Target Digital Network Analyst Challenge
An AI-assisted workspace that helps digital network analysts cut through overwhelming volumes of data — and stay in control of every conclusion.
Replace every striped block with a real screen or photo.
A tool that helps analysts think — not one that thinks for them.
The TDNA challenge posed a deceptively simple question: how might we help analysts make sense of more data than any person can reasonably read? I designed a concept for an AI-assisted workspace that surfaces patterns, explains its reasoning, and always keeps the analyst in the driver’s seat.
My approach started where I always start — with the people doing the work. Before touching an interface, I wanted to understand what analysts actually struggle with, what they trust, and where automation could help without getting in the way.
Digital network analysts work under real pressure: enormous datasets, fragmented tooling, and decisions they need to be able to defend. The volume isn’t just large — it’s faster than any human can read, and the cost of missing something that matters is high.
How might we let AI carry the weight of the volume, while the analyst keeps the judgment and control?
Analysts are pattern-seekers under time pressure. They’re skeptical by training — and rightly so. Every conclusion they reach may need to be explained and defended, so a tool that hands them an answer without a reason is a tool they won’t use.
I combined secondary research on analyst workflows with task analysis to map where the real friction lived. A few things kept surfacing — and they reshaped how I thought about the whole problem.
- 01 Volume isn’t the problem — relevance is. Analysts don’t want less data; they want the right data surfaced first.
- 02 Trust is traceable. An AI suggestion is only useful if the analyst can see why it was made.
- 03 Context-switching is expensive. Every tool an analyst jumps to costs them their train of thought.
These insights became the rules I designed against. Whenever a decision was unclear, I returned to them.
I mapped the analyst’s flow, then sketched broadly before committing to a direction — testing how much automation felt helpful versus how much felt like losing control.
Three moves that let AI assist without taking over.
A prioritized signal feed
AI ranks what likely matters most and pushes it to the top — but the analyst can re-weight, filter, and reorder at any time. The machine sorts; the human steers.
Every answer shows its “why”
Each AI suggestion opens to reveal the signals and sources behind it. Nothing is a black box — so the analyst can trust it, question it, or defend it later.
One unified workspace
Search, review, and annotation live side by side, so analysts keep their train of thought instead of losing it to tab-switching.
The hardest question wasn’t visual — it was about trust. How much should the AI do before it starts to feel like it’s deciding for the analyst?
I chose to show reasoning and sources over a single confidence score — a number is easy to over-trust, evidence invites scrutiny.
Analysts need depth, but not all at once. I layered detail so the surface stays calm and the depth is one click away.
Anything the AI does automatically, the analyst can see and reverse — control is never taken silently.
The concept landed on a principle I keep coming back to: good AI tools make people feel more capable, not more replaceable. Keeping the analyst in control wasn’t a constraint — it was the whole point.
If I took this further, I’d test the explainability patterns with real analysts and pressure-test where trust breaks down — the part I’m most curious about.
Mostly, this project reminded me why I love research: the answer was never in the data volume. It was in listening to the people who have to live with the tool.