Microsoft Research · Cambridge, UK · 2023

Making organizational knowledge visible

An interactive data visualization that helps knowledge workers see how their organization fits together, with Copilot built in.

When you join a large organization, knowledge is everywhere, but it is hard to know where to start or how it all fits together. Over four months I researched this problem, tested two navigation patterns with users and designed an interactive data visualization prototype that went to two partner teams.

Personal account, not endorsed by Microsoft.

Under NDA

This page shows no designs or internal data. All illustrations are abstract and drawn for this page. It is a personal account and is not endorsed by Microsoft.

Overview

My role

Sole designer: research, user studies, interactive data visualization

Team

Worked with partner teams in design, research and machine learning

Duration

Jun – Sep 2023 (4 months)

Type

Research concept, no launch

Problem statement

In a large organization, knowledge is spread across many apps: documents, chats, meetings, people. Each piece is easy to find on its own, but putting the pieces together into a big picture is hard.

The timing raised the stakes. In summer 2023 the first AI assistants for work tools, like Microsoft 365 Copilot, had just been announced, and there were no established patterns for how AI should show up in everyday work. The goal was to find where AI adds value, not just where it fits.

Problem statement
“How can we surface knowledge so that people get the big picture as well as specific information?”

Designing in a research lab

In a research lab, design works ahead of the product. Its job is to make ideas tangible so people can talk about them. I brought together study findings, model capabilities and open product questions, and turned them into prototypes that gave two partner teams concepts, scenarios and questions to build on.

Inputs become a prototype, the prototype starts a discussion, the discussion raises new questions.

Iterative process

  1. Empathize

    Mapped how people find organizational knowledge today

    Overwhelming: where do I even start?

  2. Define

    Defined the user scenario and persona

    Persona and “I want to…” needs

  3. Prototype

    How can knowledge workers find and navigate to knowledge? Prototyped two navigation patterns

    Tree and bubble map

  4. Test

    A design critique (9 participants) and 6 usability sessions

    Neither pattern won on its own

  5. Concept

    A high-fidelity, interactive data visualization prototype, built up in three stages

    Canvas, relationships, people and Copilot

Research

Who we designed for

The core scenario was onboarding: a new project manager who needs a broad view of the organization, depth in a few areas and people to ask.

  • I want to navigate to what I need
  • I want to dig in deeper, or take a step back
  • I want to understand how and why things are connected
  • I want to discover what I didn’t know existed
  • I want to have the freedom to change the view

9 Participants

in a remote design critique of the concept

6 Usability sessions

1 hour each, comparing the two navigation prototypes

Researching navigation patterns

Two low-fidelity prototypes tested one thing only: how people get an overview of the organization.

Tree: topics connected, opened step by step
Bubble map: topics nested in each other, explored by zooming

What worked

Each pattern had its strength. In the tree, diving in felt natural and dragging helped people understand how topics connect. The bubble map made it easy to zoom between overview and detail.

What was missing

Neither answered why things are connected or how they relate to me, and both felt static.

The decision. Neither pattern was the answer on its own, so I merged both into an infinite canvas with zoom: the connected structure of the tree, with the zoom between overview and detail from the bubble map. That became the base of the concept, and you can try a simplified version of it further down.

Concept

The result was a high-fidelity, interactive data visualization: an infinite canvas that shows how topics, people and knowledge connect, and that people can explore, reshape and share.

How the concept grows

Stage 1

Navigate

“How do I get the big picture and the details?”

Topics live as cards on a canvas, from the big picture down to the details.

Stage 2

See and understand

“Why are things connected, and how sure is it?”

Relationships become visible, so people understand why things belong together.

Stage 3

Personal and collaborative

“How does this relate to me?”

The canvas adapts to you and becomes a shared space, with Copilot built in.

The role of Copilot

In my concept, Copilot sat on top of the canvas. The idea: an assistant can show complex relationships that are hard to explain in text, answering with an interactive visualization instead of a paragraph.

Designing for uncertain AI

AI suggestions are highlighted differently from confirmed connections, so people know to check them before relying on them, and can hide the ones that don’t fit.

Try the idea

A simplified, abstract version of the canvas to explore yourself: drag to move around, open a topic, zoom to see more or less detail, and switch on AI suggestions. It shows the principles, not the prototype itself.

222

Drag to move around · click a topic to open it

100%
  • Confirmed
  • Less certain
  • Outdated
  • Suggested by AI

Conclusion

Outcome

The interactive data visualization prototype, the open questions and the research findings went to both partner teams as input for their work. As a research concept, there was no launch. The natural next step is to bring it to the product team, turn the most valuable parts into a production-ready experience and get it into people’s hands.

“Lukas presented a remarkably complex yet intuitive data visualization prototype that helped move our team’s thinking forward. […] any team would be fortunate to have him on board.”
Leor Maizel, Senior Product Designer, Microsoft
“Lukas approached design challenges creatively, and his choices were thoughtfully informed by the research at hand […]. I would warmly recommend Lukas as a UX Designer, and believe that he will thrive in projects where he can apply his ability to break down complex problems and translate these into meaningful user experiences.”
Britta Burlin, Principal Design Manager, Microsoft

Learnings

Combine, don’t pick. The strongest idea came from merging what worked in two prototypes.Make AI visible. AI suggestions only help when people can tell them apart from what is confirmed, so the design has to show which is which.A prototype is a question. In a research lab, the best prototype is the one that starts the right conversation.

Microsoft, Microsoft 365 and Copilot are trademarks of the Microsoft group of companies. This page is a personal portfolio account and is not endorsed by Microsoft.

Hero