What can AI-first thinking design, test and validate in 5 days?

When Jake Knapp wrote Sprint, building the prototype took a full day and a lot of ingenuity. Today you can do infinitely more with the tools at hand, there's been a significant shift in what's possible. Not having to worry about creating an illusion of an experience also frees up time to focus on the most important areas: creating alignment, choosing the right point of focus and thorough testing with users.

When the default plan is to create something for everyone, core design skills matter more than ever. On any key strategic project, I start with the people. That usually means meeting anyone with differing views 1:1 to understand their hopes and fears. For this project, leadership had strong instincts, but it was essential to identify all the open questions before using the sprint to provide a direction.

Before refining the scope, we ran 7 Ask the Expert interviews across leadership, commercial and product, including both UK and French specialists. The aim was to ensure anyone with a view was heard before narrowing the focus. It meant the final brief was informed by a full range of perspectives, and a clear brief could be agreed with a 360 view of business needs.

I followed this by speaking to 5 clinicians across different specialisms to gather initial research. I stored everything we learnt from experts and users alike in an AI knowledge base. This made it both quick and easy to call on the context gathered to make rapid, well-informed decisions.

With the expert input and user research in one place, I used the knowledge base to look for patterns across everything we'd heard. Three moments kept surfacing where a clinician might reach for their phone: the commute, mid-consultation, and moving around the hospital.

The research could tell us what clinicians needed, but not which need we were best placed to meet. That call came down to the sprint team's judgement: what we knew about how the system works, where it's strongest, and how the competitive landscape looks.

We prioritised the commute: a clinician walking in, head full, with the day not yet started. This formed the basis of the problem statement we worked to for the rest of the sprint. Focusing on one moment gave design decisions an anchor. Does this help a clinician arrive in control?

During ideation, we ran Lightning Demos and Crazy 8s sessions with design, product marketing and engineering.

Using AI tools for ideation meant designers and non-designers were able to get complex ideas and interfaces across in minutes. People often don’t feel confident sketching in sessions, and that’s followed by patient translation of ideas. The quality of what the group generated was high, and much of it shaped the final design.

Wednesday was build day, and it's where the gap with the old format was widest.

In a single day, using Lovable, I built a realistic end-to-end experience: sign-in, the core screens, and a full feature set.

Design is different now. We have so many modalities at our fingertips to maximise both input and output. The prototype used voice and image capture to make the experience truly seamless.

During testing, clinicians were able to use something that listened and responded in a realistic way, leading to more accurate feedback and results.

AI can give you ten good answers in a few minutes, but it can't tell you which one the business needs most.