The problem
Standing up a branded avatar was a heavy, multi-team effort — enough friction that a third of enterprise prospects dropped off before ever launching one.
Design challenge
How might we empower both novice and expert enterprise users to configure, preview, and launch a fully branded avatar themselves?
Defining success
Before designing, I aligned the team on measurable targets — so we'd know whether self-serve actually worked.
| Metric | Q1 baseline | Target post-launch |
|---|---|---|
| Time to first avatar | 14–18 business days | ≤ 3 days |
| Deployments per month | ~4 | > 10 |
| Drop-off rate | 30% | < 10% |
Aligning the team first
I ran a kickoff to restate the problem and baseline, map assumptions, define KPIs and targets, set a decision-making RACI, and agree on next steps — so design decisions downstream had a shared foundation.
Research
I partnered with a mix of enterprise customers across industries to understand how avatars actually got made today.
What we heard
Two axes shaped the whole design. Users split evenly between exploring and arriving with a fixed brand vision — and most were AI-aware but not AI-fluent, so the tool had to guide without condescending.
"Excited to try it out if it makes me more efficient — but it should guide me."
Prioritizing with Kano
I mapped 10+ candidate features on a Kano model, then sequenced the build around what would actually drive adoption — must-haves first, delighters where they'd differentiate.
What testing taught us
These fed a second iteration — a guided path for explorers, surfaced customization for specs-driven users, and a more vibrant, unified style guide.