UX chapter at ING: from nothing to 15 designers
A team of UX designers for business banking at ING Bank Śląski, built from nothing to 15 people across 6–9 squads in 3 tribes, all hired by me, and led for 5.5 years.
Eryk Orłowski
I am a design leader for digital products. I build design teams, design systems and tools in which AI works within limits set by mechanisms.
I have been designing digital products since 2003 and leading design teams since 2006. At ING Bank Śląski I built a UX chapter for business banking from scratch, up to 15 designers across 6–9 squads. At Sportradar I led the team designing tools for traders who correct the automation that sets live odds. At Efigence I lead a design team and bring AI tools into its work. I build tools with AI agents myself: Kelter, Gromatik and Exocortex. I have had a budget in every leadership role. When nobody knows the answer, not even the client or the stakeholder I would ask, I still come with a recommended answer, because that is my role.
Selected work
A design system for an energy group with many brands
A design system for an energy group with many brands. A centre of excellence made up mostly of designers from my team develops it; I give strategic support and a toolchain that measures brand consistency.
Efigence: AI in the everyday work of a design team
Bringing AI tools into the everyday work of a design team at an agency serving banks, together with rules on what can be handed to a model and what stays with a person.
Gromatik: from tools for a client to a product of my own
A platform for maintaining design systems, built from tools for a client with many brands, in which AI generates component code within limits enforced by hard rules.
Kelter: an AI agent engine that withstands model mistakes
A runtime for AI agents in which the model suggests the next step while a predictable program runs the work, so the system is resilient to model errors.
Lines of work
People and organisation
Building and leading teams, hiring, growing people, ways of working.
Systems and design systems
Design systems, many brands, publishing processes, accessibility.
AI with governance
Tools and operating models in which AI works within limits set by mechanisms.
- A portfolio engine that only states checked factsOct 2026 – in progress
- Exocortex — a knowledge system2026 – in progress
- Exocortex R&D: a lab where evidence decides about AISep 2026 – in progress
- Gatefold: an AI design process where a person makes every decisionOct 2026 – in progress
- Cognitive Mirror: an AI that says what it does not knowJul 2026 – Oct 2026
- Halvern: meeting notes that never leave your machineAug 2026 – Sep 2026
- Kelter: an AI agent engine that withstands model mistakes
- Talks on AI and strategyundated
Early signals
Solutions that came before the market caught up.
Career
- Efigence S.A., Nov 2025 – present
- TalentCanvas, Oct 2025 – present
- Sportradar, Oct 2022 – Jun 2025
- ING Bank Śląski, May 2017 – Oct 2022
- Deloitte Digital CE, Jun 2015 – Jun 2016
- offline.pl, 2003 – 2017
- UseLab, 2011
- K2 Internet, Oct 2007 – Jun 2009
- MRM Worldwide, Jul 2006 – Oct 2007
- argonauts, Dec 2003 – Oct 2005
- Wprost Online, Sep 2000 – Jul 2003
- Nov 2025 – present
Efigence S.A.
Product Design Manager
People and organisation
- Oct 2025 – present
TalentCanvas
Co-founder
People and organisation
- Oct 2022 – Jun 2025
Sportradar
Odds Product Design Manager
People and organisation
- May 2017 – Oct 2022
ING Bank Śląski
UX Chapter Lead, UX Principal Expert
People and organisation
- Jun 2015 – Jun 2016
Deloitte Digital CE
UX Manager
People and organisation
- 2003 – 2017
offline.pl
Independent UX consultant
Systems and design systems
- 2011
UseLab
UX Director
People and organisation
- Oct 2007 – Jun 2009
K2 Internet
Managing Director and co-owner
People and organisation
- Jul 2006 – Oct 2007
MRM Worldwide
UX Director
People and organisation
- Dec 2003 – Oct 2005
argonauts
Project manager and usability lab
Early signals
- Sep 2000 – Jul 2003
Wprost Online
Website manager and online advertising specialist
Systems and design systems
Latest experiments
Notes from work in progress: what I tried, how I measured it and what came out.
Fixing my model router took 90 seconds. Paying for it would take a hundred years
Choosing the right AI model for each task can be automated, but not out of the box: my router was confidently wrong until I measured it and retrained it on my own tasks. It now gets 85% right, and it's switched off, because the hardware it needs would cost more than it saves.
The boredom was in the brief
My site works, it reads well, and nobody will remember it. I asked AI for something bold and got a poster from 1973, then three different styles on one skeleton. The problem was what I had given it.
Agent or operator: twelve unlabelled variants and a rule written in advance
Same brief, same six reference images, two ways to make the screens: an AI agent on its own, or me at the design tool. I scored twelve unlabelled variants against a rule written in advance.
Contact
- [email protected]
- linkedin.com/in/eofek