AI product design
Nine weeks on designing products whose core component is uncertain. Framing the problem so a model can help, prototyping with the real thing, testing under uncertainty, and handing over artefacts engineers can build from.
- Length
- 9 weeks
- Time
- 4 h a week
- Enrolled
- 963
- Next cohort
- January 2027 cohort
- Written and marked by
- Priya Raman Lead instructor, Governance
Four modules, one assignment each
Every module ends with work that is read by the person who wrote it. The assignments build on each other, so the order is fixed.
- 01
Problem framing
Which part of the problem is probabilistic, and which part the user needs to be certain about. Drawing that line before any screen.
- 02
Prototyping with models
A prototype that calls the actual model, not a mock. Finding the failure modes in week three instead of after launch.
- 03
User testing under uncertainty
Testing a product that gives different answers to the same question. Protocols, sample sizes, and what counts as a pass.
- 04
Handover artefacts
The spec engineers actually need: what the model may say, what it must never say, and who decides when it is wrong.
Who it is for
Product designers, product managers and user researchers who are being asked to put a model into a product and have noticed that the usual process breaks. The prototype works in the demo and fails in the hands of the third user, and nobody can say whose fault that is.
It is taught from the governance side deliberately: the questions of accountability and documentation are design questions, and they are cheaper to answer in week one than in a post-mortem.
How the nine weeks run
Two taught hours a week, half of them critique of the cohort’s work. Two hours on your own on the fortnightly assignment, all of which build one product: framed in week two, prototyped in week four, tested in week six, handed over in week eight.
The final brief is a partner company’s product question, and the review panel includes the engineer who would have to build what you specified.
