VP Product.

The platform was shipping agents into regulated enterprise workflows and the product surface had grown faster than the team's ability to make calibrated tradeoffs across it. Evaluation harnesses had been built in a hurry, latency targets sat next to cost-per-task ceilings with no clear hierarchy, and the founding team was making the same product calls each week without a senior owner in the room. The CEO wanted a VP Product who had lived inside an AI-native product loop — someone who treated evaluations as a first-class artefact rather than a QA tool.

The non-negotiables were sharp. Demonstrable experience owning an evaluation harness in production, fluency with cost-per-task and latency as product constraints, and pattern recognition for selling agents into enterprise procurement. The comp band was Series A standard with meaningful equity. Geography was open across the UK, US east coast and remote-friendly within timezone overlap. The CEO was explicit that they did not want a generic SaaS PM leader with an AI veneer.

The market for AI-native product leaders is thin and skewed. The first wave of generic SaaS PMs who have repositioned around AI tend to read fluently on language models and stall on the engineering substance underneath. The smaller bench of product leaders who have actually shipped agent products is concentrated inside a short list of applied AI companies and a handful of labs that have pushed product responsibility upstream. Reference networks inside that bench are tight; movement is observed in real time.

Our read was that the shortlist would have to be built from candidates who had personally owned the evaluation-latency-cost surface, not those who had managed teams that did. Search firms without practitioner depth in this segment tend to surface the same recycled product leaders — strong on narrative, weaker on the evaluation harness they were supposedly accountable for. We assessed each candidate on the specific product calls they had owned at their current company, the metrics they had moved, and the procurement cycles they had personally led through.

The shortlist composition leaned toward product leaders with hands-on agent or evaluation-loop experience over those with broader platform leadership. We held the bar on AI-native pattern recognition rather than seniority of past title, and we screened heavily for procurement and enterprise distribution exposure.

• Product leader from an applied AI agent company.
• Senior PM who had built and owned an evaluation harness.
• Product leader with enterprise procurement scars.
• Former founder who had sold agents into regulated workflows.

The hire came from a smaller applied AI company where they had been the senior product owner across an agent platform with overlapping enterprise distribution. What made them right was the specificity of their evaluation work — they could describe, unprompted, the harness they had built, the regressions it had caught, and the product calls it had forced. The CEO described the third interview as the first time they had felt understood on the substance of their company.

The close was clean. There was a competing conversation at an adjacent applied AI company that did not progress to an offer; the candidate's clarity on the brief and the fit with the founding team carried the final week. Brief to offer ran ten weeks, with the offer accepted before the formal reference cycle had closed.

“The engagement sharpened a view we had been forming. AI-native product leadership is not a layer on top of SaaS PM craft — it is a different shape of work, with evaluation harnesses, cost ceilings and latency targets functioning as primary product instruments rather than engineering concerns. Candidates who treated evaluations as QA read confidently and missed the role; candidates who treated them as a product artefact were rarer and arrived at the brief already half-aligned. We will continue to weight that distinction heavily in early screening for AI-native product seats.”

— Craig Oliver