Start with intent
Use business-language templates instead of algorithm-first setup.
Tredence · UX Designer · 2021
Making enterprise AI setup less intimidating for non-technical teams.
Enterprise AI setup that made the first step feel clearer for business analysts and ML teams.
25usability participants
2moderated interviews
4core workflow steps
Overview
People stood at different distances from AI.
I owned research, flows, high-fidelity design, testing, and handoff for an enterprise AI workflow that had to support business users and technical experts without forcing both through the same entry point.
The problem
The blank slate was not neutral. It was intimidating.
- Expert-first entryTechnical choices appeared before users understood the workflow.
- Invisible stateSelected data, algorithms, and solutions disappeared from view.
- Unclear next stepUsers repeatedly asked, “Where do I click?” during testing.
- One level of complexityNovices needed direction while experts still needed control.
The solution
A guided path without a locked corridor.
Keep state visible
Carry selected assets and progress across the workflow.
Reveal depth
Place advanced controls one step away for experts.
Decision 01
Replace the blank slate with a meaningful start.
Business-language templates gave unfamiliar users a clear entry while keeping reusable data, algorithm, and solution assets visible.
- Intent before terminology. Begin with the problem the user recognizes.
- Reusable assets. Make available building blocks easy to scan.
- A visible starting action. Remove ambiguity from the first screen.
Decision 02
Reduce the architecture users had to remember.
The information architecture separated discovery, active work, reusable assets, and administration so people did not have to reconstruct the product map on every visit.
- Clear destinations. Separate finding, building, monitoring, and administration.
- Persistent location. Keep the current workspace and selected state visible.
- Progressive navigation. Expose specialist areas when they become relevant.
Decision 03
Validate direction without inventing precision.
Two moderated interviews exposed confidence gaps. Before-and-after Useberry testing with 25 participants validated the revised direction, but the case does not claim an unsupported improvement percentage.
25 participantsBefore-and-after testing checked whether the revised starting point and visible state were clearer.
- Visible selected state. Keep data, algorithms, and solutions in view.
- Business-language labels. Reduce unnecessary technical translation.
- Expert controls nearby. Preserve power without forcing it into the common path.
Outcome
The revised workflow made the starting point clearer.
The design direction was validated through moderated research and before-and-after testing, then implemented by the development team.
Templates framed AI work in recognizable business language.
Selected assets and progress remained available across steps.
Guidance supported the common path while advanced controls stayed available.
Evidence boundary: two moderated 45-minute interviews and before-and-after Useberry testing with 25 participants support the direction. No unverified improvement percentage or business-impact metric is claimed.
Final product
The enterprise AI workspace.
The supplied final-product cutout is the only visual used here. Supporting diagrams and process imagery have been removed from the case-study presentation.
