Six weeks to AI-native engineering at Joulen

Joulen builds smart solar and battery storage technology, and the business is growing fast — with an ambitious roadmap for the next generation of its platform. Its leadership believed an AI-native SDLC would raise the velocity of every team, and they were right. But believing it and working that way are different things.

The tools were available and the team was curious. What was missing was the working practice: context engineering, getting the most from AI harnesses, keeping quality high as throughput rises, and making spec-driven development an everyday habit. Joulen didn't want a vendor to deliver the roadmap for them — they wanted their own people to come out the other side working differently.

The engagement followed the three-stage shape of our AI Enablement service: collaborative discovery, embed and transfer, then withdraw to autonomy. Discovery opened with a workshop with Joulen's executive team — priorities agreed, success defined, sequencing set — and within a week of signing we were embedded on-site. Two weeks in, we delivered a landscape report: an independent read on where capability sat, and where the bottlenecks would surface as delivery accelerated.

The transfer stage worked systematically through the engineering organisation, team by team — structured, hands-on sessions that built each team's AI practice in the context of their own work: the context engineering, harness skills and spec-driven habits the teams had been missing, plus DORA metrics rethought for agentic delivery, shaped together with Joulen's engineering leadership.

Alongside the sessions, we built the team a shared AI-enablement repository — enterprise context, conventions and a curated set of skills — so every engineer's coding agent starts with a built-in understanding of Joulen: the business, the codebase conventions, and how the estate fits together. Nobody re-explains the company to their tools each morning, and the repository is a foundation the team owns and extends.

Throughout, we measured rather than assumed. Adoption surveys — run on Signals, our own feedback product — tracked how the shift was landing: who was working AI-first, who was confident, and where friction remained. By design, the engagement grew lighter every week: collaboration first, then facilitation, then stepping back as the teams became self-sufficient.

By the engagement's conclusion, Joulen engineers were routinely running multiple coding agents concurrently, exploring and experimenting with cutting-edge technologies, and putting real momentum behind the platform roadmap that prompted the shift. The adoption survey told the same story: 100% of engineers working with AI daily, and the product team using it most days.

The shift outlived the engagement. The shared context repository now lives in Joulen's own tooling, and the conversation inside the business has moved on to what comes next: enterprise context, governance, the layer beyond engineering. Every primary objective was met inside the original six weeks — no extension. We handed over and stepped out.