What Makes a Great Leader in DeepTech & AI (And Why Most Executive Hires Fail Within 18 Months)
Half of executive hires fail within 18 months. After years sitting across the table from boards, founders, and CHROs, I've started to understand why. They're all trying to do the same thing: look into their crystal ball and answer one question. What will predict someone being a great leader? You'd think we'd have cracked […]
Half of executive hires fail within 18 months. In AI and DeepTech, the number is arguably higher — because the gap between what the role demands and what most hiring processes actually assess for is wider here than almost anywhere else in technology.
After years sitting across the table from founders, boards, and investors building frontier technology companies, I've started to understand why. They're all trying to answer the same question: what will predict someone being a great leader in an environment where the science is moving faster than the business model, and the commercial pressure is mounting faster than the technology is ready?
There is no shortage of frameworks and assessments. Yet multiple independent studies — from the Center for Creative Leadership to Heidrick & Struggles — keep landing on that same brutal failure rate. In DeepTech and AI, we see it play out in a particularly recognisable pattern.
The presence trap in AI & DeepTech hiring
A common mistake in AI and DeepTech executive search is confusing technical credibility with leadership capability. Or in other words, judging the wine by the label.
There is a certain type of candidate who interviews brilliantly in this space. They can speak fluently about transformer architectures, decentralised systems, or quantum error correction. They command a room of engineers. They make you feel like they have everything figured out — technically.
But leading an AI or DeepTech company requires a fundamentally different skill set from doing world-class research or building a technically elegant product. The ability to make hard resource allocation decisions when a research roadmap is not converging. To manage investor expectations when the science timeline is slipping. To build and retain a team of exceptional engineers and scientists in a market where they have more options than ever. Technical depth gets you in the door. It does not help you navigate the day-to-day reality of leading a company through those challenges.
What we see, and what research confirms, is AI and DeepTech companies over-indexing on technical pedigree and under-indexing on the qualities that actually determine whether an executive will succeed in the role.
What the research shows
The executive search industry, particularly in DeepTech, relies far too heavily on gut feel and pattern matching. We have decades of research on leadership effectiveness. Most of it gets ignored in favour of publication records and lab pedigree.
Self-awareness is foundational, but rare. Organisational psychologist Tasha Eurich found that while 95% of people think they are self-aware, only 10–15% actually are. In AI and DeepTech, this manifests in a specific way: leaders who have been the smartest person in most rooms they have inhabited often have significant blind spots about how they land with people who do not share their technical background — commercial teams, investors, regulators, customers.
Cognitive flexibility beats domain expertise. The ability to update your thinking when new information arrives, hold contradictory ideas without ego getting in the way, and change course when the situation demands it. In AI and DeepTech, where research directions pivot, foundational assumptions get invalidated, and competitive landscapes shift in months, this quality separates leaders who adapt from those who double down on failing strategies.
Attitudes drive most failures, not skills. Leadership IQ tracked over 20,000 new hires and found 89% of failures came down to attitude: coachability, emotional intelligence, motivation, temperament. Only 11% were due to lack of technical skills. In an industry where technical skills are typically exceptional, the attitude dimension becomes even more critical — and even more frequently overlooked.
The track record problem in frontier technology
"They led research at a tier-one AI lab." "They were employee number eight at a blockchain unicorn." Sounds impressive. Maybe it is. But track records in DeepTech are particularly noisy.
The question that rarely gets asked: what was the context? Were they operating in a resource-abundant research environment with long time horizons — and are they now being hired into a capital-constrained startup that needs to ship a commercial product in 18 months? Were they a key decision-maker, or were they surrounded by exceptional people who carried the operational complexity while they focused on the technical work?
The patterns we see most often in AI and DeepTech executive failure: the research leader who thrives with long time horizons but cannot operate under commercial urgency. The technical co-founder who can build the product but cannot build the organisation around it. The big-company AI executive who can run a mature machine learning platform but cannot build the go-to-market motion from scratch.
Past performance is not irrelevant. But it is only useful if you understand why someone succeeded in their previous context, and whether that context actually maps to the one they are walking into.
What I actually look for in AI & DeepTech leaders
After years of placing executives in AI and DeepTech companies, here is what I have learned to pay attention to.
How do they talk about failure? In AI and DeepTech, failure is structural — most research directions do not pan out, most product hypotheses are wrong, most go-to-market assumptions need revision. Leaders who can discuss what they got wrong, what they learned, and what they would do differently are the ones who actually improve over time. The ones who deflect or rationalise every failure are the ones who repeat it.
