Industry

Nonprofit / Child Safety Tech

Serivces

AI-Enabled Research + Design Ops

Turning a feature factory into a feedback loop

90%

more user research feedback in 9 month period

~33%

faster feature development and testing

90%

more user research feedback in 9 month period

~33%

faster feature development and testing

90%

more user research feedback in 9 month period

~33%

faster feature development and testing

PROBLEM

Shipping fast, listening slow

The old design process is fading. The one thing that won’t fade is knowing what people need and how they feel. Our agile process shipped features without research to inform them, and our tools were scattered.

Challenges

Features proposed without user research.

Fragmented tooling slowed design and delivery.

Design was a bottleneck rather than something the whole org could do.

Challenges

Features proposed without user research.

Fragmented tooling slowed design and delivery.

Design was a bottleneck rather than something the whole org could do.

Challenges

Features proposed without user research.

Fragmented tooling slowed design and delivery.

Design was a bottleneck rather than something the whole org could do.

a black and white photo of wavy lines
APPROACH

Agents do the legwork, people make the calls

We rebuilt research so it runs all the time instead of once in a while. Then we built a spec-driven workflow so there was always something real to test.

Solutions

Agents propose PRDs every two weeks, then product prioritizes them based on user need

A listening agent scans Slack and support channels and creates monthly themes.

Drop a PRD into the spec-driven workflow and get a prototype or production code built on our design system.

Synthetic users automatically test any prototype created by the system.

Solutions

Agents propose PRDs every two weeks, then product prioritizes them based on user need

A listening agent scans Slack and support channels and creates monthly themes.

Drop a PRD into the spec-driven workflow and get a prototype or production code built on our design system.

Synthetic users automatically test any prototype created by the system.

Solutions

Agents propose PRDs every two weeks, then product prioritizes them based on user need

A listening agent scans Slack and support channels and creates monthly themes.

Drop a PRD into the spec-driven workflow and get a prototype or production code built on our design system.

Synthetic users automatically test any prototype created by the system.

pink green and blue abstract painting
a black and white photo of a pattern

THE PIVOT

Drift happens. Ownership fixes it.

We connected legacy systems through agents, then spent a couple months tuning the workflow to stop drift and find where a human needed to step in. We also named a decision-maker at every step, for quality, privacy, and impact.

THE PIVOT

Drift happens. Ownership fixes it.

We connected legacy systems through agents, then spent a couple months tuning the workflow to stop drift and find where a human needed to step in. We also named a decision-maker at every step, for quality, privacy, and impact.

THE PIVOT

Drift happens. Ownership fixes it.

We connected legacy systems through agents, then spent a couple months tuning the workflow to stop drift and find where a human needed to step in. We also named a decision-maker at every step, for quality, privacy, and impact.

OUTCOME

Design for everyone, research by default

Most UX support requests now go through the spec-driven workflow, and that work dropped from a full sprint to half a sprint. Research and design move 30% faster, and features reach testing about a third sooner. The process got faster, and users stayed at the center of it.


  • 90% more research collected in a 9 month period

  • ~33% faster feature development and testing


a black background with orange and blue lines