Rebuilding Trust at Platform Scale
How a design team in free fall became the org Salesforce trusted to design its biggest AI bets.
At the end of 2021 I was leading design for enterprise Commerce Cloud when I was asked to take over a team that was struggling: the design org for Experience Cloud, the platform Salesforce customers use to build portals and storefronts. Three designers were left from a team of seven. The ones who stayed were strong and completely burned out.
The real problem wasn't headcount or skill. It was trust. Partners had stopped trusting design's judgment, and design had stopped trusting partners.
Two years later the same org was 26 designers across two continents, and it was the team Salesforce handed its most important AI work to. This is how that happened.
Design had a reputation as a black box. Priorities were opaque, commitments slipped, and PMs had started filing tickets instead of having conversations.
Inside the team it looked like design theatre. People were producing polished demos instead of outcomes, afraid to show early work, and avoiding hard conversatons. It was the bottom of the "5 Dysfunctions of a Team" pyramid: no trust, so no constructive conflict, so nothing honest got said.
"Some of us felt more like a service team than strategic partners. We were trying our best to support everyone, but we didn't feel empowered to influence priorities upstream."
— Karen, DesignerI spent my first few weeks listening. I talked to product leadership about their priorities and where trust had broken, and to engineering about the delivery relationship. I brought my team together to evaluate what wasn't working, and what was.
The pattern was clear:
Partners wanted a strategic collaborator and were getting a service desk.
Designers wanted protection, and a way to move upstream and shape direction.
Nobody had a shared picture of what design was working on or why.
The team had almost no fluency in the product or the market it was designing for.
Safety first.
Before any process, I made it safe to be honest. A simple example: Katie, the most junior designer, was presenting a visual builder component to Khush, our detail-oriented SVP of Product. I watched her get backed into a corner, not understanding what he was actually asking for. So I stepped in, repeated his point back, and asked him to help us understand the underlying goal. Once Katie understood it, she explained the friction it would create, and I backed her. Khush saw it immediately and we found a compromise in the room. Afterward Nicole, the most senior designer, told me she'd never seen anyone navigate Khush that way, curious and humble, and defuse something that used to leave the team feeling bruised.
Nicole told me she'd never seen anyone talk to Khush that way — curious and humble — and defuse something that used to leave the team feeling bullied.
Then transparency.
I replaced the black box with an operating model partners could see into: a real intake process, a capacity tracker maintained in partnership, a clear Definition of Done that restored and ensured quality.
Inviting PMs into the tracker was a harder sell than it sounds, because visibility cuts both ways. It meant they had to prioritize against a fixed capacity instead of filing tickets and hoping.
The first real test was a stalled CMS overhaul. The PM team was under pressure to build a sweeping vision one designer had shaped, but when I dug in, it was still too abstract to hand to engineering. Rather than ship it as-is or scrap it, I worked with the team to break it into pieces we could unblock sequentially, then used that plan with partners to buy the runway to get new hires onto the hard parts.
We shipped something the team was proud of, and it started to rebuild trust across teams. Alongside it I stood up a designer-led research program and trained the team on design sprints, so they were shaping direction instead of reacting to finished PRDs.
Leaders and a second site.
As we grew past ten, I promoted and coached two managers, Nicole and Chelsea, into trusted partners for PM and Eng leadership. I also negotiated with senior leadership to build a new design site in Bangalore, led by a Senior Director, to match the growing partner investment there and alleviate burnout from time zone coordination.
As team health and quality climbed, so did executive trust. Which earned us increased scope and impact.
We were entrusted with the Salesforce mobile platform and Lightning UI, the front-end components for the entire Salesforce platform. Additionally, teams for Lightning App Builder (the visual Salesforce screen editor), IdeaExchange (customer request portal), and Salesforce Wallet (consumption-based pricing tools) joined my org. Then the AI work started arriving, because we had become the org leadership trusted to get it right.
App Store rating moved from 4.1 to 4.7, with engagement up 30%.
We defined the conversational UI patterns for the assistant used across the full product suite, and built Prompt Builder so non-technical users could configure prompts safely, no code required.
The obvious framing for an AI development platform was better tooling. I bet the discovery on a different problem: intent, not tools, reframing the work around what a developer is trying to build rather than the steps to build it. That was a real risk, since it meant setting aside the approach everyone expected. I took it to the Partner Advisory Board for feedback, and it earned direct sponsorship from Marc Benioff, and it launched at Dreamforce '25.
With expanded scope came challenges. We had become a junk drawer: teams bolted on with no shared context, and coherence suffered.
The case for fixing it came from a few directions at once. Our annual planning sessions made it clear that AI needed a focused, cohesive design effort rather than something bolted onto existing teams. Separately, my managers were leading multiple teams that had little in common in terms of users and product strategy. And team retros plus trends from the company's biannual survey were both surfacing the same complaint, that nobody had context outside their own corner of the org.
Three separate signals pointing the same way was enough to act on.
The fix was a re-org, and initially it wasn't popular. I got the team through it by being clear about the problem it solved and staying close while we adapted. I gave one leader with a strong knack for innovation focused ownership over AI instead of a sprawling remit, gave two others new, consolidated charters that felt more unified, and stood up cross-departmental workgroups, team liaisons, and a shared working model so the org could support programs that ran across team lines. It let us ship things we couldn't have otherwise.
designers across the US and a Bangalore site built from scratch, with zero regrettable attrition through the entire transformation.
developed from within, so the org could scale without me in the middle of everything.
the team Salesforce trusted with its top-priority AI programs, from Agentforce patterns to Agentforce Vibes.
design went from bottleneck to the partner other teams asked to be moved under.
The rebuild was never about one lever. It was about designing the conditions where good work becomes possible: safety so people tell the truth, transparency so partners can trust you, and enough leadership depth to hold it together as it grows.
The hard part wasn't the operating model. It was governance, the unglamorous work of maintaining trust every week so it didn't quietly decay.
Scope stops being something you lobby for. It starts arriving on its own.
