I fix AI adoption, train developers, and build the rest.

A typical engagement runs in three phases. First I embed with your team — observing real workflows, current tool usage, and where the friction actually sits. Then I design and configure the setup: agent and tooling configuration, prompts and playbooks, and the integrations that make the workflow practical for your stack. Finally, I work alongside your developers in pair sessions, code reviews, and iteration cycles until the new practices stick. Scope is agreed per outcome, not per hour.

We agree on baseline metrics before the engagement starts — typically PR cycle time, time-to-first-commit on new tasks, share of pull requests touched by AI tools, and developer satisfaction. After the workflows are embedded, we re-measure. The goal is a defendable shift you can show your CFO, not a vibes-based 'it feels faster'. If the numbers do not move, we have a problem to solve together rather than a result to celebrate.

Training is educational: workshops, hands-on sessions, and multi-day programmes that teach your developers how to use AI coding tools effectively. It is the right fit when the gap is primarily knowledge. Consulting goes further: I diagnose why adoption is not sticking, design the workflow and tooling setup, and embed it with your team over several weeks. Often the answer is both — first set up the system, then train the wider organisation on how to use it well.

It is almost never the technology. The recurring patterns are tools rolled out without configuration for the actual codebase, no clear guidance on when to use AI versus write by hand, no shared prompts or playbooks, and no follow-through after the initial training session. My psychology background combined with hands-on engineering means I work on both layers — the human one and the technical one — and diagnose the real bottleneck before suggesting solutions.

That is the whole point. I leave behind documented playbooks, configured tooling (CLAUDE.md, agent configurations, prompt libraries), and at least one internal champion who can extend the setup. The success criterion is not 'Luka is still in the building' — it is that productivity gains compound after I am gone. Top-up training and ad-hoc consulting are available when something genuinely new comes up, but they are optional, not built-in dependencies.

Off-the-shelf tools are excellent for general tasks. Custom AI matters when you need deep integration with your stack, use of proprietary data, or enforcement of complex business rules. The systems I build automate multi-step workflows — syncing CRM data, running inventory checks, routing approvals — which a chatbot or simple API wrapper cannot do. The decision usually comes down to whether the workflow is your competitive edge or just a productivity helper.