AI design sprints that leave a decision

An AI design sprint is a timeboxed sprint where the model is a teammate. Faster prototypes. Same kill criteria. If you leave with a demo and no decision, you ran a show, not a sprint.

People hear "AI" and forget the rest of the sentence. The rest of the sentence is design sprint. Problem, options, prototype, test, decide. The model does not get a vote. It gets a job: draft, generate, cluster, stub the first version so humans can argue about something real before lunch.

Enterprise design sprints already had this shape. The AI version is not a new religion. It is speed of options. I would rather see twenty ugly drafts in a morning than three polished slides on Friday.

How to run it

Frame the problem before anyone opens a tool. Then let the model flood the table. Then kill most of it. Humans test with humans. A facilitator keeps the room honest. Someone who owns the process has to be there. Someone who does the work has to be there. Someone who can kill the idea has to be there. A gallery of interested stakeholders will turn it back into a workshop.

Two days can be enough. Inkeros ran a two-day innovation sprint. Stretching to two weeks usually means nobody wanted to decide. The clock is a feature. Without it, the model will generate forever and you will call it exploration.

Design thinking is the parent method. Lean startup is the loop after the sprint if the idea deserves a longer life. Enterprise AI is the habitat: a hated workflow, not a playground. AI training is how the people in the room know what the model is for.

Infinitev generating landing pages to test ideas fast is the spirit: AI in service of a learning loop. Steal the loop.

What to train

You can teach the sprint mechanics and the AI craft in the same week if the problem is live. Collective Campus trains design thinking and sprints and AI skills for teams. Two links, because this hub sits on both. Do not buy a week of tool tourism and call it a sprint.

Geos: AU, NZ, HK, SG, the US, the UK and Canada. A sprint in a New York corporate is not a sprint in a Singapore bank. The clock is the same. The constraint is not.

What we publish here

AI design sprints that ship a decision. Design thinking that survives the board pack. Inkeros. Infinitev. The design thinking and AI training guides.

If your last "AI sprint" produced a demo and a parking lot, you did not sprint. You performed. Book a shorter room and a harder no.

The model will happily generate a strategy. Do not let it. Strategy is a human kill. Use the model for volume: sketches, copy variants, process maps, sample data, a first prototype. Then argue. The argument is the work. Teams that skip the argument because the output "looks done" ship plausible nonsense. Plausible is the danger. Ugly and testable is safer.

I would score a sprint on one question: what did we decide to stop, start or continue? If the answer is "we learned a lot," you learned how to avoid a decision. Write the decision on a page. Date it. That is the artefact. The prototype is evidence. The decision is the product.

Do not invite the whole leadership team to day one. Invite them to the decision. Leadership in the working room turns the sprint into a performance. The people who do the work go quiet. The model fills the silence with plausible slides. You will clap. You will not have tested anything.