I spent the 2nd and 3rd of July in Cornwall for Agile on the Beach. It is held at Falmouth University's AMATA campus, which is a genuinely nice place to spend two days: sunshine, sea air, and a crowd of people who are there to think rather than to sell you something. It made a real change from a typical development conference, and I came away with more to think about than I expected.
Prototyping with AI
I started the first morning with a workshop on prototyping with AI. The headline takeaway was a framework called COSID: Context, Objectives, Scope, Ideas (questions), Deliverables. It is a structured way to frame a prompt, and it works. Give an AI model a COSID-structured brief and you get something usable back. Give it a vague one-liner and you get a vague answer.
What I kept thinking about was how well COSID maps onto writing good tasks. When refining work it is easy to leave out the context and the deliverable and end up with something that means different things to different people. COSID is not just a prompting technique, it is a way of forcing clarity before you start. I want to try using it as a task template.
The workshop also reinforced something I have believed for a while: prototyping is cheap. A few hours building a rough version of something tells you more about the actual requirements than any amount of upfront design. Compared to a week or more of development time, the cost of getting it wrong without prototyping is significant. I want to push for more of it earlier in the delivery process.
The Traitors: Agile Edition
This was the most theatrical session of the two days. The workshop was themed around the TV show The Traitors: teams were given missions, faced sabotage, and used collaborative decision-making under uncertainty. The introduction was long, I will be honest about that, but once it got going the theming was brilliant. It created a dynamic that a standard retrospective exercise simply would not have.
The mechanic that stood out was using roundtables as retrospectives mid-activity. Rather than saving reflection for the end, the team paused, discussed what was happening, and adapted. It worked well. There is something worth borrowing there: shorter, more frequent retrospection rather than one big session at the end of a sprint.
Put people in a high-pressure, unfamiliar situation and you learn very quickly how a team actually operates. The format was a fun way to surface real team dynamics without the awkwardness of being told you are doing a team-building exercise.
Collaborative Modelling
The session I got the most from technically was on collaborative modelling, presented through the lens of a health insurer that had managed to get its IT and business teams speaking the same language. The pipeline they described went like this:
Domain Storytelling gets the business and IT into a shared conversation. Event Storming turns those conversations into domain events the whole room agrees on. Context Mapping draws the service boundaries before anyone writes a line of code. Then, and only then, you let AI handle the mechanical parts like generating API contracts.
The speaker made the point clearly: you cannot skip steps. Jump straight to AI without the collaborative modelling and you will get beautifully generated garbage, because the input is still wrong. Get the collaboration right first and AI becomes a genuine force multiplier rather than a shortcut to a bigger mess.
I would like to run this pipeline with my team. We build services that cross domain boundaries, and going through Domain Storytelling and Event Storming together would give us a shared model to work from and make the AI-assisted parts further down the pipeline much more effective.
The Beach Party
Wednesday evening was a party on Gyllyngvase Beach. It was exactly what it sounds like: a stretch of Cornish coastline, good weather, and a few hundred people from the conference. I had conversations I would not have had in a breakout room. That matters. The social side of a conference is where a lot of the value actually lives.
Purpose-Driven Leadership
The talk I was most glad I attended on day two was on purpose-driven leadership. The speaker had spent fifteen years leading teams, including building agile communities entirely on volunteers, where anyone can walk away at any time. His answer to that challenge was leading through purpose rather than authority or process.
This landed for me because I work in the public sector, and the mission is a real part of why I do the work. The speaker used the image of the Roman legionnaire's eagle: a physical symbol of what the unit stands for, something soldiers would risk everything to protect. He called it a north star, something a team can point at when they need to explain why they do what they do and how they decide how to do it.
I want to find that for my team. Not a vision statement that sits in a slide deck, but something concrete enough that it actually shapes decisions day to day. Working in the public sector gives us a strong foundation to build that from.
How Spotify Almost Killed Discover Weekly
The final talk I went to was on how Spotify's culture of autonomous teams made Discover Weekly possible. Daniel Ek said afterwards that he would have killed it himself if it had been down to him alone. Designers and senior engineers were sceptical. The team shipped it anyway, because Spotify's model gave them the room to.
What struck me was how evidence-based their decision-making was. The feature got axed at one point. It came back because the numbers said it should. Not gut feel, not advocacy, just the data making the case. That is the kind of product culture that is easy to admire and genuinely hard to build.
Autonomy without evidence is just noise. Spotify's model worked because the teams had both the freedom to act and the discipline to prove it.
A Good Conversation
Between sessions on day two I had a one-to-one with another DevOps engineer. It was the kind of conversation that is hard to have at a desk: shared problems, different approaches, no agenda. That human element is easy to undervalue when you are deciding whether a conference is worth the time, but it is often what you remember most.
The Boat Trip
The conference finished with a trip up the Fal River. Sandwiches, cake, good weather, and an hour or two of unhurried conversation with people I had met over the two days. A good way to end it.
What I'm Taking Away
Taking home
- Use COSID as a template for writing tasks, not just AI prompts
- Pilot the collaborative modelling pipeline (Domain Storytelling → Event Storming → Context Mapping) with my team before the end of the year
- Find the team's north star: something concrete enough to shape decisions, not just sit in a slide deck
- Push for prototyping earlier — a few hours of rough building beats a week of wrong development
Agile on the Beach is not a development conference. That is the point. Two days away from the keyboard, in the sunshine, talking to people who think about the same problems from different angles. Well worth going.