Executives now need to prove what they built with AI

AI is changing what senior leaders have to prove. Twenty years of operating history once helped an executive earn confidence about the next 12 months because accumulated experience signaled pattern recognition and judgment under uncertainty. AI weakens that shortcut. Young technical talent faces a demanding standard of its own: leaders still have to build with AI, judge what it produces and persuade an organization to change.

That standard creates pressure at both ends of the leadership pipeline. Experienced transformation executives may know how to move processes, technology and people while having little direct experience deploying AI, while younger builders may know the technology but have little experience winning trust across an organization. JC Christian of executive search firm Christian & Timbers says the same core considerations that mattered 20, 10 or 50 years ago remain relevant. Because Christian works in executive search, his assessment comes from a business directly involved in how companies evaluate and fill executive roles.

The hiring question is how much credibility either experience or AI fluency deserves on its own. Experience has to connect to current execution, while technical fluency has to connect to judgment and organizational results. That combination changes what companies should examine in hiring, promotion, succession planning and leadership development.

AI is changing what counts as executive proof

Experience earned its value partly because judgment under uncertainty is difficult to teach and difficult to fake over a long career. A leader who had worked through enough business cycles could bring pattern recognition and institutional knowledge that others in the room lacked. Accumulated history therefore gave employers a rational basis for extending the benefit of the doubt. Scarcity strengthened that credibility because much of the relevant knowledge and interpretation lived with a relatively small number of people.

AI changes the evidentiary value of that history by giving leaders more ways to work directly with specialized information and systems. Past experience consequently has less power to substitute for evidence of what an executive can do now. A leader can still draw on years of transformation work, while employers can increasingly ask what that leader has actually built, deployed and changed with AI. Experimentation and deployment give companies more current evidence to examine.

The same demand for current evidence applies to technical fluency. Someone who can produce impressive AI work may demonstrate initiative and technical agency while giving an employer little evidence that they can persuade a finance leader, operations leader or business unit head to adopt a different process. Executive work includes that organizational step because a technically successful system creates limited business value when people refuse to use it. Technical ability therefore establishes one part of executive credibility.

That partial signal leads to a more demanding standard for both groups. Companies need leaders close enough to AI to understand and shape what is being built, with enough judgment to decide whether it is useful and enough organizational influence to change how work gets done. Christian’s executive-search observations provide one way to test the first part. He starts by asking a candidate to explain their own work.

The first test is whether a leader has actually built with AI

Christian begins with the problem an executive tried to solve. He asks candidates to describe a problem they solved using AI and treats the first 30 seconds as a barometer of whether they really understand the work. “Most people can’t even accurately describe the problem itself, let alone get anywhere near a solution,” he says. A candidate who cannot define the problem precisely gives Christian reason to question whatever follows.

Once the problem is clear, Christian moves to what the executive built and how deeply the executive understands it. Hands-on AI means being close enough to the work to connect the initial business problem with the system created to address it. Transformation leaders can meet that standard through close involvement with the implementation. They need enough involvement to explain what was built in concrete terms.

That explanation leads to deployment because a working AI system becomes meaningful to a business when it enters the way people work. Christian looks for models placed into workflows and agents applied to business processes; an agent is an AI system used to perform or coordinate tasks toward an objective. He then asks what changed after deployment. The sequence is concrete: define the problem, explain the build, show its use in a workflow and identify the resulting change.

A mainstream AI product by itself provides much weaker evidence. “I talk to people who are like, we just started using Copilot and now we’re AI native,” Christian says. His response is explicit: “And it’s like, no, unfortunately you’re not. You’re at the very beginning of the AI-enabled adoption curve.” For hiring, the distinction matters because tool access provides little evidence that a leader has redesigned work around that tool or produced a meaningful outcome.

