A profile is not a person.

i. The category mistake

The category mistake

Gilbert Ryle used the term category mistake to describe a specific kind of conceptual error: treating something as if it belonged to a category to which it does not belong.

His classic example is the visitor who is shown the colleges, libraries, laboratories, lecture halls, and administrative buildings of a university, and then asks: "But where is the university?" The visitor has not missed a building. He has misunderstood the kind of thing a university is. The university is not one more building alongside the others. It is the institutional organization of those buildings, practices, roles, and relationships.

The same mistake appears in AI-led hiring.

A candidate profile is a representation. It may include a CV, a LinkedIn page, a database record, an AI-generated summary, a keyword match, a skills graph, a ranking, or a fit score. These artifacts may be useful, accurate, and well produced. They help make the market visible. They help structure search. They help compare signals.

But none of them is the candidate.

The candidate is the person whose career is being represented. That person has motivations, values, constraints, decision patterns, fears, ambitions, commitments, and a particular moment in which they may or may not be open to change. They have a history of choices made under pressure. They have a way of reading power, conflict, ambiguity, silence, risk, and responsibility. They have reasons for moving, and reasons for not moving.

Adding more representation does not cross the category boundary. A longer CV does not become a person. A richer LinkedIn profile does not become a candidate. A fluent AI summary does not become an assessment. A fit score does not become judgment.

The operational consequence is direct: if a senior search treats the profile as the candidate, the search is built on the wrong object. The shortlist becomes a function of profile similarity rather than candidate-mandate fit. The interview becomes pattern verification rather than interpretation. The final decision becomes the selection of the most plausible representation rather than the right person for this organization, in this role, at this moment.

ii. Competence is not comprehension

Competence is not comprehension

This distinction matters because AI is genuinely competent at many tasks inside executive search.

Daniel Dennett used the expression competence without comprehension to describe systems that can perform sophisticated tasks without understanding the reasons or meanings that human observers associate with those tasks. A thermostat regulates temperature without understanding comfort. A search engine retrieves documents without understanding their significance. A spell-checker corrects language without understanding what is being said.

This is the right way to think about AI in senior search.

AI can process CVs at scale. It can classify career trajectories, detect adjacent industries, summarize profiles, extract structured information, compare candidate histories against role requirements, rank signals, identify hidden title equivalences, and generate plausible interview questions. None of this should be denied. AI can do real work in executive search. At KiTalent, we use it for mapping, deduplication, document processing, signal organization, and market intelligence.

But competence in profile processing does not entail comprehension of the candidate.

A system that summarizes a career does not understand the motives behind that career. A model that infers technical alignment does not grasp whether the candidate can operate inside the politics, pace, values, and informal power structure of the client's organization. A tool that generates a fit narrative does not own the judgment that the narrative invites.

The problem is not that AI is useless. The problem is that fluency at the representational layer is mistaken for understanding at the candidate layer.

AI can possess mapping competence without candidate comprehension.

iii. Hiring language is not self-interpreting

Hiring language is not self-interpreting

The language of senior hiring is dense, familiar, and deceptively simple.

CVs and job descriptions are full of phrases such as "strategic leadership," "ownership mindset," "executive presence," "stakeholder maturity," "transformation experience," "commercial acumen," and "culture fit." These phrases are not meaningless. But they are not self-interpreting either.

Their meaning depends on use.

"Ownership" does not mean the same thing in a private-equity-backed turnaround, a regulated bank, a founder-led technology company, a family-owned industrial group, or a post-merger integration. "Transformation experience" may mean having designed the transformation, inherited it, executed someone else's plan, survived the politics around it, or merely occupied the right title while it happened. "Stakeholder maturity" may mean directness in one organization and diplomatic restraint in another.

An AI system can process these phrases as linguistic forms. It can detect their presence, classify their density, compare their distribution, and rank candidates according to their proximity to a brief. What it cannot do is participate in the practice that gives those phrases their assessment meaning.

It cannot ask the follow-up question that reveals whether a candidate's transformation experience was real or cosmetic. It cannot reconstruct the language-game of the candidate's previous organization and compare it to the language-game of the client's mandate. It cannot test whether "ownership" in the candidate's narrative corresponds to the kind of ownership the role will actually require.

This is why hiring language must be interpreted, not merely processed.

iv. Interviews are not data extraction

Interviews are not data extraction

If a senior interview were simply an exercise in collecting more information, AI would be a natural substitute. It could ask standardized questions, transcribe answers, summarize responses, and score keywords faster than a human consultant.

But a serious senior interview is not a data extraction session.

It is a reason-giving practice.

The candidate is not merely asked to provide facts. They are asked to explain decisions, defend trade-offs, revise claims under pressure, account for contradictions, clarify motivations, and take responsibility for the picture of themselves that emerges through the conversation.

The consultant is not merely collecting data. The consultant is testing the candidate's reasons against the mandate: why this move, why now, why this organization, why this kind of risk, why this operating environment, why this leadership problem, why this level of ambiguity, why this cultural reality.

This matters because senior hiring is not just about what a candidate has done. It is about how they understand what they have done, why they made the choices they made, what they learned, what they protect, what they avoid, what they are ready to carry, and what they are likely to do when the role becomes difficult.

