When Salesforce announced Einstein GPT, dozens of teams raced to integrate generative AI with no shared standards. I organized the cross-cloud collaboration that built the framework, presented to the CDO, and influenced the reorg that created Einstein Copilot.
After ChatGPT launched, every Salesforce cloud was told to integrate Einstein GPT within a single engineering release. With no playbook, each team invented its own patterns independently — producing a fragmented experience mixing conversational interfaces, prompt-based generation, and inline suggestions, sometimes on the same screen.
Beyond bad UX, this meant platform-scale technical debt, an inconsistent product story, and a missed window to define Salesforce's enterprise AI point of view. The org structure incentivized teams to compete, not collaborate. No one was solving the coordination problem.
I reached out to design strategist peers across clouds and pitched a simple case: cooperating would let everyone ship faster at higher quality. I organized a kick-off workshop and established a recurring weekly working group of design, product, and engineering leads across multiple clouds — not a committee, but a working session that made decisions.
Our UX audit revealed teams weren't just building inconsistent UI — they were conflating fundamentally different types of AI interaction. I led the group in defining six distinct modes of engagement with generative AI, each with its own data profile, trust model, risk implications, and design principles. We partnered with Salesforce's VP of Ethics to embed responsible AI as a structural consideration, not an afterthought.
This gave the organization a shared vocabulary. Instead of debating UI, teams could identify which mode they were designing for and apply the right patterns.
The framework surfaced two questions no individual team could answer: What is the consistent entry point for Einstein GPT across Salesforce? And is it fundamentally a conversational experience? I presented these findings — backed by weeks of cross-cloud analysis — directly to the Chief Design Officer.
The CDO advocated for a unified approach, and Salesforce reorganized to create a dedicated team for Einstein Copilot — rather than each cloud building its own version.
Einstein Copilot beta launched with a consistent sidecar UX across Sales, Service, Marketing, Commerce, Slack, and Mobile.
The six modes of engagement framework directly informed the new team's design direction, giving a 70,000-person company a shared language for AI experiences.