Christian applies the same focus on outcomes when assessing company-level evidence. Public companies are eager to discuss returns from AI, and he assumes some public claims may be aggressive, while a complete absence of results concerns him more. “If we’re not bragging at all, that’s typically a red flag,” he says. For an executive whose remit specifically covers deployment, he makes the implication personal: “If you’re the AI lead of deployment at a company and they don’t have any AI ROI, it’s pretty bleak for being able to claim your role there was a success.”

That outcome test moves executive assessment toward demonstrable work. A history of leading three transformations can remain relevant because it provides evidence about difficult change, but employers now have a more specific follow-up: what did the candidate actually change in this transformation? Production evidence answers only part of the question because a technically working system can still fail where organizational value is created. The next test is adoption.

Building becomes valuable when people adopt what was built

Christian places the harder part of AI transformation beyond the initial technical work. “Adoption is actually the biggest problem,” he says. “The technology is an easier problem. It’s more solvable.” Many transformation roles center on changing how an organization works, and curious executives who remain close to the technology can learn enough to work effectively with the tools. Getting an organization to alter established behavior requires another set of capabilities.

Those capabilities matter because adoption determines whether technical work becomes business value. An AI system can solve its intended technical problem and still create insufficient value if stakeholders keep their existing workflows. Resistance can make the outcome worse than wasted development effort because a disrupted process can impose costs of its own. As Christian puts it, without buy-in, “adoption’s going to be a mess. You’re going to get zero value, maybe even negative value.”

That adoption problem explains a recurring decision Christian hears from clients. “A lot of leaders say, ‘Hey, I have a 20-something-year-old whiz kid who lacks gravitas and transformation experience completely, but has built all this cool stuff and knows AI best.’ And I’m considering, do I promote them and make them my VP of AI adoption, or do I look externally?” The candidate has direct evidence of building. The unresolved question is whether that evidence extends to leading change across functions.

Cross-functional change raises a trust problem for a builder moving directly into leadership. AI adoption often asks established leaders to change their team’s processes, accept unfamiliar systems and sometimes reconsider how responsibilities are divided. Technical expertise alone does not confer the organizational standing needed to win those decisions. Christian has therefore seen strong younger builders placed in chief-of-staff or second-in-command positions alongside more seasoned executives, pairing direct AI capability with experience navigating an organization.

That pairing also exposes a weakness in how companies sometimes describe the experienced side of the equation. Terms such as “gravitas” and “executive presence” are subjective, so assessments based largely on feel can reproduce incumbent preferences about how a senior leader should look and sound. Younger AI builders can consequently be screened out through criteria with little connection to whether they can do the work. If organizational influence is the capability that matters, companies can test it more directly.

The direct test is whether a person can persuade skeptical leaders to change how their teams work. A candidate who can bring a resistant finance leader, operations leader or business unit head through an adoption decision has demonstrated something observable. That evidence captures influence rather than polish and gives employers a way to evaluate change leadership without treating familiarity with established executive norms as the capability itself. Experienced candidates can be assessed against the same evidence.

Experience remains useful under that test because someone who has repeatedly moved organizations through difficult transformations may have learned how to establish trust, handle competing incentives and work across process, people and technology. That history still leaves current AI capabilities to be demonstrated: identifying a useful AI problem and staying close enough to deployment to understand what is changing. Younger builders face the inverse gap when technical production outruns cross-functional leadership experience. The two profiles therefore arrive at the same role with different evidence still to provide.

Those gaps suggest a sequence for evaluating credibility. First, a leader identifies the business problem accurately enough to build against it. The leader then stays close enough to the resulting AI work to judge its quality and consequences, then persuades others to adopt the resulting change. Each stage produces evidence needed for the next, reducing reliance on tenure or broad claims of fluency.

Judgment becomes crucial in that sequence because adoption can make a weak system consequential. A leader has to decide whether an AI-generated recommendation deserves to shape a workflow, whether an output is sufficiently reliable and when apparent technical success hides a business problem. Wider access to information can support those decisions, but faster AI production can also make careful review easier to skip. That tension changes the value of knowledge itself.

