How we usually run a sprint

We identified a key surface in the Messenger experience and wanted to iterate very quickly on it to generate ideas that could be built in the short term. In this kind of sprint, the classic approach would have been to gather designers around a table, start by reviewing all existing documentation, and then generate ideas together. We would vote, explore a few directions in design, critique them, and end with a final share-out. But I was wondering if AI could replace all of that.

How I ran this one with AI

The first thing I did was feed the AI with as much information as possible: sprint brief, user research, data, existing experiences, and competitive screenshots. Then I invoked a brainstorming agent through the BMAD method, using all of this context.

The agent guided me through multiple techniques like assumption reversal, cross-pollination, and Scamper to generate a large number of ideas based on the main user problems and the data. At the end of the session, all ideas were already prioritized.

I was so surprised by the quality of what came out. I know my brain pretty well, and I can tell, it cannot generate this level of output.

Designing and iterating with the team

I still had to move into Figma and build the explorations manually (old-school guy, I know). But thanks to the agents, I knew exactly which flows to prioritize, and I could focus directly on designing them, almost screen by screen.

Critical feedback was really important at this stage. That’s why I gathered the team to run open critiques, with no filters, to challenge the ideas and improve them.

At the end of the sprint, we presented the work to the leads and received very positive feedback. And because I didn’t want to end with a standard presentation, I built a website, to showcase the ideas, the strategy, and the prototypes in a more compelling way.

I got sad

At the end of this process, I was very proud of the output. But I also felt incredibly lonely and that loneliness made me sad. I looked at everything we had produced, all the screens, all the flows, the full story, and realized there was no one around to react to it.

I know it was a fast and straightforward sprint, but I missed the small things. The icebreakers, the team dinners, the spontaneous idea exchanges, the moments where everyone gets excited about something together.

The future of AI

AI was better than people. For brainstorming and structuring ideas, it was faster, more exhaustive, and always grounded in data. I didn’t need any colleagues to make this sprint effective. But I really missed them during the process.

AI was better than people.

So I started wondering what this means. Can the future of working with AI be collaborative?

I think that time spent together will become more valuable than ever. Not just to produce work, but to see how people think, how they react, how ideas evolve in conversation. At the same time, the efficiency that AI brings is hard to ignore. It feels like both are necessary, and we need to find a way to combine them.

It probably also depends on the type of sprint you want to run. Do you want to move fast and generate a lot of high-quality output? Or do you want to bring the team together, build momentum, and involve everyone in the decisions? Both approaches have value, and maybe can coexist.

Conclusion

I love AI. Sometimes more than I love myself (wait… what?). I feel incredibly empowered when I use it because it compensates for a lot of my design weaknesses. But I also miss people.

I don’t think I’ll stop working this way, I just don’t want to do it alone. Which means we really need to make AI collaborative before we all burn out.