“All this happened, more or less.” K. Vonnegut
Something funny.
10 Jan 2026, Warsaw, Poland
Clear sky, 19 Celsius
newspaper
tabloid
trendy
UX generalist on AI products
What do I work on everyday
Starting conversation
Disclaimer: For the last 2 years I’ve worked as a designer for a suite of AI powered products. The project presented below is a concept inspired by real-word enterprise tool I work on.
My role
Lead designer, working closely with a team of developers and Product Owner
Responsibilities
User Research, UX Design, UI Design
Target users
All employees of the company
Timing
June 2024 - now
Environment
  • Quickly changing technology makes it hard to plan work ahead cause we don’t know what will be launched and how good it will be
  • Users expect the product to work 1:1 as the AI tool use privately (chatGPT, Gemini, Claude, Copilot)
  • Users assume that AI works like magic, as shown in promotional videos, and that new features should be available as soon as they hit the market
  • Users' AI literacy is evolving rapidly, influencing their expectations and satisfaction with our product.
Ongoing research
  • 40+ in-depth user interviews
  • Usability testing via Maze, with over 150 participants
  • Annual large-scale user surveys (thousands of respondents) to track UX metrics such as CSAT, NPS, and accessibility, and identify key improvement areas
Problems & Solutions
1.
Users perceived the interface as visually outdated, difficult to read, and unresponsive across different screen sizes.
Solution:
The product was fully redesigned to be responsive and modern, establishing a strong foundation for future development through the introduction of a Material UI component library.
2.
Words and features that were obvious to AI‑savvy users were confusing for many employees, creating a high learning curve for average users.
Solution:
User flows were adjusted to better match typical user needs by:
  • simplifying language,
  • hiding advanced or less frequently used features from the main experience,
  • adding contextual explanations in the UI,
  • sending educational communication via email and Teams.
3.
As the product grew, chat history became overwhelming and difficult to navigate.
Solution:
New navigation patterns were introduced, including:
  • keyword search across chat titles, messages, and tools,
  • filters, and sorting,
  • ability to pin favorite conversations.
4.
Depending on their role and environment, users wanted to interact with the product in different ways (e.g., on the go, in short sessions, or during long focused work).
Solution:
New interaction modes were introduced:
  • dictation,
  • voice to voice interaction,
  • mobile support.
5.
Many AI tools already existed within the company, but users were often unaware of them or did not know how to find the right one.
Solution:
A centralized catalog of AI tools (internal and external) was created, with search and filtering capabilities to improve discoverability.
6.
There was a growing need for users to build, customize, and share their own AI-powered tools.
Solution:
Users can create AI tools through defined processes, while admins manage approvals, publishing, and the full tools catalog in one central place, replacing manual email workflows.
7.
Users struggled to use AI efficiently, and many new features went unnoticed as the product evolved.
Solution:
Educational support was added both inside and outside the UI:
  • guidance on model selection,
  • feature tours and announcements,
  • company-wide training sessions.
  • reducing reliance on AI expertise at critical points in the process through automatic selections and universal default options.
Results
  • Executed a comprehensive product redesign and delivered over 100 new features
  • Achieved an 85% Customer Satisfaction (CSAT) score (from over 3,000 responses)
  • Boosted user productivity: 72% of users save over 1 hour weekly, 22% save over 4 hours weekly
  • Reduced tool publication processeses from 1 week to 1 day