As scarce knowledge gets cheaper, judgment becomes more valuable and easier to neglect

Brandon Sammut, chief people and AI transformation officer at Zapier, has seen what knowledge scarcity can do to an organization. His remit puts him directly inside the organizational changes created by AI transformation. At a previous company, a couple of engineers were the only people who understood one area of the product architecture. “As a result, they were effectively untouchable,” Sammut says.

That dependence imposed a cost on the engineers as well as the organization. “That’s not super healthy or high functioning for anyone in that equation,” Sammut says. When essential knowledge is concentrated in a few people, those people gain protection but can also become trapped in the work only they know how to perform. Sammut sees AI as accelerating the democratization of knowledge that began with the internet, making more expertise available beyond the people who previously held it.

Greater access changes which capabilities Sammut expects to carry more value. He identifies judgment, taste, trustworthiness, reliability, accountability and the ability to influence without authority as increasingly valuable. These capabilities determine what someone does with accessible knowledge and whether others can rely on the result. Christian’s hiring tests address a separate question, what evidence a candidate can show, so the two perspectives operate at different levels rather than requiring the same explanation.

Greater access can also create more opportunities to develop judgment because ideas can be tested much faster. As an illustration rather than a measured finding, consider a problem that someone might previously have encountered a dozen times over several years: AI can allow that person to pressure-test it dozens of ways in an afternoon. More repetitions can expose a leader to alternatives, objections and consequences at a much higher rate. Faster exposure can develop judgment when the person genuinely evaluates those outputs.

That opportunity comes with a review problem. Tony Castellanos, executive vice president of people at Nextdoor, approaches the issue from a people-leadership role in which AI-generated work affects how individuals and organizations operate. “I’ve tried to read most of the docs that have come my way, but honestly, they’re just flowing in so fast,” he says. Generation capacity can therefore rise faster than review capacity, allowing apparent productivity to increase while each individual claim or decision receives less scrutiny.

Repeated success creates another pressure because trust changes review behavior. Castellanos describes the process this way: “You have a new hire, you don’t automatically trust everything that they do, but over time they deliver, they get things right, and you more and more trust them, just like you’re going to more and more trust an LLM,” using LLM to mean the large language model producing the output. As confidence accumulates, he says, “As they’re sending things to you to proofread, to audit, to check, you’re probably doing less and less and less of that over time.” Reliability can therefore weaken the habit of checking the system that appears reliable.

Castellanos illustrates that progression with a hypothetical six months of working with a model, validating its output and repeatedly exercising judgment. Sustained success would reasonably increase confidence, but that confidence can reduce how often the human performs the checking through which judgment was exercised. AI can consequently increase the opportunities for judgment while encouraging users to scrutinize a smaller share of its output. High production volume is weak executive evidence when the decision process behind that production becomes less demanding.

To restore scrutiny, Castellanos proposes productive disagreement without requiring a human to check every AI output forever. He calls for a “third skeptic,” potentially another AI system, that challenges the relationship between the user and the model the user has learned to trust. When two systems disagree, the human has a reason to examine assumptions and evidence again. Disagreement matters because it forces an actual decision instead of allowing accumulated trust to settle the question automatically.

That disagreement returns the decision to the leader: “What’s the right call? What’s the right thing for me? What’s the right thing for my organization?” The independent challenger can do its job simply by exposing a decision that had become automatic and requiring the human to own the result. The human still has to decide what to accept. Accountability therefore remains with the executive even when AI participates on multiple sides of the analysis.

That accountability complicates the demand for hands-on AI evidence. Companies need executives who can show that they have built and deployed real systems, while AI makes sophisticated-looking output increasingly easy to produce. The same tools can become sufficiently trusted that users gradually surrender opportunities to question their results. Executive assessment therefore has to distinguish AI production from the judgment governing what reaches a workflow, what gets challenged and what ultimately shapes a decision.

