Artificial Intelligence
AI Product Manager / GTM Engineer / Engagement Lead
The commercial surface of AI, and it is becoming genuinely technical.
What the job actually is
Owning an AI capability as a business outcome rather than a feature — what ships, at what quality bar, priced how. The interesting development is GTM Engineer, a title that barely existed two years ago and now appears across the corpus: someone who writes code in service of revenue, building the demos, integrations and internal tooling that close deals. It sits between archetypes 1 and 7 and is a genuinely good landing spot for an engineer who likes customers.
Titles this hides behind
The loop
Prepare like this
- Have a position on eval-gated launch — the specific bar you would hold a release to, and what you would do the week it is missed.
- Know token unit economics well enough to price a feature. Cost per request, per resolved task, and how it moves when the agent retries.
- Be able to say what you would cut. Every strategy case is secretly a scoping test.
- Prepare the failure story: an AI feature that shipped and underperformed, and what you learned about the gap between demo and production.
Red flags in the posting
"AI Transformation" with no product surface. Several postings under this banner are internal change management with no artefact to point at afterwards. Ask what shipped last quarter and who used it.
Compensation
A day that is actually representative
| Time | What you are actually doing |
|---|---|
| 09:00 | Pricing model. Cost per resolved task including retries, and what the tail does to margin. |
| 10:30 | A quality-bar conversation with engineering about what threshold justifies launch. |
| 13:00 | Customer call where the real question is what they will pay for versus what they say they want. |
| 15:00 | Roadmap trade-off, with an eval result as the input rather than an opinion. |
| 16:30 | Writing the launch note that explains honestly what the system does not do. |
Liability rested with the company for its assistant's statement.
Use it to say: “It is a pricing and scoping input, not just a legal one: what the product may assert is a design decision.”
A ban followed a leak and pushed usage onto personal devices.
Use it to say: “It informs how I think about enterprise adoption — the sanctioned path has to be better than the unsanctioned one, not merely permitted.”
Week 1 — mechanics
Days 1–3: Read the AI foundations book here, chapters 5 to 7.
Days 4–5: Build a trivial prototype yourself. Engineering credibility is visible within minutes.
Week 2 — economics
Days 1–3: Model cost per resolved task with a realistic retry rate, and the p95 user.
Days 4–5: Write your eval-gated launch position: the bar, the set, and what you do when it is missed.
Week 3 — rehearse
Days 1–2: The 85%-accuracy answer, out loud.
Days 3–4: A build-versus-buy case and a pricing structure you would defend.
Day 5: Five stories, including an AI feature that underperformed and what you learned.
The book for this field
Product & Delivery
Prioritisation that survives contact, writing that lets engineering start, discovery when your users are not consumers, measurement and its failure modes, dependencies and honest status, saying no well, pricing and unit economics, and scoping a non-deterministic feature.
Cross-cutting
Skills every archetype tests
These are shared across every field on this site — the same question asked in different vocabulary — so preparation here compounds rather than being spent once. The book above is specific to your field.
Practice the 22 AI questions → · Back to Artificial Intelligence →