Case study 02 2025 · Enterprise property-tech

The project that “couldn't fail”

Ten teams, three agentic products, one company bet — and no one accountable for the experience end to end. A year spent going from doing, to enabling, to scaling.

Role
UX Manager and IC
Timeline
Four quarters, 2025
Scope
10+ product development teams, 3 agentic experiences
Contribution
Research, vision, framework, workshop facilitation, operating model

The setup

Customer problem

Following up with prospects and residents eats the day. The overwhelm of the task shadows the moments of connection that actually matter.

Company bet

Ship three agentic experiences — leasing, maintenance, and resident — that take the mundane communication work off people's plates.

The measure

A single company-level goal: tens of millions of customer actions taken through the AI tools within the year.

Why it was failing

1
New tech, old systems
Existing UX patterns were optimized for manual workflows, and platform dependencies slowed every team down.
2
Siloed workflows
Teams assumed feature ownership rather than shared platform responsibility, so the same problems were solved three different ways.
3
Misalignment
Ownership was split across teams and leaders held partial, local views. No one owned the experience end to end.

The arc

Q1 · Doing

A vision to ground the work

Research first, then a concrete picture of the agent experience.

Q2 · Enabling ICs

A shared language

A framework and playbook so designers had clarity without me in the room.

Q3 · Enabling managers

Alignment on risk

Making the gaps visible to leaders who could act on them.

Q4 · Scaling

A new operating model

Ownership of the whole experience, held by name.

Q1 — Doing

Start where the handoff hurts

As ten-plus teams raced to automate, my research question was narrow on purpose: how do customers know when a task has come back to them?

Automation only helps if the human handoff is legible. If people can't see what needs them, the agent creates work instead of removing it.

Managers

Start from signals, then dig in. They need to know where to look, not what to do next.

Front-line staff

Work from a structured list. They need direction and sequence, not a dashboard.

Both

Notification fatigue. They want one place, with enough context to act.

fig 2.1 — the vision for agents
Q2 — Enabling ICs

Turn a vision into a shared language

A vision only scales if other people can use it. I converted it into an agent experience framework that named what was essential — and what was negotiable.

That grew into a playbook: the new design patterns, plus the processes for testing and validating agent behavior before it reached customers.

fig 2.2 — experience framework
fig 2.3 — pattern playbook
Q3 — Enabling managers

Make the misalignment impossible to miss

Leaders weren't resisting — they each had a partial view. I ran a workshop that put the whole experience on one wall, then converted the discussion into three durable artifacts.

1
Named the specific gap
Surfaced the one workflow where two teams' assumptions collided, and settled it in the room.
2
Defined levels of ownership
Split platform-level from experience-level decisions, and implemented it with both UX groups.
3
Wrote the company-wide vision
One statement of how customers should experience our AI, usable by any team as a test.
fig 2.4 — the workshop
fig 2.5 — levels of ownership
Q4 — Scaling

Change the operating model, not the people

Alignment decays unless it's built into how the org works. Two changes made it stick.

For UX

Designers now own the platform vision for AI experiences, and hold their teams to solutions that serve the whole experience — not just the feature.

For product development

I partnered with product leadership to redraw ownership: one product owner accountable end to end, instead of ownership split by feature.

What changed

Customer actions taken through the AI tools, against the annual company goal.
100s
Deals closed with the agentic experiences named in them.
1
Accountable owner for the end-to-end experience, where there had been none.
Before
After
Ownership by feature
One product owner per bet
Local context
Shared platform context
Leadership misalignment
Clear accountability
Designers waiting for direction
Designers designing end to end, together
What I'd carry forward

The design work was never the bottleneck. The ownership model was.

I'd draw the levels of ownership in the first month rather than the ninth — and I'd bring leaders into one room earlier, because misalignment was never disagreement. It was ten partial views nobody had put side by side.

Want the version with the artifacts?

Emailmelissa.wittmayer@gmail.com All workBack to index ↑