Context
Bosch is a multinational automotive and industrial group, and it had to answer a question that reaches every large engineering organisation sooner or later: what happens when its own people start using AI tools. My team and I were Bosch’s AI supplier for that work. We ran the hackathons on our own platform and coached their engineers directly.
Business problem
AI coding tools arrive in an enterprise from the bottom up. Engineers install them, or ask to, long before anyone has decided what may be pasted into a model, which instances are approved, or what good use even looks like. The risk is not that the tools are bad; it is that adoption happens without a standard, and then has to be unwound.
My mandate
Act as Bosch’s AI supplier for engineering enablement: run the hackathons, coach their employees in prompt engineering and AI-accelerated development, and put the security standards and the internal tooling practices in place around that.
Decisions
- Two hackathons, run for Bosch employees on our own platform. A hackathon produces something that either works or does not by the end of the day, which is a faster teacher than a slide about prompting.
- Onboard the teams to the tools they would actually use — Cursor, Claude Code and Copilot — with context management taught as part of the tool, not as an advanced topic afterwards.
- Teach prompt writing as a skill with rules, so quality did not depend on which engineer happened to have a feel for it.
- Establish security standards and responsible-use guidelines for enterprise AI tool adoption, so the answer to “may I use this on our code” was written down rather than decided case by case.
- Set up MCP best practices and custom skills inside their internal Claude Code and Cursor instances, so the practices lived in the tooling the engineers open every morning instead of in a document.
Delivery
Recurring workshops and training sessions run by my team and me on prompt engineering and on AI-accelerated development, two hackathons on our own platform, the security and responsible-use guidelines written for their context, and the MCP setup and custom skills installed in their own instances.
Outcome
Bosch employees work with AI-assisted development tools inside a set of standards their security people agreed to, on internal instances, with prompting and context management taught rather than improvised. No adoption or usage numbers are published for this engagement, so none are given here.
Reuse
The hackathon format, the prompt-writing and AI-IDE curriculum, the security and responsible-use guidelines and the MCP practices are the same material I use for other engineering organisations; each engagement adapts the guidelines to the client’s own security posture and tooling.
Evidence
The hackathon count, the tools, the guidelines and the MCP setup are project records. Every outcome here is qualitative on purpose: no adoption, productivity or usage figures are published for this client, and none are estimated here.