The Hiring Question Changed — What Leaders Now Have to Screen For

The headcount meeting used to be a familiar argument about volume, cost, and timing. A manager arrived with a queue that no longer fit the design of the role, a spreadsheet that proved the team was busy, and a careful sentence about how one additional analyst would stabilize service levels while process improvements continued. The room asked whether the volume was permanent, whether the role was scoped correctly, whether the work could be simplified. Everyone knew that script.

Now someone asks a different question, in a practical tone. Why can’t AI do this? The question is not hostile. It sounds accountable for cost. It sounds current. And it lands harder than the manager expected, because much of what the spreadsheet documents as work — messages drafted, cases summarized, reports prepared, tickets classified — now looks, on paper, suitable for assistance. That is the first rupture. The manager came prepared to justify additional staffing, and instead has to explain the nature of the work. That shift is not a communication problem. It is the moment the hiring conversation changed for good, and it is the moment executive screening has to change with it.

Why does “Can AI do this?” mislead leaders about the role?

“Can AI do this?” is a substitution question. It asks whether a task can be performed by a tool. It has a productive answer for the visible parts of most roles. AI can draft the response. It can prepare the summary. It can rank the queue. It can classify the ticket. Companies that adopt these capabilities well see genuine relief.

The problem is not the answer. It is what the question leaves out. A senior role is rarely a clean list of tasks once it is lived. It is a bundle of formal work and tacit work. The formal work lives in the job description: review cases, prepare analysis, answer customers, draft recommendations, produce reports. The tacit work lives underneath: notice when the facts do not fit the category, remember which promise was made last month, sense when an exception is not an exception but a signal, know when speed will create a larger cost later. The tacit work often appears as friction. Someone slows down a case. Someone reads the raw notes instead of the summary. Someone escalates a decision the queue treated as complete.

That friction is easy to miss and easy to automate away. It is also where consequence lives. A leader who asks only “Can AI do this?” answers whether the visible activity can move. A leader who asks “What does this work require?” answers whether the judgment inside the work can move. The two questions are not variations of the same thing. One protects throughput. The other protects the company.

What is the difference between relief and redesign?

Relief is what leaders feel first. The queue drains. The dashboard improves. The manager stops asking for headcount. The team looks modern. Relief is real. It is also silent about ownership. It shows that effort has moved. It does not show where judgment now sits.

Redesign begins earlier, before the tool goes live. The work is separated in daylight. Activity that should never have required a person is named as activity — the weekly exception report someone assembled by hand, the policy language copied from one field to another because the process was built before the work changed. Analysis is named as analysis; the software may prepare it, but a defined subset of cases still requires the raw material to be seen. Interpretation is named as interpretation. Judgment is named as judgment. The commitments the role creates for the company — the promise, the appeal, the recommendation a person will later be asked to defend — are named separately, with a named owner. The role description is rewritten before the tool goes live, not after, so everyone knows who owns the recommendation when a disputed case surfaces months later.

Relief without redesign is the pattern that carries the largest hidden cost. The dashboard improves. The team is less buried. Something about the work feels less owned. The first pass arrives already written. The messy case is compressed into a summary before a supervisor sees it. Approval no longer distinguishes between real judgment and acceptance of the tool’s frame. A company can become leaner and less capable at the same time. Search work that ignores this distinction places leaders who make the queue faster and cannot say what the role now protects.

Why does AI enter the organization without a role description?

A human role enters a company through a known architecture. The architecture may be flawed, but it exists. Human resources can define the job. Finance can price it. A manager can assign it. The person can be trained, coached, corrected, promoted, or removed. The role can be placed on the chart. The person can be told, “This is yours to own.”

AI enters differently. It enters through design. It enters through a tool a team adopts before the policy catches up. Through a summary feature embedded in software already approved for another purpose. Through a ranking model added to a queue. Through an agent allowed to complete a workflow. Through a default. No new box appears on the org chart. No title changes. No manager gets a new direct report. Yet the work begins to move through a nonhuman actor before a human being sees it, frames it, questions it, or approves it. The formal map stays intact. The sequence underneath has already changed.

This is why the hiring question matters more than it looks. It is one of the first places leaders can see whether they are redesigning work deliberately, or allowing work to reorganize itself around the nearest tool. The seat the manager is asking for may be the wrong instrument for the situation. But the seat is also often the last place in the organization where anyone is still forced to state what the work protects. Remove the seat without answering that question, and the question does not disappear. It becomes invisible, distributed across a workflow whose ownership no one has been asked to name.

What should executive screening now test?

