SHORT ANSWER
A technical product manager owns products where the main complexity is technical rather than customer-facing: APIs, platforms, data pipelines, infrastructure and AI systems. The job is the same as any product manager — decide what to build and why — but the customers are often engineers, and the trade-offs are architectural. It requires enough technical depth to argue with an engineer without pretending to be one.
Answered by Andrew Crossley, AI Product Consultant & Product Partner · Updated 2026-08-11
AI products have made the technical PM the default shape of the role. If you cannot reason about latency, cost per call, evaluation and failure modes, you cannot scope an AI feature responsibly.
Companies frequently advertise 'technical product manager' when they mean 'product manager who will not slow engineering down'. Reading the job description properly saves both sides an interview loop.
Platform and API surface, data models, integration contracts, non-functional requirements, and the sequencing of technical work against commercial deadlines. Adoption by internal or external developers is usually the headline metric.
Read and follow code without writing production code. Query a database in SQL. Understand APIs, authentication, rate limits and webhooks. Read a system diagram. For AI products: prompting, retrieval, evaluation and cost per request.
You are not choosing the architecture. You are making the commercial consequences of each option visible — what it costs, what it forecloses, and how long it delays the thing customers are waiting for.
With AI build tools, a technical PM can produce a working prototype in an afternoon. It replaces a fortnight of specification argument and tends to settle disagreements faster than a document.
Time to first successful API call, integration completion rate, error rate by endpoint, p95 latency, cost per active user. Feature counts tell you nothing about a platform.
FROM EXPERIENCE
On CallFlow AI the whole product hinged on numbers a non-technical PM would not have asked for: latency per turn, cost per minute of call, and the false-positive rate on intent detection. Getting those wrong would have produced a demo that impressed people and a business that lost money on every call.
At Just Eat the same discipline applied at a different scale — integration reliability across hundreds of partner systems mattered more to the numbers than any feature we could have shipped that quarter.
Ownership of an API, platform, data or AI product; working directly with engineering on architecture trade-offs; defining non-functional requirements; and measuring adoption, reliability and cost rather than feature output.
SQL, reading code and system diagrams, understanding APIs and authentication, and — for AI products — prompting, retrieval, evaluation design and unit economics. Writing production code is not required.
No. Most technical PMs I have worked with came from support, QA, data or solutions engineering. Evidence of shipped technical products beats the degree at interview.
Roughly £65,000-£90,000 at mid level and £90,000-£130,000 at senior level in 2026, with AI-focused roles at the top of each band.
An app product manager owns an end-user surface and optimises activation and retention. A technical product manager owns the systems beneath it and optimises reliability, adoption and cost.
IN SHORT
What the role actually involves and how it differs from a classic PM.
Read moreContract, fractional and advisory engagements — current availability.
Read moreTurning LLMs, RAG and agents into products that hold up in production.
Read moreSeven stages from first idea to first revenue, with one output each.
Explore the Crossley MethodTHE FRAMEWORK
MORE ANSWERS
You become a product manager without a degree by producing evidence instead of credentials: ship something real, own a metric in an adjacent role, and document decisions publicly. Support, sales, operations and QA are the highest-converting entry routes because they give you customer contact and data. Hiring managers screen for judgement and shipped outcomes; almost none check for a degree at interview stage.
Four skills carry most of the job: customer discovery, prioritisation under uncertainty, written communication, and data literacy. In 2026 add two more: AI-assisted execution, and evaluation design for AI features. Frameworks, roadmapping tools and ceremonies are teachable in weeks; judgement about what not to build is the skill that separates senior PMs from everyone else.
Product managers usually fail for structural reasons, not talent ones: no clear mandate, no owned metric, no direct customer contact, an organisation that rewards shipping over outcomes, and an unwillingness to create conflict by saying no. Four of those five are fixable by the company. The fifth — avoiding conflict — is the personal skill most often missing in PMs who stall at mid-level.
Product managers use AI to compress the artefact half of the job: synthesising interviews, drafting specs and tickets, building clickable prototypes, summarising support and review data, and pressure-testing decisions. The best PMs also use it inside the product, designing evaluation sets and quality metrics for AI features. What they do not delegate is the decision, the customer conversation or the accountability.