Illustration: three colleagues posting checklists and notes on a wall

Cowork.

Agent-first Adobe CX Enterprise
Scroll to read
cowork

I inherited a project that had been flailing for years.

When I moved back to IC, my VP wanted me to tackle EC Home. One of the most senior-leadership-visible surfaces at Adobe, and quietly underdeveloped while the products around it kept shipping AI.

cowork

I shipped the obvious win first.

I took the bones of what Home was and Spectrum 2-ized it. Ran a quick usability test. The finding was almost too simple to admit: making it look better measurably improved user sentiment. Enough to sell PM and eng on the bigger investment.

EC Home in its current state, before the Spectrum 2 refresh.
EC Home, current state.
EC Home after the Spectrum 2 refresh.
EC Home, Spectrum 2-ized.
3.8 5.1

Felt the page was modern & intelligent 1–7 scale

44% 27%

Negative sentiment

22% 36%

Positive sentiment

Current Home
"It looked poorly laid out and was indeed very confusing."
Updated Home
"...there is an ai chatbot so that seems helpful."
cowork

leadership wanted pillars for summit

Pillars was just creating more routes. We already had AI Assistant on the home page, being used as a feature. I didn't understand why AI wasn't already the front door.

If users still have to navigate, we haven't actually simplified anything.

the reframe

Pillars reinforced our org silos. The work stayed fragmented.

Today, multiple homes.

Each product owns a home. Users have to know which product they want before they get there.

Pillars, routing dressed as consolidation.

Collapsed surfaces, but the entry still routes. A navigation layer between the front door and the destination.

Cowork, the architectural reframe.

One entry. Agents orchestrate across products. Users start with what they want done.

If every user's tool belt is the same regardless of where they land,
why are we creating separate landings at all?

cowork

Three weeks.
A storyboard.
A new strategic direction.

The vision was new but years in the making, drawing on Nextgen AI Assistant and other prior work. Three weeks was the window to assemble and land the vision. I couldn't sell this with comprehensive design at that timeline. I needed a storyboard that got execs reacting to the strategic argument fast. Claude Cowork's release was the moment to argue for the agent layer at the cloud level which would become strategic direction for the cloud.

New paint on a few doors wouldn't have carried that argument.

CX Enterprise storyboard.
cowork

Three of us. One room in SF. Whiteboard, then Cursor.

Vic, Jessi, and me. We whiteboarded the story end-to-end, then built live in Cursor while I kept laying out designs in parallel.

Whiteboard from the SF planning session, the story sketched end-to-end.
cowork

AI as the third teammate.

Rather than telling Cursor each individual spec (type, padding, color tokens), I used Claude to write a skill that encoded the design system. The skill was tuned for the larger typography of the demo stage and the conversational rhythm of the interface. Cursor consumed the system on every build decision, which kept me in design mode while the build kept moving.

AI-native isn't adding AI features. It's restructuring the working method so AI consumes the same design system the rest of the team uses.

cowork

The SVP of Product wanted to white-label Cowork. We said no.

When white-labeling came up mid-sprint, Jessi and I made the call not to do it. Two reasons: we didn't have the time, and a white-labeled version would have diluted the Adobe story we were on stage to tell.

cowork

The framework emerged from pushback on the word "coworker."

Adobe Design was morally opposed to "coworker." The resistance pointed at a real design gap: we hadn't been deterministic about when AI has authority and when humans do. The framework maps four modes to stakes and control, plus one zone deliberately off-limits.

cowork

Summit Main Stage.

Three weeks of Long nights.

cowork

Cowork vision sold at Summit. The validation is happening now.

The week prior to Summit I ran the delegation framework with researchers at a customer onsite. Findings are reshaping how the framework applies in production and how we build a thoughtful framework into our experience going forward.

cowork

Forward-deploying the delegation framework with Marriott.

Summit shipped. Now I'm working with research to fully flesh out the delegation framework, and forward-deploying a prototype with Marriott to learn fast and ship fast.

Cowork · Adobe Summit, main stage · 2025

Q&A

brookebell.ai LinkedIn
Up next Adobe AI