What 30 Enterprise CTOs Actually Ask About AI and Jobs

We recently spent an afternoon with about thirty enterprise CTOs, presenting our work on labor market trends and the effect AI is having on them. The presentation turned out to be the smaller half of the session. What followed was a long run of questions, and almost none of them were about models, benchmarks, or vendors.

They were about people, and they were unusually blunt. Does hiring get slower now that the market is full of fabricated signals? What happens to the middle manager? If our processes are poor, are we not simply going to build poor agents? And how do we square the argument that AI creates more roles with the very public view that it will destroy them? Thirty CTOs are not a representative sample of anything. But the consistency of the concerns was hard to miss.

What follows is how we answered them, arranged so that a CHRO or workforce planning leader can act on it. One thread runs through all of it. AI changes the distribution of work well before it changes the number of jobs, and the organizations making progress are the ones starting at the level of the task rather than the level of the headcount plan.

Relabeling automation as agentic is not a strategy

Several CTOs described the same internal confusion: work that was automated years ago is now being repackaged as an agentic opportunity, and nobody can tell what is genuinely new. Our position is straightforward. If traditional automation is already working, the next step is not to rename it. It is to do the harder analysis of where agents create incremental impact over what already exists. Unless the most critical tasks in a process are themselves automatable, meaningful FTE savings will remain difficult, however promising the broader agentic opportunity looks on a slide.

The related question was sharper: if we have poorly designed processes, will we not just build poorly designed agents? That is absolutely true. Agent effectiveness depends directly on the quality of the underlying process, which means the deliberate part of the work is choosing where agents belong at all. Making an entire onboarding process agentic raises a question worth answering out loud — is the organization comfortable with a completely non-human onboarding experience? There are adjacent tasks where the answer is easy and the value is real: translating employee manuals into multiple languages, or drafting the first version of a job description. The discipline is selective application, protecting the human touchpoints that carry disproportionate weight.

Why our view on roles differs from the public one

One CTO put it directly: you are telling us there will be more roles, but we do not hear that from leaders like Elon Musk. How do we reconcile the two?

Much of the apparent disagreement is about time horizon and about which population is being counted. In enterprise environments, transformation does not happen overnight; we are looking at a two-to-three year horizon rather than an immediate reset. Within that period, increased efficiency will likely reduce the number of low-performing roles and compress external contractor spend. Those are real reductions. At the same time, demand for high-quality, AI-aligned talent continues to grow, because building agentic operating models requires capable, AI-literate people at scale rather than a small central team. Our own analysis shows AI-enabled software engineers remaining in strong demand. Both statements can hold: fewer of one kind of role, more of another, on a timeline measured in years rather than quarters.

Hiring gets noisier before it gets faster

The first question of the session was whether talent acquisition will now take longer because of the volume of misinformation in the market. It may, and this is exactly where human recruiters become more important rather than less. Validating candidate quality, assessing authenticity, and conducting in-person or onsite evaluation are the activities that separate signal from noise, and they are precisely the activities that get cut first when a TA function is measured on cost per hire. Leaders hiring seriously into AI capability over the next two years should be funding that judgment layer, not thinning it.

Three layers of talent are moving at once

Early-career talent came up repeatedly, usually framed as a problem. I would frame it differently. Today's entrants are fundamentally unlike prior generations: native to cloud, AI-enabled, digital-first environments, and able to contribute faster and more adaptively than early-career talent in previous eras. The constraint is rarely their capability. It is whether the organization has designed work that lets that capability land.

Mid-level managers face the largest adjustment. Their role shifts from primarily coordinating people to increasingly driving execution through AI tools. The managers who become more valuable, not less, are those with strong business context, judgment, and decision-making ability — the parts of the job that were always underweighted relative to coordination.

At the top, something structural is happening. Coca-Cola's James Quincey and Walmart's Doug McMillon have both acknowledged that while they could initiate this shift, the magnitude and duration of the change required a new generation of leadership to fully execute it. That is a striking admission from two leaders of that standing, and it is not really about understanding AI conceptually. It is about the energy, speed, and conviction needed to sustain a multi-year transformation. For current and aspiring senior leaders, the implication is an unusually clear opportunity: those who invest early in AI fluency, rethinking operating models, and driving agentic ways of working are the candidates for the next set of these roles.

Retention now runs through the tech stack

Asked how to retain talent in this environment, the honest answer is that retention increasingly depends on the organization's ability to modernize both its skills and its technology. Strong people want to work with current, relevant tools and to operate in an AI-enabled environment. If the stack is dated, the work is dated, and the capability an employee builds this year is worth less on the market next year. People stay where they can build modern capability and remain close to how work is actually evolving. That makes tech stack investment a retention lever, not only an IT decision.

What the skill shift actually looks like

AI is reshaping skills in ways both subtle and significant. Human-centric skills are rising in importance — creativity, problem solving, and storytelling are becoming critical for interpreting and applying AI-driven output. AI governance is emerging as a foundational capability across every function, with growing attention to risk, ethics, and responsible use. Verification skills, and the judgment that underpins them, are becoming a distinct requirement. And as teams operate at higher scale and speed, complex coordination and the ability to identify new market opportunities matter more.

Roles are also becoming more integrated across functions. DevOps and infrastructure management increasingly intersect with finance, driven by cost optimization and cloud economics. HR roles are evolving to require stronger data and analytics capability so decisions can be made on evidence. The direction is a blend of human judgment, cross-functional awareness, and AI fluency.

The last question of the session was the most practical: how do we actually begin? The starting point is not a broad reskilling program. It is a clear understanding of how work itself is changing at the task level — identifying which tasks are being automated, which are being augmented, and which are newly created, then mapping the skill shifts that follow. From there the work is targeted skill stack upgrades in the roles and functions where AI impact is highest, building AI fluency alongside judgment and storytelling, and enabling cross-functional capability.

Two conditions determine whether this holds. The tech stack has to move with the skills strategy, because employees cannot adopt new skills inside outdated tools and workflows. And the whole exercise has to be treated as continuous rather than as a program with an end date. Skills will keep shifting, and the organizations that build dynamic, continuously updated skills architectures will be the ones still making informed decisions in three years, when the current round of forecasts has already been overtaken.