Christian expects AI fluency itself to become less distinctive as the technology spreads. He compares its likely trajectory with financial literacy: an ability that may look unusual among executive candidates now can become an ordinary baseline expectation. Once that happens, saying that an executive works comfortably with AI will communicate relatively little. More useful evidence will come from how that executive decides when an output is right, when it is wrong, which consequences matter and what the organization should do next.

Hiring systems still measure yesterday’s proxies

If judgment and adoption carry more evidentiary weight, executive selection systems have to expose them. Succession plans, promotion criteria, leadership programs, P&L assignments and long functional apprenticeships were largely designed to reward accumulated experience and create evidence that someone has operated at increasing levels of responsibility. Those mechanisms can still develop useful leaders, but direct contact with AI deployment has to be examined separately. Adding AI fluency to an existing checklist leaves the underlying assessment problem intact.

Interview systems show the same mismatch. Christian says many executive interview processes remain centered on familiar behavioral questions, formulas for extracting examples and the eventual gut checks of interviewers, even as companies ask what new questions might identify the capabilities they now want. Gut judgment becomes especially problematic when it turns into an assessment of “executive presence,” because subjective impressions can obscure the concrete capability underneath. Evidence of problem definition, deployment and organizational influence gives interviewers more specific material to evaluate.

For experienced candidates, that evidence means deeper questions about what they have deployed. General familiarity with models or productivity tools provides limited information about whether they can identify a suitable business problem, stay close to implementation, evaluate the result and lead adoption. Many AI transformation jobs center on organizational change, with engineering as a separate discipline. Executives still need sufficient proximity to the work to understand what happened and take responsibility for the decisions around it.

For younger builders, the corresponding evidence concerns organizational leadership. Technical strength can support promotion into an adoption role, while influence still has to be demonstrated through skeptical stakeholders, trust and changed behavior across functions. Conventional “gravitas” gives companies a subjective proxy for those capabilities, so observable behavior is more useful. When direct cross-functional experience remains limited, pairing a strong builder with a seasoned executive in a chief-of-staff or second-in-command role is one way to develop that evidence rather than treating the pairing as a universal organizational design.

Leadership development follows the same evidentiary shift because AI broadens the problem spaces executives can enter. Castellanos expects expertise to remain important while people use AI to reach knowledge that previously stayed within a function, and Sammut’s account shows how concentrated information can shape status and organizational dependence. Wider access raises the value of accountability, reliability, judgment and influence because someone still has to decide what to do with that information. Those decisions then have to survive contact with the organization.

That requirement changes what familiar leadership signals can establish. Tenure can show that someone has survived hard decisions, AI fluency can show that someone understands important new tools, and organizational influence can show that other people are willing to follow. Hiring and promotion systems can turn those signals into observable tests by asking leaders to define a problem, explain what they built, defend the judgments they made and show how they changed behavior. The resulting evidence comes from the work the executive can explain and own.

Main highlights

• Test executive proof through current AI work: Hiring and succession teams can ask candidates to define the business problem, explain what they built, show how it entered a workflow and identify the resulting change. This makes current execution a stronger complement to career history.
• Treat adoption as evidence of leadership: AI builders prove executive readiness when they can persuade skeptical stakeholders to change established workflows. Companies evaluating technical talent for senior roles can test cross-functional influence directly rather than rely on subjective measures such as executive presence.
• Protect judgment as AI output accelerates: Faster access to expertise raises the value of judgment, accountability and reliability while increasing the volume of work requiring review. Organizations can introduce independent AI challengers or other review mechanisms to surface disagreements and keep consequential decisions under human scrutiny.
• Rebuild hiring around observable capabilities: Interview, promotion and leadership-development systems can test problem definition, hands-on AI deployment, decision quality and organizational influence. These measures give companies stronger evidence of readiness as AI fluency becomes a baseline executive skill.