Status: Ready to publish

Platform: Substack (primary) + LinkedIn companion

Word count: ~1,400 words (Substack) / ~220 words (LinkedIn)

Publish: Week 1

SEO target terms: executive search DACH, AI hiring, transformation leadership, cognitive environment fit

The Machine That Hires Like the Past

How AI is accelerating the most expensive mistake in executive search — and what it cannot fix

Substack version

AI hiring tools are now used in the majority of enterprise recruitment processes across Germany, Switzerland, and Austria. They screen CVs in seconds, score candidates against job descriptions, rank applicants by predicted performance, and flag cultural fit before any human has read a word. They are fast, scalable, defensible, and wrong in exactly the same way that human hiring has always been wrong — only faster and at greater scale.

The mistake is not new. It predates AI by decades. It is structural, not procedural: every major hiring assessment tool ever built has been designed to evaluate the candidate. Nobody evaluates the environment the candidate is walking into. AI did not create this blind spot. It automated it.

To understand why this matters for transformation hiring specifically, you have to understand what AI hiring systems are actually doing. They are trained on historical data — past hires, performance outcomes, tenure patterns, career trajectories. The best-performing systems identify which candidate profiles have historically succeeded in roles like the one being filled. They are, by design, pattern-matching machines that look backward. In stable functions — financial control, legal, procurement — this is a reasonable approach. Past patterns predict future patterns with enough reliability to be useful.

Transformation roles are different in kind, not just degree. A Head of Transformation at a PE-backed Mittelstand business in 2026 does not look like the last Head of Transformation at a PE-backed Mittelstand business in 2019, because the organisations are not the same. The operating environment has changed. The cognitive demands of the role have changed. The speed required, the ambiguity tolerated, the political complexity navigated, the board relationship managed — all of it has shifted. An AI system trained on historical transformation hires will recommend candidates who succeeded in a world that no longer exists. It will filter out the profiles that are best suited to the world that does.

This is not a speculative risk. It is visible in the numbers. Between 40 and 60 percent of senior transformation hires fail within 18 months — a figure that has not improved despite two decades of assessment technology, competency frameworks, and structured interview processes. The market has optimised the tools without questioning the premise. The premise is that if you find the right person, the hire will succeed. The premise is wrong.

The same person in the wrong environment will fail. The same person in the right environment will perform at a level that looks, in retrospect, like genius. This is not a philosophical claim. It is the most consistent finding in decades of industrial-organisational psychology research, and it is the finding that the entire executive search market has chosen to ignore because assessing environments is harder to sell than assessing people. Organisations want to be told they have found the right hire. They do not want to be told that their environment may not be capable of activating what that hire brings.

AI makes this worse in a specific way. When an organisation deploys an AI hiring tool, it introduces a new layer of authority into the process — algorithmic authority, which is harder to challenge than human judgment and easier to hide behind. When a hiring committee decides against a candidate, they can be questioned. When an algorithm scores a candidate below threshold, the decision has the appearance of objectivity. The environment is never discussed because the algorithm never assessed it. The algorithm cannot assess it. Algorithms are trained on data about people. There is no training data for the specific combination of a CDO's cognitive profile meeting a particular German family-owned business's tolerance for disruption in the third quarter of a restructuring programme. That combination has never existed before. It will never exist again in quite the same form. And it is precisely the combination that determines whether the hire succeeds.

This is the problem that APT was built to solve — not as a philosophical exercise, but as a commercial reality that German, Swiss, and Austrian companies are confronting right now. The transformation mandate is not going away. The pressure to digitise operations, restructure post-acquisition, build new business models inside legacy organisations — this is the defining leadership challenge of the DACH business landscape for the next decade. Eight hundred and forty-two thousand German SMEs face a leadership transition in the next five years. The Mittelstand is not equipped for this. Most of them know something is wrong with how they hire. Almost none of them have been offered a structural explanation.

