Before
Your engineers still write code, line by line. Three of them are 3× faster than the rest.
Tools installed. Adoption stalled.
Your team has the tools.
Let's make it actually work.
I fix broken AI coding setups, build custom plugins for your stack, and train developers to ship fluently — Claude Code, Cursor, Codex, whatever you use.
Claude Code
Cursor
Windsurf
GitHub Copilot
Codex
You bought the tools. You ran the workshop. Six months later, a few devs fly — the rest still paste into a chat window.
- Licences aren't the lever. Your fastest devs aren't smarter — they have a tuned CLAUDE.md, the right hooks, and a few custom skills.
- Workshops decay. Without project-level setup and pair-coding, habits revert within weeks.
- Generic configs fight your codebase. The agent feels useless because it doesn't know your stack, conventions, or tools.
- Every month of waiting widens the gap. Competitors who master AI ship faster, learn faster, compound the advantage — and it shows up in your roadmap, not theirs.
Adoption quality is the lever — project-level setup, custom tooling, and habits embedded in real code review.
What changes
When adoption quality is in place, the day-to-day shifts.
After
Your engineers supervise agents — and ship at a pace they couldn't reach alone.
What's on the table
Three things I do. We'll work out which you need on the audit call.
Fix your AI setup
Your CLAUDE.md / AGENTS.md / .cursorrules is generic boilerplate. Your developers still paste into ChatGPT instead of running their agent properly. The tools are installed but the workflow never clicked.
What you get
- Project instruction files rewritten to match your stack, conventions, and team workflow
- Tool configs (hooks, slash commands, IDE settings) tuned to your repos
- A short walkthrough with your team so they own the new setup
Plugin engineering
Your AI agent doesn't know about your Linear / your DB schema / your internal API. There's an MCP for that — but nobody on the team has time to build, integrate, or maintain it. Or you've installed seven community MCPs and the agent is overwhelmed.
What you get
- Custom MCP servers, skills, agents, and slash commands built for your stack
- Off-the-shelf MCPs integrated cleanly, with auth and security review
- Working code in your repo — deployed, documented, and yours to extend
Programmer training
Your team has the licences. Some devs are 3× more productive; others run the agent in chat mode like it's 2023. You don't need another video course — you need someone in the trenches with your code.
What you get
- Hands-on workshops on your actual codebase, not toy projects
- Pair-coding sessions with each developer until the workflow clicks
- Code review of AI-assisted PRs, so habits stick after I leave
Hire one. Hire all three. Your call.
Most engagements start with one row. We'll know which after the audit call.
Wondering if I'm your person?
Ten quick questions in, you'll know whether I can actually help — and how.
How we'd work together
Three steps. The first is free, and the only one you have to commit to.
01
Free audit conversation
We talk for about 30 minutes. You tell me what your team is using, what's broken, and what 'better' would look like. I tell you which of the three service rows — or none — would actually move your numbers. No commitment, no slides.
02
Diagnostic artefact
After the call, I send you something written — a short take on where your team stands, where the gaps are, and which services would move the needle. Sometimes a one-pager, sometimes an annotated map of your setup. Yours to forward internally.
03
Custom engagement
We work the plan together. Remote by default, on-site when it matters. Deliverables land in your repo, not a PDF. Habits land in your team — through pair-coding and PR review until they stick after I leave.
Start with the audit conversation. It's the only step you have to commit to.
About me
Luka Breitig
Software engineer & AI implementation consultant
I'm a software engineer and trainer based in Poland with an academic background in psychology and business (B.Sc. Osnabrück, M.A. Zeppelin University). I help developers and teams adopt modern AI coding tools through hands-on workshops that focus on practical workflows rather than theory.
I work primarily with engineering teams adopting AI coding tools across stacks — Elixir, TypeScript, Python, whatever you ship. My background in psychology and business means I read team dynamics as carefully as I read code.
How I work with engineering teams
- I work in your repo, on your stack — not toy projects or vendor demos.
- I ship artefacts (skills, MCPs, CLAUDE.md, hooks) that you own and can extend.
- I pair with individual developers until the habit is theirs, not mine.
What I bring that an internal effort doesn't
- I've shipped these setups across many stacks — I know which patterns fail and why.
- I have no internal politics, no project I'd rather be on, no team-lead role I'm protecting.
- I leave when the team can run it without me — that's the goal from day one.
Practitioner first, trainer second. I build production software with the same AI tools I teach — this portfolio site included.
Training or consulting?
Both bring AI-assisted coding into your team. They're booked, billed, and delivered differently.
Training
Scheduled programme
- What you book
- Workshop or programme with a fixed agenda
- What you get
- Skills your developers carry into every project
- Measured by
- Workshop days, participants, satisfaction
- Who buys
- L&D managers and training providers
- When I leave
- Your developers carry new skills
Implementation consulting
Embedded engagement
You're here
- What you book
- Hands-on engagement embedded in your team
- What you get
- Configured tooling and workflows your team adopts
- Measured by
- Adoption stickiness and productivity shift
- Who buys
- Tech leads, engineering directors, CTOs
- When I leave
- Your stack carries new defaults
Still not sure? Many engagements start with a workshop and continue as consulting — or the other way round. Get in touch and we'll work out which fits.
Frequently asked questions
Common questions about AI implementation consulting and team adoption
Training teaches your team what AI coding tools can do and how to use them effectively — it builds knowledge and skills in a structured workshop format. Implementation consulting goes further: I work directly with your team's codebase, configure tooling for your specific stack, set up project-level instructions and custom skills, and pair with developers until the workflows are embedded in daily practice. Training gives your team the knowledge; consulting ensures the knowledge turns into lasting habits and measurable results.
You absolutely can, and the best clients have tried. The difference: I've shipped these for dozens of teams across stacks. I know the failure modes — over-specified rules that block the agent, generic conventions that the agent ignores, hooks that fight your CI. You get a working file in days, not weeks of iteration.
Focus on outcomes, not activity metrics. Useful indicators include cycle time (how quickly features move from idea to production), PR throughput, time spent on boilerplate versus creative problem-solving, and developer satisfaction surveys. Avoid measuring lines of code or commits per day — these incentivise the wrong behaviour. I help teams establish a baseline during the audit phase, then track concrete improvements over the engagement. Most teams see measurable gains within the first two to three weeks of properly configured agentic workflows.
The hero says tool-agnostic and means it. The patterns that matter — project-level instructions, custom tooling, agent-aware workflows — port across stacks. CLAUDE.md, AGENTS.md, .cursorrules, and Copilot custom instructions are different filenames for the same underlying problem: telling the agent what your codebase actually is. I've configured setups across Claude Code, Cursor, Windsurf, Codex, and GitHub Copilot; the artefacts I ship in your repo are tuned to whichever you've standardised on. If you're still choosing between tools, that's a useful conversation for the audit call.
Engagements can be fully remote — pair-coding, PR review, and configuration all run cleanly over video and async, which keeps things lean and avoids travel overhead. On-site days are possible where they make sense: an audit kick-off, a focused workshop day, or moments where face-to-face meaningfully shifts team momentum. Travel is billed separately at cost. I'm based in Poland and travel readily across Europe; further afield works when the engagement warrants it.
Have a question about your team's AI adoption?
Tell me what's broken. I'll fix it.
Setup, plugins, training — or all three. Start with a conversation.
No commitment needed — just a conversation about your team's challenges.
Trusted by
Organisations I've worked with on AI-assisted coding adoption.