AI moves the lines between every role. I help your team adapt.
Tools like Claude Code don't just make your developers faster. They change what everyone on a product team can do. The teams that get ahead are the ones that adapt on purpose, fitting these tools into how they already work instead of bolting them on.
The tool is the easy part
Dropping Claude Code or Cursor into a team takes an afternoon. Changing how the team works around it takes longer, and that's where the value is. When one person can suddenly do what used to take three, the old boundaries between roles stop making sense. The job titles stay the same. The work underneath them shifts.
Most teams that struggle with AI aren't struggling with the tools. They're struggling because nobody redrew the map of who does what.
How each role flexes
Here is the shift I see on product teams that adopt AI well:
- Product managers build, not just spec. A PM can now put working front-end code in front of users, not just a Figma mockup or a written ticket. The loop from idea to feedback gets shorter, and the handoff gets thinner.
- Designers ship their own systems. Instead of handing a design over and hoping it survives implementation, a designer can build the design system and wire it into the codebase directly, keeping the craft intact all the way to production.
- Front-end developers move toward product. With the mechanical work of turning designs into components largely handled, front-enders are freed to own outcomes: the flows, the edge cases, and the details that decide whether a feature actually works.
- Full-stack and backend engineers move toward architecture. As more code gets generated, the scarce skill becomes judgment: system design, data models, and the guardrails that stop a fast-moving team from shipping a mess.
Guardrails matter more, not less
AI makes it cheap to produce code. That raises the value of everything that keeps code trustworthy: review, testing, clear architecture, and a shared sense of what good looks like. Teams that loosen those in the name of speed pay for it later, usually all at once.
So I don't help teams cut corners to go faster. I help them strengthen the guardrails, so they can trust what AI produces and keep the quality bar exactly where it was.
What I do
I work with your team, not around it. A typical engagement looks like this:
- Map how you work today. Where the time goes, where the friction is, and where AI actually helps versus where it just adds noise.
- Pilot on real work. We adopt the tools on a live piece of work, not a toy example, so the team feels the change firsthand.
- Set the guardrails. Review, testing, and architecture practices that let you trust AI-generated code instead of fearing it.
- Coach each role. Practical, hands-on support as PMs, designers, and engineers grow into their new shape.
- Embed it in the process. So the new ways of working stick long after I've gone.
Who this is for
Product teams that want to adopt AI seriously: in-house teams, scaleups, and studios who care about shipping quality software and want to move faster without losing the practices that got them here.
About me
I'm Ben Leavett. Over 17 years I've led engineering and product teams at Microsoft, SwiftKey, and Glovo, building products from zero to millions of users. I use these tools daily to ship real software, so I know where they help and where they get in the way. I run Cadence Digital, based in France and the UK.