Jacek Janczura Founder-CTO · Enterprise AI
Book a call

Consulting · Engagement model

Engagements with a defined outcome and an end date.

I take on a finite piece of work with an agreed outcome: an architecture and risk read, an interim leadership stretch, one production AI workflow, or an engineering organisation that needs to work with AI safely. Every engagement ends in a handover to the people who will run it.

Send a project brief Book a 30-min call

Finite, outcome-defined, handed over

We agree what has to be true at the end and how we will know it is true. I work inside your repository, your review process and your stakeholder cadence rather than beside them, and decisions get written down as they are made. When the outcome is met, ownership sits with your team: standards, documentation and a system they can change without me.

You inherited a mess

Technology turnaround

Legacy debt off the critical path, the compliance gaps auditors flag closed, and a dated plan for the first 90 days — without freezing the roadmap.

See the turnaround angle You have a demo, not a system

Production AI systems

Document-heavy workflows and customer-facing assistants taken from prototype to something that survives real data, real integrations and an audit.

See the systems angle

What I take on

Four engagements

Scoped one at a time, each with an outcome we agree before it starts.

  1. Architecture & risk review

    A structured read of the system you have, or the one you are about to build: boundaries, data flows, failure modes, security and compliance exposure, cost, and the trade-off behind every recommendation. Findings come back prioritised into now, next and later.

    In practice The same read I ran on our own AWS platform against the Well-Architected Framework, and on client systems before they went into regulated production.

  2. Interim CTO / AI lead

    Hands-on leadership while you hire or recover: architecture decisions, delivery standards, roadmap, technical budget and the stakeholder conversations — with the decisions written down and the mandate ending in a handover.

    In practice Three years as CTO of a bootstrapped enterprise AI company: engineering from 0 to 10, 50+ organisations on the platform, ISO 27001 and SOC 2 Type 2 in about three months.

    Read the case

  3. Production AI thin slice

    One workflow, end to end, in your environment: process and retrieval design, the evaluation harness, human-review thresholds, the integration, and a deployment path your team owns. Narrow on purpose, so it can actually ship.

    In practice A member-facing pension assistant: prototype in a day from platform components, on-premises MVP in twelve sprints under Dutch financial regulation.

    Read the case

  4. AI transformation & engineering enablement

    Introducing teams to AI-accelerated software engineering: setting up the best practices, teaching people to build their own internal tools, and setting up environments where AI can be used securely. Workshops on prompt engineering and AI IDEs, MCP practices and custom skills, and guidelines for responsible use.

    In practice Delivered for enterprise engineering teams at a global automotive group: onboarding to Cursor, Claude Code and Copilot, security standards for enterprise AI tool adoption, and MCP best practices in their internal instances.

    Read the case

Where I help, and where I don't

A good fit

  • AI has to land inside a real back-office process — documents, contracts, reports, applications — and the output has to be defensible to whoever signs it off.
  • A customer-facing assistant has to work in a regulated environment, on-premises or under data-sovereignty rules.
  • A prototype impressed everyone and now has to survive real data, real integrations and an audit.
  • The platform is slowing the team down: an entangled monolith, no quality gates, reliability and cost drifting.
  • A certification or a pentest finding is coming and the roadmap cannot stop.
  • Engineers need to use AI tools without leaking data or losing review discipline.

Not a fit

  • A seat to fill on a long-running team with no defined outcome.
  • Research without a workflow behind it, or a model trained for its own sake.
  • Work where nobody on your side will own the system after the handover.
  • A number in a pitch deck that I cannot evidence — I would rather say no than dress it up.
  • Deployments where evaluation, security and observability are scheduled for phase two.

Send a project brief

The goal, where you are now, the constraints you cannot move (security, compliance, timeline) and who will own it afterwards. That is enough for a straight answer about whether I can help.

Send a project brief Book a 30-min call

Remote-first, on-site when it matters; NDA on request