The structural explanation is this: performance is not a property of a person. It is a property of a person in a specific context. The cognitive profile that makes a leader exceptional in a high-ambiguity, fast-moving, politically flat organisation is the same profile that will produce visible failure in a consensus-driven, process-heavy, hierarchically structured one. Not because the leader deteriorated. Because the environment dampened the capability. This is not a question of culture fit in the soft, HR-colloquial sense. It is a question of whether an organisation's operating conditions — its decision-making architecture, its tolerance for challenge, its feedback mechanisms, its genuine versus stated autonomy — can activate the cognitive profile being placed.

AI hiring tools will never answer this question because they are not designed to ask it. They are designed to find candidates. Finding candidates is only half the work. The half that matters — assessing whether the environment can use what the candidate brings — is the half that has been systematically skipped, by human processes and by the AI systems trained on human processes.

APT assesses both sides. The CognitiveEdge Assessment maps candidates across five dimensions that research consistently identifies as the primary cognitive predictors of transformation performance: Absorption & Drive, Ambiguity Tolerance, Decision Calibration, Inductive Reasoning, and Resilience & Recovery. APT Place assesses the hiring organisation across the five corresponding environmental dimensions before any search proceeds. The match between the two — Cognitive Environment Fit — is the variable that drives every placement recommendation.

This is not a better version of what the market already does. It is a categorically different approach — one that treats the environment as a performance variable with the same rigour applied to the candidate. It is the approach that should have been standard in executive search for the past thirty years. It was not because it is harder, slower, and requires an organisation to accept the possibility that the problem is not only who they hire but the conditions they have built.

AI was supposed to make hiring smarter. In many respects it has — it has eliminated the most egregious forms of early-stage screening bias, it has reduced time-to-shortlist, it has made the mechanics of search faster and cheaper. But it has not made senior transformation hiring more likely to succeed because it has not changed the premise. It has accelerated the old mistake. It has made organisations more confident in answers that have always been half-formed.

The leaders who drive transformation — the ones who navigate genuine uncertainty, see patterns before the data confirms them, act before the map exists — have always been filtered out by processes designed for a different kind of performance. AI has made those filters faster. The leaders who survive them are increasingly the ones who are easiest to justify algorithmically, not the ones most likely to succeed in the specific environment that needs them.

Something important is being optimised away. APT exists to recover it.

LinkedIn companion version

Post this natively on LinkedIn the same day the Substack publishes. Do not paste the full essay. This is a standalone piece built around one idea.

AI hiring tools are now standard across DACH enterprise recruitment. They are fast, scalable, and wrong in exactly the same way human hiring has always been wrong — only at greater scale.

Every AI hiring system is trained on historical data: past hires, performance outcomes, tenure patterns. They are pattern-matching machines that look backward. In stable functions, this works. In transformation roles, it fails — because the best candidate for a 2026 transformation mandate does not look like the best candidate from 2019. The organisations are different. The cognitive demands are different. The AI recommends profiles that succeeded in a world that no longer exists.

But the deeper problem is not the AI. It is what the AI was trained on.

Every assessment tool built in the last fifty years — human or algorithmic — was designed to evaluate the candidate. Nobody evaluates the environment. Nobody asks whether the organisation's operating conditions — its decision architecture, its tolerance for challenge, its actual versus stated autonomy — can activate what the candidate brings.

You can find the right person and place them in the wrong environment and watch them fail. Not because they changed. Because the conditions dampened the capability.

AI accelerated the old mistake. It made organisations more confident in answers that were always half-formed.

The environment is a performance variable. It has always been. We just built an entire industry around pretending otherwise.

What does your organisation's environment actually demand from a senior transformation leader — and have you ever assessed it?

Published by Tobias Temmen. APT — Adaptive Performance Talent. Rare minds, right seats. apt-match.com

Publishing notes:

- Substack: publish Tuesday or Wednesday, 07:00–09:00 CET

- LinkedIn: publish same day, 30 minutes after Substack goes live

- LinkedIn post: do NOT include a link in the post body — add apt-match.com link in first comment to avoid algorithmic penalty

- After 1 hour on LinkedIn: reply to every comment personally — early engagement velocity determines reach

- Suggested Substack subtitle: How AI is accelerating the most expensive mistake in executive search — and what it cannot fix