Why group by employer type
Company size predicts less about an AI PM interview than the kind of company does. A frontier lab, a large consumer platform, an AI-native startup and a bank adopting AI test different things, even for the same title. The directory below groups employers that way, with each careers page and our preparation page for the company.
Frontier labs
Labs that train their own models, such as OpenAI, Anthropic, Google DeepMind, Mistral AI, Cohere and Sarvam AI. Expect technical depth on evaluation and model behaviour, platform questions about launches and deprecations, and values conversations that weigh candour and safety judgment. Polished answers that agree with everything tend to land badly.
Big tech
Large companies with long-established PM loops, from Google, Meta and Microsoft to Amazon, Apple and Spotify. The familiar product sense, execution and leadership conversations remain, with AI depth added inside each. Scale changes the arithmetic: a cent per request is a large line item at a billion requests.
AI-native product companies
Companies whose product is the model's behaviour, such as Perplexity, Cursor, Harvey, Glean and ElevenLabs. They look for product taste, speed and hands-on fluency, and several run build exercises. Expect questions on quality bars and unit economics, because gross margin is often the company's central problem.
Enterprises and regional leaders
Banks, payment companies, marketplaces and enterprise software vendors bringing AI into large operations. That includes Indian employers such as Flipkart, Swiggy, Razorpay and PhonePe, and European ones such as Klarna and Revolut. Governance, permissions, regulation and change management dominate; getting the model to work is usually the smaller task.
Reading a posting
Look past the title. Postings say AI product manager, product manager for generative AI, technical product manager for ML, or simply product manager with AI in the responsibilities. The useful signals are what the team measures, who its users are (consumers, developers or employees) and whether the posting mentions evaluation. A team that writes about evals has usually learned where the hard part is.
Sources and further reading
44 employers building AI products, by type
Every row links to our preparation page for that company and to its own careers site, where open roles are listed. We do not track openings, and we only describe a loop where public evidence supports it, so treat your recruiter as the final word.
Search careers pages for these as well, then judge each listing by what the team measures rather than by its title:
Frontier labs
Mission and values rounds that reject strong candidates, technical depth on models and evals, and product questions about products that did not exist a year ago.
Big tech AI product teams
The classic product sense, execution and leadership loop, with an AI layer on top: evals, cost, and when the model should not be in the product at all.
AI-native product companies
Product taste, speed, and increasingly a live build: scope a prototype, ship a narrow path, add an eval, defend the trade-offs.
Enterprises and regional leaders adopting AI
Permissions, governance, procurement and change management. Getting a model to work is the easy half; changing how thousands of employees work is the other.
We host no job listings. Careers links go to each employer’s own site, which is the only place an opening can be confirmed.
Questions about this topic
We do not publish hiring counts, because we cannot verify them and openings change weekly. Large technology companies and AI labs advertise AI PM roles most visibly; each careers page has the current list.