AI removes the busywork. It doesn't change what's worth building.
AI-native transformation is a new discipline, and it gets redefined every time AI capability moves. The busywork of product is going away faster than most organizations have adjusted for. What survives is the part that was always the real job: understanding user needs across the market, understanding the business levers, and building what serves both. This page sets out my approach.
AI-ready product foundations
Good product decisions have always depended on high-quality context: real data based on clear instrumentation, direct contact with users, and agreed criteria for what good looks like. Now agents act on that context directly, so gaps in it still produce unnecessary features, built faster and putting more strain on the review chain.
How I work:
- Audit what the team actually decides on, and find the data that genuinely informs those decisions.
- Rebuild instrumentation around the decisions that matter, moving from top-level conversion numbers to the micro-conversions and engagement signals that explain them.
- Put product people back in direct contact with users, so judgment has a source outside the analytics stack.
- Define what good looks like before automating any part of a decision, so success criteria exist before an agent is asked to optimize against them.
AI in the product
The AI work that matters has actual ARR impact. I treat AI capability as part of the product and revenue model, competing for roadmap space on the same terms as everything else.
How I work:
- Start from the commercial outcome the business already tracks, and work backwards to where AI capability changes it.
- Ship narrow, high-conviction capability into the core workflow first, ahead of any general-purpose layer.
- Reposition the value proposition around better outcomes for the customer, rather than feature breadth that locks them in.
- Build the platform and integration strategy that lets AI capability compound instead of staying a set of one-off chatbots.
At a PE-backed e-commerce platform business, that meant two AI initiatives delivered, three more fully specified, and a roadmap for agentic commerce, inside a broader transformation that took the business from a Rule of 40 around 11 to around 40 within two quarters.
AI-native ways of working
This is the newest part of the discipline and the part still being written. Its shape is already visible: smaller teams holding pace, specification and research synthesis compressing from weeks into days, and development lifecycles being redrawn around what agents can reliably do. The largest multiplier here is still the oldest one, building less of what was never going to matter. AI makes the remaining work faster, which only pays once that decision is right.
How I work:
- Cut the roadmap items that were never going to move the numbers, which frees more capacity than any tooling change.
- Point that capacity at real customer contact, and use AI to multiply what it yields, analyzing every sales call and support conversation instead of a sampled handful.
- Redraw the discovery and delivery model around what the team can now do, including how research, specs, reviews and handovers work.
- Hold the commercial commitments fixed while the working model changes, so the transformation is measured on outcomes rather than adoption.
I work on this in two directions. I have operated a product organization through a significant reduction in headcount while holding the same commercial targets, which addresses the same problem: fewer people, unchanged expectations, and a delivery model that has to be rebuilt rather than stretched. And I take practice back from AI-native startups, advising founders across legal tech, sustainability and climate tech, and teaching product strategy at TU Berlin's startup incubator.
Judgment over confidence
A critical mistake is spreading through organizations moving fast on AI: cutting junior talent because everyone can now look competent with the right prompt. Confidence used to be a rough proxy for seniority. It no longer is, and anyone with a good tool can produce polished, confident, mediocre work. Telling the difference is the scarce skill now.
How I work:
- Separate output quality from delivery confidence when assessing people and work, because the two have come apart.
- Build taste deliberately through structured feedback on real decisions, rather than leaving it to accumulate on its own.
- Keep junior people close to customer contact and real tradeoffs, which is where judgment has always come from.
- Coach product leads directly, in regular sessions over months, so the capability stays after the engagement ends.
At an EdTech platform I ran individual coaching plans for five product leads alongside the transformation itself, and mentored emerging leads specifically so the new operating model would outlast my involvement.
Get in touch – let's work together