Skip to content
Christy Yao
Actian · Enterprise AI

AI that guides data teams from signal to decision

I designed a context-aware AI system that helps data engineers know what to do next during an investigation, turning a complex workflow into guided decision-making.

Product
Actian Data Observability
My role
UX Designer
Timeline
Jun–Aug 2025 · 12 weeks
Team
JEKRJF
+1
TL;DR

AI existed. It just didn't help when it mattered.

Actian is an enterprise data platform serving 10K+ enterprise customers and 42M users, including 24 Fortune 100 companies. Its Observability Platform helps data teams detect and resolve issues across complex systems. During the platform's 2025 global launch, over twelve weeks, I defined how AI should show up here: not a feature bolted on, but part of the workflow.

The problem

Guidance gap

Engineers had every signal. What they lacked was knowing what to do next.

The solution

Context-aware AI

AI surfaces inline, scoped to what someone is already investigating.

What I delivered

AI system

A reusable AI component library and a foundation for future AI features.

Key outcome

~25%

An early team estimate. Clearer guidance got new engineers to a confident first action sooner.

1 / 3

Problem

Users had data. They didn't know what to do next.

During a live incident, engineers open dashboards, check logs, compare signals. The data is there. After each step they get stuck: what do I look at next, what actually matters, am I even on the right track?

The whole project lives in the red box. Detection was solved. Guidance was not.

For the user

More time guessing. Less confidence in each decision. Every manual step adds pressure and slows the resolution.

For the product

Slower resolution means unreliable data reaching clients. For a platform built on trust, that is the worst outcome.

What makes AI actually useful.

I reviewed twelve AI tools, studying how people start with AI, how systems respond, and what builds or breaks trust.

The platforms that felt most useful were not the most capable. They were the ones where AI responded to what people were already doing.

The shift wasn't a new surface. It was reorganizing the one that existed around the decision.

Old question

How do we make AI easier to find?

Real question

How do we help people know what to do next, in the moment they need it?

Key insight. Access wasn't the issue. Knowing what to do next was.

NDA · Protected work

The rest of this case study is protected. 6 chapters are under NDA.

What's above covers the problem and the reframe. What's protected covers the design work: personas, explorations, and what shipped.

The password is on the resume you received from me.
Don't have it? Email me and I'll reply the same day.

Inside

  1. 01
    UsersThe primary persona: who investigates, what they reach for, and the moment they get stuck.
  2. 02
    ExplorationThree AI participation models, and why two of them broke the workflow.
  3. 03
    The solutionThe three-stage flow, walked screen by screen.
  4. 04
    FeaturesThe feature set, each mapped to a moment of uncertainty.
  5. 05
    Delivery & impactThe component library, handoff artifacts, and what the numbers held up.
  6. 06
    ReflectionWhat I would defend, and what I would do differently.