deeply and can bridge it to the business. It is, in other words, exactly what strong IT organizations do.

As AI handles more of the execution layer in software development, the value of enterprise architecture, business relationship management, and stakeholder navigation increases. Zehren made the point that enterprise architecture in particular has been declared dead so many times it has become background noise. The current moment is an argument for its expansion, not its retirement.

Eighty-four percent of software development specialists are already using AI across the full software development lifecycle, according to Info-Tech's survey data. What Zehren described from conversations with engineering teams at large technology companies goes further.

Six months ago, those developers were spending significant time reviewing and fixing bugs in AI-generated code. Today, at the organizations that have scaled these workflows thoughtfully, that remediation burden has dropped to the point where AI-generated code ships to production with human oversight but without the rework cycle. That is a phase transition, not an incremental improvement.

Zehren's framing for what this means inside IT organizations was clear: the right move is not headcount reduction but redeployment. The execution work that consumed the majority of IT resources in a traditional delivery model is compressing. The strategic work, defining what gets built, why, and for whom, is expanding to fill that space. For IT leaders who have spent careers wanting a seat at the strategy table, this is the mechanism that creates it.

The finding that landed with the most practical weight: organizations with a standalone, board-governed AI strategy are three times more likely to reach measurable business value than those that have folded AI into the broader IT strategy document. Fifty percent of surveyed organizations have that dedicated strategy in place. Thirty-five percent do not.

This is where the energizing argument meets the operational one. The CIO who wants to lead on AI needs the governance structure that gives technology decisions organizational authority. Without board-level commitment to a standalone AI strategy, every deployment decision gets made in a vacuum, every budget request is a negotiation from scratch, and the function that should be driving the work is instead defending it.

Getting there is a structural ask, not a technical one. It requires IT leadership to make the case to the board that AI deserves dedicated governance, real accountability, and a funding model built around it. That conversation is harder than any infrastructure deployment. It is also the one that determines whether everything else works.

The organizing idea of the keynote was a reframe of the CIO's operating posture. For most of the past decade, the CIO role has been institutionally defined by scrutiny, security reviews, compliance holds, vendor risk assessments, and technical debt management. Important work, all of it. But the cumulative effect has been an identity built around constraint.

Zehren's prescription was the "CIO yes, and." It is not a blanket yes. The "and" matters as much as the yes. The "and" is the set of conditions: a governance model, a realistic sandbox, a funding structure that kills low-value projects quickly, a data foundation that makes AI useful rather than decorative. The yes is the posture that says IT will be the organizational partner that makes things possible, not the function that makes things complicated.

Forty-two percent of organizations have already deployed AI across multiple departments and are reporting measurable impact. Those results did not come from organizations where IT was sitting at the back of the room waiting to be consulted. They came from organizations where IT decided to lead.

One of the most practical points of the session addressed why so many organizations have sandboxes that nobody uses. Usability is almost always the cause, not security. A sandbox that does not connect to synthetic versions of real production data sources gives innovators inside the business nothing to build against. IT enthusiasts in every corner of the organization, the people closest to the use cases with the highest potential value, need environments that mirror reality closely enough to produce something testable.

Twelve well-constructed sandboxes across an organization, built to production standards and stocked with synthetic data, are worth more than a hundred sandboxes that exist only on paper. Info-Tech announced a new demo lab at LIVE 2026 specifically to address this, giving attendees hands-on time with its AI agentic prototyping service to build their own productivity agents on site.