How do they bridge technical and commercial thinking? The best AI and DeepTech executives I have placed can hold a rigorous conversation with a research team at 10am and a persuasive conversation with a customer or investor at 2pm. They do not simplify the science dishonestly, but they can translate it into value without losing the room. This is rarer than it sounds.
What is their relationship with uncertainty? In frontier technology, certainty is a liability. The leaders who perform best in AI and DeepTech environments are those who can say "I do not know yet, but here is how I would figure it out" — and mean it. Leaders who need to project certainty in an inherently uncertain environment make worse decisions and create cultures where people stop surfacing bad news.
How do they talk about their teams? Leaders who consistently claim credit for collective scientific or commercial achievements tend to struggle retaining the exceptional engineers and researchers that AI and DeepTech companies depend on. The best leaders I have placed are obsessively focused on creating conditions where exceptional people can do their best work.
Can they disagree with you in the room? Leaders who cannot challenge someone they are trying to impress during an interview will not have the difficult conversations a DeepTech role demands — telling the board the research timeline needs to extend, telling the investors the product market fit is not there yet, telling the team a fundamental technical assumption was wrong.
The culture question in AI & DeepTech
"Culture fit" is a phrase I hear constantly in AI and DeepTech hiring. It often means one of two things: "someone from a top research institution" or "someone who thinks like we do." Both are problematic.
The tension in most AI and DeepTech companies is between research culture — which values depth, rigor, and long time horizons — and commercial culture, which values speed, iteration, and customer feedback. The best leaders do not choose one side. They create environments where both can coexist productively. That requires cognitive flexibility and genuine respect for both ways of working — not just tolerance for the one that is not their native mode.
Better question: "Will they strengthen the tension we need between research excellence and commercial urgency?" That is almost always the real culture question in an AI or DeepTech company. Not whether the candidate reminds the hiring committee of themselves.
Why AI & DeepTech companies keep getting this wrong
Time pressure. AI and DeepTech executive hiring often happens under intense pressure — a funding round has closed, a product needs to go to market, a technical co-founder needs a commercial counterpart. The urgency leads to shortcuts that are expensive to reverse.
Anchoring on credentials. "DeepMind, OpenAI, or Google Brain" feels like a safe signal. But research pedigree and commercial leadership capability are genuinely different skills. Using the former as a proxy for the latter is one of the most common and costly mistakes we see.
Interview theatre. Traditional interviews in AI and DeepTech often become technical seminars. They test for what the candidate knows, not for how they would actually operate in the role. Structured behavioural interviews, with rigorous reference-checking and real scenario challenge, consistently outperform unstructured technical conversations as predictors of executive success.
Confirmation bias. Once a board or founding team gets excited about a candidate from a prestigious lab or a successful AI company, every subsequent signal gets interpreted positively. The warning signs become "manageable." They rarely are.
A different approach for AI & DeepTech hiring
Start with the real challenge, not the role. Before defining the ideal candidate, get specific about the situation. Is the company moving from research to product? From product to scale? From a European base into North America? What has failed before, and why? The answer determines the profile — not the other way around.
Assess for learning agility, not just domain expertise. The AI and DeepTech landscape changes fast enough that what worked three years ago may be irrelevant. The executives who perform best over time are those who update quickly, not those who have the deepest knowledge of yesterday's paradigm.
Build in real friction. The process should include genuine challenge — not adversarial grilling, but real conversation where candidates have to think on their feet, handle disagreement, and demonstrate how they reason under pressure. A comfortable process selects for comfort, not for leadership capability.
Reference for the specific context. "When things got difficult, how did they show up?" "What surprised you about them?" "Who did they struggle with, and why?" These questions reveal far more than standard strength-and-weakness conversations, particularly in the high-stakes, high-ambiguity environments that define AI and DeepTech leadership.
The uncomfortable bit
Most executive hiring in AI and DeepTech optimises for reducing reputational risk rather than finding the best person. Companies hire candidates who look impressive on paper, come from the right institutions, and will not embarrass anyone if things go south.
But leadership in frontier technology is inherently risky. The candidates who will actually transform your organisation often have unusual trajectories — commercial experience outside the research mainstream, sector transitions that look lateral on paper but built exactly the capabilities the role demands, a willingness to disagree with the conventional view of what the technology can and cannot do.
The best AI and DeepTech hiring decisions I have been part of were ones where the client was willing to think differently about what they actually needed. Not "who looks like a leader in our space" but "who will lead us where we need to go." Ask harder questions.
Need advice on a senior hire in AI or DeepTech?
Schedule a consultation with our specialist team today.
Schedule a consultation with our specialist team today.