The course in fifteen seconds
I have been designing digital products since 2003 and leading design teams since 2006; I build teams, design systems and tools in which AI works within limits.
Every fact of the course in one place, each with its source. Nothing here is behind an exercise.
Contact
- E-mail: [email protected]
- LinkedIn: linkedin.com/in/eofek
Course card
- 01I have been designing digital products since 2003 and leading design teams since 2006.
- 02At ING Bank Śląski I built a UX chapter for business banking from scratch, up to 15 designers across 6–9 squads.
- 03At Sportradar I led the team designing tools for traders who correct the automation that sets live odds.
- 04At Efigence I lead a design team and bring AI tools into its work.
- 05I build tools with AI agents myself: Kelter, Gromatik and Exocortex.
- 06I have had a budget in every leadership role.
1.1 · A chapter that outlived its lead
- 07The chapter grew from nothing to 15 designers working across 6–9 squads.
- 08I hired the entire chapter team myself.
- 09I decided which competences a role required; the tribe lead decided on the need and the headcount.
- 10I created a career ladder for designers with about 6 levels, from junior up to Principal Expert UX and Chief Expert UX.
- 11I introduced design reviews and knowledge sharing between squads as a steady rhythm of work.
- 12The measure of the chapter working was that the same problem was not being solved by more than one team.
- 13From the start I wanted the chapter not to depend on one person.
- 14Rituals, standards, career ladders and people in their roles kept working after I left in 2022.
1.2 · Changing how a team works
- 15My team was 2 seniors and 1 junior retraining from QA to UX design.
- 16The designers first worked in a service model central to the tribe, and after a change of concept were embedded in product squads.
- 17The designers moved into product squads because the domain is very demanding and moving them between projects made it hard to get up to speed.
- 18I redefined the designers' work in the team: from mainly UI design to the full UX and UI toolkit.
- 19I extended the team's craft through research with traders, workshops, mentoring, changing the scope of tasks and taking part in discovery.
2.1 · Adoption you can't order
- 20Measured on the real model, the chat gave no facts from outside the package, fell back in 3.3% of answers and took 9.5 s for the slowest 5% of answers.
- 21AI adoption in the team is gradual and voluntary.
- 22We treat the concern that AI produces uncreative work as accurate and show a division of work in which creativity stays with the designer.
- 23The team follows Efigence's official policy on the use of AI in projects.
- 24I took part in launching AI Sagres, the group that leads AI projects at Efigence.
2.2 · Who decides when agents build the tools
- 25The starting point was 7–9 brands from acquisitions, ranging from a PDF brand book to full component libraries in Figma.
- 26I came up with, designed and built the toolchain myself, working with AI agents.
- 27The toolchain is a monorepo of 12 command-line tools with 78 test files, built in June 2026.
- 28A person reviews every drift between Figma and the canon and decides what to do with it.
- 29A measurement on the live stack showed that 5 repetitions of the same prompt gave 5 byte-wise different results.
- 30I designed the "agent in the product" route and suspended it on the same day, before implementation.
- 31Every run records its settings and a fingerprint of its input, so you can trace where a result came from.
3.1 · Training, career ladders, mentoring
- 32I designed and ran commercial UX training for the teams at Oponeo and Onet.
3.2 · Designing for learners who don't share your assumptions
- 33The one I'm proudest of isn't the biggest. I ran the discovery phase for an e-learning platform that ORPEG, the centre for Polish education abroad, was building for its remote schooling programme: lessons for children of Polish citizens who live abroad for a while. My job was to identify and document the requirements, so I spent a lot of time listening to the everyday problems of teaching these children. Some of those stories stayed with me. One child didn't know what a pear was. Another had never heard of Santa Claus, for cultural reasons. And all of them had to work through the Polish national curriculum. That changes how you think about content, language and what you can assume a user already knows. I like doing work that matters to someone, and this one did.
4.1 · One system, many brands
- 34A centre of excellence at the client develops the design system and delivers its architectural decisions in stages.
- 35A new company in the group is one configuration file, with no code changes.
- 36Contrast is checked against WCAG 2.2 for every mode, including mourning, and failing the threshold stops the pipeline.
- 37I co-hosted a webinar on the European Accessibility Act with Michał Jankowski and a lawyer.
The lab
- 38In Kelter the kernel is the runtime and the language model is only a policy.
- 39The same job runs on a 0.5 B local model and on a frontier model after a change of 1 line of configuration.
- 40I came up with, designed and built Kelter myself, with AI agents; it is my intellectual property.
- 41The hypothesis card is frozen and its checksum published before anything is calculated.
- 42As of 30 September 2026 no hypothesis has a result yet.
- 43I was a panellist in a discussion on the agentic economy at ExFest, taking a sceptical position.
- 44At AI Lessons Unlearned I gave the keynote "Who will deliver your 2030 strategy?".