Designing an AI maturity strategy
Moving a whole engineering department from individual AI tool use to a shared, measurable and sustainable way of working.
Context
An engineering department of seven teams, covering data and reporting, machine learning, backend, frontend and infrastructure. I had helped teams take their first steps with generative AI before, by setting up local LLM solutions at a previous company. This time the goal was bigger.
Challenge
Engineers were using AI one person at a time. Good tricks stayed with whoever found them, and there was no shared way to tell what was actually helping.
Approach
- Designed an AI maturity strategy built on four parts: context engineering, spec-driven development, sharing best practices, and learning and development.
- Formed and led a team of AI champions to carry the strategy into every engineering team.
- Started with a shared context foundation, so AI tools understand the team’s real systems and language.
Outcome
The context foundation serves as the bedrock of all LLM queries. This has saved countless hours of engineering and product work from debugging, Q&A, brainstorming, presenting, developing new features, and writing and improving documentation.
What I learned
AI adoption is a people problem first. The tools change every month. Shared habits, clear context and a safe space to experiment are what last.