There is no cross-company data on what AI fluency actually looks like inside a working team. Nobody has it, including us. The design partner programme is how it gets built — a small group of companies measuring the same thing the same way, long enough for the numbers to mean something.
8 slots, because that is how many one founder can work with properly. When they are taken, the page says so.
Two paragraphs are enough: what you use AI for today, and what you cannot currently answer about it.
This page describes the programme, not a contract. The rate, the data scope, the term and the exit are agreed in a written partner agreement before anything starts — and you get the draft before you decide.
Written out in full, so you can decide from this page whether it is worth a call.
One working call, plus written feedback when something in the product is wrong. No steering committee, no slide reviews.
Aggregated assessment and competency results from your workspace, feeding the cross-company benchmark. Nothing individual, nothing identifying — the scope is written down on the data page. The fluency score and the connected usage signals are not built yet; partners are who we build them with.
Including the one that ends the pilot. A partner who quietly stops using the product teaches us nothing.
Being referenceable helps us. Staying an unnamed logo-free participant is a valid answer and does not change anything else in this list.
Partner-raised issues are prioritised ahead of the public backlog for the duration of the programme, and you see where they land in the changelog.
The person answering is the person writing the code. No support tier in between.
Your team's assessment and competency distribution from the first month, and the Hoffman 60 / 30 / 5 read as the fluency scale lands. The cross-company comparison follows once an industry has the 20 tenants our k-anonymity threshold requires — partners see it before any aggregate is published.
Your rate is agreed with you before you commit, written into the partner agreement, and held for 24 months from general availability. Cloud billing has not launched, so there is no list price to discount — that is exactly why the number is fixed in a document rather than on this page.
Copilot, Claude, ChatGPT seats — something with a line in the budget and no answer to what it returns.
Small enough that one person can drive the rollout, large enough that a fluency distribution is more than noise.
In the EU we work with your works council rather than around it — the data scope is designed for that conversation, not against it.
Write what you use AI for and what you cannot answer about it today. You get a reply from the founder, not a sequence.
