"Head of AI" is one of the fastest-growing titles in tech — and one of the least standardized. Before you post the role or brief a recruiter, you need to figure out which of three very different jobs you're actually trying to fill.
Most early-stage startups need Archetype 2 — or a hybrid of 2 and 3. Hiring Archetype 1 when you need Archetype 2 is one of the most expensive mismatch mistakes in startup AI hiring.
For the archetype most startups actually need, here's what to look for:
Still writes production code. Not just reviews PRs — writes and ships. At an early-stage company, a Head of AI who isn't hands-on quickly becomes a bottleneck rather than a force multiplier. Has shipped AI systems, not just trained models. Experience deploying AI in production — handling latency constraints, monitoring model drift, building reliable inference pipelines — is different from research experience. Look for production ML in their portfolio. Product-sensible. They understand that a working AI system that doesn't solve a real customer problem is a science project. They push back on technically impressive features that don't move the product. Can attract AI engineers. At the level of Head of AI, they'll need to recruit and retain AI engineers. Their technical reputation and ability to build a team matters as much as their individual output.The best candidates for this role are almost never actively applying. They're typically:
AI lab alumni (OpenAI, Anthropic, Google DeepMind) are a starting point, not a destination. The question is whether their experience is on the applied side — systems and deployment — versus the research side. Lab experience alone doesn't answer that question.
Standard interview loops miss the signal for this role. Instead:
Deep technical system design. Not a whiteboard puzzle — a substantive discussion of an AI system they've built. How did they think about evals? What broke in production and how did they fix it? What would they do differently? Product judgment. Present a real AI use case your company is considering. How do they think through feasibility, trade-offs, and prioritization? Do they lead with what's technically interesting or what's most likely to move customer outcomes? Team building. At the Head of AI level, how they'd hire and develop AI engineers matters as much as what they'd build themselves.If you're building a research-first company (foundation model training, novel architectures, academic-adjacent work), a Head of AI focused on applied systems won't be the right hire. And if you're earlier than product-market fit, you may not need a Head of AI at all — an experienced applied ML engineer who can own AI development is often the right first hire.
Q: When should a startup hire a Head of AI? A: When you have a clear AI use case in your product, the engineering team is too small to own AI development alongside everything else, and the AI work is complex enough to require dedicated leadership. This is usually Series A to B. Earlier than that, an applied ML engineer who can also lead AI direction is often a better fit. Q: What's the difference between a CTO and a Head of AI? A: A CTO owns the full engineering organization and technical strategy. A Head of AI owns the AI/ML function specifically — the models, the data pipelines, the AI product strategy. At companies where AI is the core product, these may overlap significantly. At companies where AI is one of several technical capabilities, they're distinct roles. Q: How do you evaluate a Head of AI candidate? A: Ask for a deep dive on an AI system they've built end-to-end — from problem framing through production deployment. Look for production experience (not just research), product sensibility, and evidence that they can build and lead a team. The ability to distinguish applied from research experience is a core screening signal. Q: What compensation does a Head of AI expect? A: Head of AI compensation is meaningfully higher than a senior software engineer — both in base and in equity — reflecting the combination of technical depth, product judgment, and leadership expected. The exact range varies significantly by stage, company, and candidate background. A recruiting partner with AI hiring experience can help calibrate against current market expectations. Q: Does Recruiting from Scratch fill Head of AI roles? A: Yes. We've worked with AI-first startups on AI and ML leadership searches, including distinguishing between research-oriented and applied profiles. We source from applied ML engineers with production experience — not just candidates with impressive-sounding lab backgrounds.Tell us about your open roles and we'll start sourcing within 48 hours.