Traditional executive screening evaluated whether a candidate could do the work of the role — depth in the function, track record with the stakes, ability to build a team, ability to deliver against a plan. Those criteria are still necessary. They are no longer sufficient. When AI absorbs the executable parts of a senior role, the human failure mode shifts. It stops being couldn’t do the analysis and starts being accepted the analysis without pressure-testing it. Screening has to reach into the parts of the role AI cannot do.

Framing quality. The strongest AI-augmented executives can state the question the tool is answering, and can rephrase it when the situation demands. In an interview, this is the candidate who can tell you what a good version of a specific decision looks like before the model produces a recommendation. Weak candidates only critique what the tool proposed. Strong candidates frame what the tool should have been asked.

Escalation instinct. The senior person who never overrides a well-scored recommendation is not calibrated; they are absent. The one who overrides everything to prove they are still needed is a different problem. Screening has to find the middle: candidates who can describe cases where they slowed the line despite a clean summary, name why, and stand by the outcome. Ask for the case, ask for the second one, ask for the third one.

Comfort with disagreement. Much of the tacit work in a senior role now surfaces as a moment where the machine’s answer and the human’s read diverge. Executives who avoid that moment — by re-reading the summary until it feels right, by pushing the case back into the queue, by delegating the disagreement — will not run the exception well when it matters. Screening should stage that divergence in the interview, not hypothetically, and watch what the candidate does with it.

Ownership of the sentence. Every consequential piece of work needs a human sentence: I own this choice. That sentence cannot be implied by a button. It cannot be assumed because a person was copied on the record. It cannot be manufactured after the fact by pointing to an approval log. It has to be designed into the work before the work moves. Executive candidates who cannot describe which sentences in the role are theirs to own — and which the tool should never be allowed to author — are not ready for a seat where AI is already speaking on the company’s behalf.

What does this mean for the search committee brief?

The brief is where the shift becomes concrete. A pre-AI brief listed responsibilities, scope, comp, and reporting structure. An AI-augmented brief has to add two paragraphs the old one did not need.

The first names what the role protects. Not the whole role — the specific commitments the seat is accountable for that the tool must never be allowed to close on its own. A commercial recommendation to a vulnerable customer. A risk exception above a defined threshold. A public statement that binds the company. A hiring decision. A promise that will be cited in a later dispute. The brief should be able to name these. If the operating team cannot, that is the finding — and the search work should surface it before candidates are assessed against a role that has not yet been designed.

The second names the shape of the decision loop the incoming executive will inherit. Where does the AI produce output first, and the human review after? Where does the human frame first, and the AI produce inside that frame? Where is the executive expected to stop the line, and on what grounds? A candidate can be brilliant on execution and wrong for a loop that lives further downstream than they have ever worked. A senior operator whose entire career predates AI’s arrival in the decision path may be the right person, or the wrong one, but the answer is unreadable without stating what the loop now looks like.

Search that does not surface these questions places seats that will feel modern and mis-owned within a year. Search that does surface them protects the sentence the company needs the person to be able to say: I own this choice.

Frequently asked questions

Is AI going to reduce the number of senior hires? Sometimes. More often it reshapes them. Roles that were volume-driven often need fewer seats but stronger decision rights. Roles that carry consequence often need a more senior operator because the work has become more interpretive, not more voluminous. The right answer follows from the work; it cannot be inferred from a productivity trend.

How do we test judgment in an interview if the candidate has not worked in an AI-augmented role yet? Stage the situation. Give the candidate a case with a clean AI-generated summary and a set of raw notes underneath. Ask what they would do next. Ask why. Ask what would have to be true for them to override. Judgment is legible in how a candidate handles the divergence between summary and source, not in what they say about AI in the abstract.

Should we screen for AI fluency or AI skepticism? Neither, on its own. Screen for the ability to say what the tool should be asked, what it should be allowed to close, and what it should never be allowed to author. Fluency without discipline overtrusts the machine. Skepticism without fluency stalls the work. The senior operator you need can hold both.

Jose J. Ruiz is CEO and Managing Partner of Alder Koten and Chairman of Anker Bioss. He develops these ideas at book length in AI in the Org Chart: A Leadership Guide to Implementing AI Without Losing Human Judgment, Accountability, and Trust, published by Elavant Press and available on Amazon.

If you are redesigning a senior seat around an AI-augmented workflow — or briefing a search where the role’s decision loop has changed shape — start a conversation with Alder Koten. The right operator, framed against the work as it now runs, is what protects the sentence the company needs the person to be able to say.