Claude Enablement Program

Teaching a whole company to work with Claude — and letting the savings compound.

What it is

A hands-on training and enablement program delivered to a 25+ person company, covering how to actually work with large language models day to day: token and context optimization, project mapping (giving Claude the right context for a codebase or process), documentation practices, and best practices for working with LLMs. The goal wasn’t a one-off workshop — it was to make the whole team fluent enough to build with AI on their own.

The challenge

Most teams adopt an AI assistant and then either underuse it or burn money on it — bloated context, no shared conventions, no idea what “good” looks like. Meanwhile the people closest to the pain (support, ops) can’t build their own fixes, and the dev team is stuck maintaining infrastructure by hand. The bottleneck isn’t the tool; it’s know-how.

What we did

• Token & cost optimization — taught the team how to structure prompts, context, and projects so Claude does more for less spend.
• Project mapping & documentation — a repeatable way to give the model the right context, so answers are accurate and repeatable across the team.
• LLM best practices — patterns for reliable, verifiable AI work instead of one-shot guessing.
• Enablement, not dependence — hands-on sessions so people could keep building after we left.

Results / impact

• −37% monthly Claude usage cost through better token and context practices
• Helpdesk started building their own code, automations, and solutions to optimize their own work — no longer waiting on the dev team
• The dev team migrated an unmanageable bare-metal setup to OVH Managed Kubernetes (with Claude), cutting infrastructure cost from 20,000 PLN/month to 2,500 PLN/month — an ~87% reduction (~210,000 PLN/year)
• AI capability spread across roles, not siloed in engineering

Stack & tools

Claude · Claude Code · MCP · prompt/token/context optimization · OVH Managed Kubernetes (infra migration outcome).

Status: Delivered.