No AI system participates in that responsibility. An AI can generate questions, summarize answers, and organize interview evidence. But it does not own the judgment. It cannot be accountable for whether a candidate's reasons genuinely fit the client context or whether the hire will hold under pressure.

v. Culture fit is not a soft skill

Culture fit is not a soft skill

This is also why culture fit should not be treated as a soft skill.

Soft skills are capabilities. Communication, adaptability, stakeholder management, influence, conflict handling, emotional regulation, and structured problem-solving can be observed, developed, benchmarked, and compared across contexts. They are part of what a person can do.

Culture fit is different. It is not an extra soft skill added to the competency list. It is a relational judgment between a person's identity and an organization's operating reality.

It concerns values, motivation, decision patterns, leadership style, tolerance for ambiguity, political judgment, pace, conflict, authority, and the kind of environment in which the candidate's way of working actually makes sense.

Treating culture fit as a soft skill is a category mistake. It treats a relational identity-context judgment as if it were a portable capability.

This does not mean culture fit should be vague or subjective. Quite the opposite. When culture fit is handled carelessly, it can become a cover for bias, familiarity, or homophily. The answer is not to automate it into an opaque score. The answer is to define it explicitly, assess it structurally, and interpret it in relation to a real mandate.

AI can help map capability signals. It cannot perform identity-fit judgment.

vi. The operational boundary

The operational boundary

The practical conclusion is a division of labor.

AI belongs where the task is representational, comparative, documentary, or classificatory. It is useful for market mapping, candidate universe construction, company adjacency analysis, title equivalence, CV parsing, document screening, deduplication, signal organization, reporting, and preparation.

Consultant judgment belongs where the task is interpretive, relational, motivational, cultural, or accountable. It is required for candidate engagement, mandate positioning, motivation testing, identity assessment, leadership-style interpretation, stakeholder calibration, culture-fit judgment, shortlist accountability, and final recommendation.

This boundary is not a slogan. It is methodological.

AI expands and structures the field of visibility. Consultant judgment interprets what becomes visible.

AI helps construct the candidate universe before outreach. The consultant decides who should be approached, how the mandate should be positioned, whether the candidate is genuinely motivated, whether the career narrative holds under questioning, and whether the person belongs on the shortlist.

Where AI can do work better and faster than a consultant, it should be used. Where judgment matters, judgment must remain with the consultant.

vii. Why blind CVs are weak evidence

Why blind CVs are weak evidence

The same argument explains why we do not send blind CVs.

A blind CV is a representation stripped of the very conditions that would allow it to function as meaningful evidence. It removes the candidate's identity, current employer, real availability, compensation expectations, motivation, mobility constraints, and the conversation that produced the document.

As a diversity tool in some early-stage hiring processes, anonymization can be useful. It can reduce certain forms of bias when the task is initial screening and when the role is standardized enough for anonymized comparison to make sense.

But blind CV selling in executive search is different.

In senior search, a blind CV is weak commercial proof. It proves that a document can be produced. It does not prove that the candidate is real, reachable, interested, assessed, available, motivated, or aligned with the mandate. In the age of generative AI, this weakness becomes even more obvious: plausible candidate documents are now cheap to produce, and plausibility is not evidence.

A blind CV is representation without epistemic accountability.

viii. Proof-First Search as situated evidence

Proof-First Search as situated evidence

Proof-First Search reverses that logic.

The model is not a discount, a shortcut, or a disguised contingency process. It is a different way of organizing trust in a high-stakes professional service.

Instead of asking the client to commit the major fee before evidence exists, the firm runs the search and presents a validated shortlist of real, named, interviewed, motivated, available candidates against a real mandate. The evidence is not that a plausible profile exists. The evidence is that actual candidates, in the actual market, under the actual constraints of the mandate, are reachable, interested, assessable, and worth interviewing.

That is why Proof-First de-risks the process. It shifts trust from promise to situated evidence.

The blind CV says: here is a representation.

The Proof-First shortlist says: here is proof that the market has been entered, the candidate has been engaged, the motivation has been tested, and the fit has been interpreted.

AI helps construct the field from which that shortlist is drawn. Consultant judgment produces the shortlist as evidence.

ix. What this argument does not claim

What this argument does not claim

This argument is sometimes misread as a defense of human recruiters against technological progress. It is not.

The point is not to slow technology down. The point is to put it in the right category and ask it to do the work it can do well.

It is also not a defense of unstructured human judgment. The selection research on the limits of intuition is correct. Algorithms and structured models often outperform unaided human judgment in tasks where the variables are stable, the evidence is codifiable, and the prediction target is clear.

The alternative to AI-led assessment is not gut feeling.

It is structured, evidence-led, consultant-owned judgment, supported by AI where AI is epistemically appropriate and constrained where it is not.

The relevant distinction is not human versus machine. The relevant distinction is the kind of evidence required by the task.

If the task is mapping, AI is powerful. If the task is identity-fit judgment, AI is insufficient. If the task is document production, AI can accelerate. If the task is candidate understanding, the profile is not enough.

x. The conclusion

The conclusion

A profile is not a person.

Representation is not understanding.

Competence is not comprehension.

Mapping is not judgment.

Once these distinctions are clear, the division of work follows. AI should help make the talent market visible. Consultants must remain responsible for interpreting the candidate.

That is why AI belongs in mapping, why it should not replace identity assessment, why culture fit is not a soft skill, why blind CVs are weak evidence, and why Proof-First Search creates a stronger form of de-risking in senior hiring.