The AI PM Upskilling Path
Learn the fundamentals, shadow a research scientist an hour a week, and build one model end to end.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 90%
Nika's concrete route for a non-technical PM to become an AI PM. It rejects both extremes — you never need to code or train models in the job, but learning the fundamentals anyway changes how you think and stops you trusting tools blindfolded. The three moves are: understand how AI product development differs from regular product development, shadow the research scientists already in your building, and ship one small model end to end using no-code tools.
Origin
Marily Nika's own recommended path, and the structure of her three-week AI Product Management course on Maven.
Core principles
- 01You will never need to actually train or code in the AI PM job — but learning it anyway gives you the confidence to not trust tools blindfolded.
- 02Learning the fundamentals first is like learning classical piano before the songs you actually want to play.
- 03Research scientists are not siloed anymore; a lot of what you need is on arXiv and in research blogs.
- 04The learning comes from doing it, not from reading about it or following Twitter.
How to run it
- 1
Map how AI product development differs
Learn the ML project life cycle — scoping, data acquisition, training, evaluation — and how it diverges from the standard product development life cycle you already know.
- 2
Shadow a research scientist one hour a week
If your company already has AI researchers, go talk to them and shadow them — spend an hour of their week watching what they actually do. Nika says this alone opens your mind to the potential you can identify.
Pro tip This is free, requires no permission, and is the highest-leverage step available to a PM inside a company that already does ML.
- 3
Get your hands dirty with the fundamentals
Take an online course (Nika names Stanford's Intro to AI on Coursera; CareerFoundry, General Assembly and Coding Dojo for cohort-style learning) and pair up with someone in the same boat so you don't drop out.
Pro tip Choose the format that fits how you learn — self-paced if you have the discipline, cohort-based if you need the accountability of a group.
Watch out Don't be intimidated into skipping this on the grounds that no-code tools exist; the point is understanding how the tool you depend on was built.
- 4
Build one AI product end to end with no-code tools
Use a no-code training tool (Nika recommends AutoML from Google Cloud) to train a custom model on data you have already collected and labelled, and take it all the way to a working demo.
Pro tip Present it to someone and take questions — the presentation is where the real understanding gets forced out.
Watch out No-code tools train the model but will not collect the data for you — you still need a labelled corpus.
- 5
Stay current from primary sources
Read technology newsletters (Nika names The Download from MIT Technology Review, and TLDR) plus academic and research blogs and arXiv, on the assumption that everything will become AI by default and AI will get sprinkled into non-AI publications anyway.
In the wild
In Nika's course, students who did not know how to code built a model that took an X-ray image found online as input and flagged what appeared to be wrong with the patient, then extended it into a recommender for next steps — using scraped X-ray images and no-code training tools.
→ A functioning end-to-end AI product built in three weeks by people with no prior coding ability; one student pair went on to raise funding for what they built.
A company inspecting wind turbines used to send people up huge ladders to check each one manually. They flew drones to photograph the turbines, uploaded the images to AutoML, and trained a model to identify which turbines needed maintenance.
→ Inspection triage dropped from roughly three weeks of work to a few hours, with crews dispatched only where needed.
Common mistakes
Being intimidated out of starting
Nika's central message is that a non-technical background is not a barrier — you can learn this, and the job itself never requires you to train or code. The overwhelm is the only real blocker.
Learning by consuming instead of building
Reading about AI and following Twitter doesn't produce the intuition. The course's value came from students building and presenting their own end-to-end product.
Is it for you?
Best for
Non-technical product managers who want to move into AI/ML product roles without going back to school.
Not ideal for
Engineers or data scientists already inside the ML stack, for whom the fundamentals step is redundant.
From the transcript
“I encourage people to just we talk to them and Shadow them and spend an hour of their week just just talking to them and…”
“number one figure out how it differ from General product management”
“you shouldn't be overwhelmed by these Technologies if you don't have a technical background because you can learn these things and as a PM you…”
From the episode
AI and product management
Marily Nika (Meta, Google)