Redesigning Data Operations

Led a redesign of Cognite's data management experience, grounding the work in real user behaviour rather than existing tooling or architecture.

The Challenge

As Cognite Data Fusion scaled across customers, sites, and teams, configuration and change management became increasingly difficult to reason about.

While powerful tooling existed - particularly CLI- and Git-based workflows - users struggled to maintain a clear mental model of:

• What was currently configured
• How resources were interdependent
• What would change if they acted
• Where risk or unintended impact might occur

The challenge was not adding new capability, but redesigning the data operations UI so complex systems could be understood, validated, and evolved with confidence.

What I Did

I led the redesign concept, grounding the work in real user behaviour rather than existing tooling or architecture. I combined:

• Ethnographic research to understand how users reason about configuration, risk, and responsibility
• Usability testing to validate mental models early
• AI-assisted prototyping to explore the problem space at high fidelity and speed

This allowed me to move quickly from insight to working concepts that stakeholders could meaningfully critique.

Design Approach

Ethnography first

Rather than starting from system structure, I focused on how people talked aboutconfiguration work - where they lost overview, what they double-checked, and what made them hesitant to deploy changes.

Across interviews and testing, a consistent need emerged: a visual way to understand interdependencies, environments, and impact - before acting.

AI-native prototyping

I used AI tools to prototype interaction ideas rapidly at near-production fidelity. This wasn't about visual polish, but about accelerating learning and sense-making in a highly complex domain.

Personas at speed

I used AI to rapidly synthesise lightweight personas from Cognite's existing user research repository (Dovetail), internal reports, and historical research - alongside fresh interviews and usability sessions conducted during the project. Drafted in minutes, not months. Grounded in real, accumulated research. Explicitly treated as "good enough" and refined through ongoing validation.

Outcome

"One thing awesome is I feel that you understood what I described - so maybe there is some LLM involved here. Nice."

"It's connecting some of the dots that are a bit scattered in Fusion today - here you get more of one pane of glass to see something."

Good enterprise UX isn't about removing complexity - it's about making it understandable at the moment it matters.