I led early AI platform projects at Salesforce, including Einstein Copilot and Prompt Builder.

Shaping an AI-first strategy

As the design leadership bridge between the AI Cloud and Experience Services organizations, I collaborated with multiple teams to shape Salesforce’s AI strategy and define how users interact with AI across products. Our work centered on three strategic pillars:

1. Assistive for experiences that respond to users in reactive, proactive, or agentic ways.
2. Embedded for integrating AI directly into familiar tools such as Experience Cloud’s builder.
3. Immersive for experiences that enable AI to generate complete interfaces based on user intent. One example is my collaboration with the AI Frontiers team to design Generative Canvas for Lightning, a concept that turns a sales professional’s stated intent into a working interface.

My team of product designers focused on the Assistive pillar by designing Einstein Copilot, Salesforce’s systemwide conversational assistant. We created new components for mobile and desktop, defined the user experience for interacting with the Copilot panel, and worked with Sales Cloud, Service Cloud, Marketing Cloud, and Commerce Cloud to align functionality and address internal use cases. This work became the foundation other clouds built their AI workflows on.

Scaling Copilot across Salesforce

The Experience Services organization at Salesforce is responsible for building products that empower internal teams to create new features and external customers to build applications. During the planning phase for Einstein Copilot, the goal was to design an experience that was extendable, modular, and easy for other clouds to integrate with. My team collaborated with multiple Salesforce clouds to understand their customers and ensure Copilot addressed their specific business needs.

Einstein Copilot was also designed to be easily adopted by both internal and external customers. It launched within the Lightning experience, then expanded into Lightning App Builder, Setup, and Experience Builder, where administrators could add AI to their Salesforce instances directly in context.

In addition to leading high-level experience design, my team of product designers created new base components and interaction patterns specific to Einstein Copilot. I sat between the design system team and Experience Services engineering so the components we built could carry every Salesforce cloud, not just our own.

After months of collaboration across the company, and extensive component and pattern development, Einstein Copilot shipped inside Experience Builder, Setup, and Lightning Builder in 2024.

Reaching that scale forced the team to rethink how Salesforce builds, tests, and launches cross-platform AI experiences.

From surprises to a shared vision

“Dave, I’m surprised by what is built at the end of a release.”

Vice President, Product Management

That sentence, said to me in a one-on-one, changed how I lead and collaborate. It revealed a pattern that had settled into the release cycle: goals set by the SVP were translated into projects by the VP and then handed to Directors and Senior Directors to execute. Teams began designing and building, but as engineering uncovered complexities, priorities shifted and timelines slipped. By the time end of release demos arrived, leadership was often surprised by what was delivered. No one lacked effort or intent. The levels of the organization had lost visibility into one another’s work.

Partnering with PM and Engineering leaders, I introduced a new operating model built to surface scope changes early and keep leadership current on progress throughout the release. Trust between design, product, and engineering grew with it.

Designing a process for clarity and influence

To understand where gaps were forming during each release, I interviewed all PMs and EMs in Experience Services. Three findings came out of those conversations:

• PM and UX leadership were not confident that UX had sufficient influence within the organization.
• New features were often not built to spec and failed to fully address customer problems.
• Leaders were frustrated with the current development process.

These pointed to one hypothesis: the VP needed regular visibility into work in progress throughout the release cycle.

When requirements shifted or engineering identified scope changes, designers iterated quickly, but these updates rarely reached the VP level. As a result, leadership lacked confidence that teams were still solving the right problems.

To fix this, I partnered with my PM and Engineering peers to create a new monthly, design-led review that included the VP and senior directors. The new review process would be:

• Designer led: focused entirely on prototypes rather than documentation.
• Empowering: giving designers direct opportunities to present their work to senior leadership.
• Collaborative: allowing PM leadership to identify misalignment early and redirect efforts as needed.

These reviews gave leaders consistent visibility into progress and put designers, PMs, and engineers in the same room to clarify scope and direction in real time. When disagreement surfaced, it got resolved there instead of at the end-of-release demo. As the design leader, I organized agendas, wrote summaries, and shared recordings across the organization so every team saw the same priorities.

After the next release, I interviewed PM and Engineering stakeholders again. The feedback was positive across the board: executive leaders had clearer insight into plans and progress, and designers were energized by presenting directly to PM leadership. One benefit I had not planned for: PMs began defining requirements earlier in the release cycle to prepare for these reviews, which gave designers two to three extra weeks for collaboration and refinement before development began.

Building a culture of UX excellence

As the leader of the Mobile team, I noticed a growing backlog of P2 and P3 bugs filed by both designers and customers. The backlog looked like a maintenance chore, but it gave me the case I needed to convince PM and Engineering to reserve room in each release for quality work. Together, we established a new process called UXcellence.

My approach was to take a long-term view of quality across the Salesforce Mobile App. I partnered with PM and Engineering to set a goal of resolving a small number of issues in every sprint and tracking progress across the entire year.

The UXcellence initiative turned the Salesforce Mobile App’s growing backlog into a shrinking one. The App Store rating rose from 4.4 to 4.7 stars, and the sales team reported that customers cited the higher rating in purchasing decisions. It also put design in sprint planning and retrospectives, the rooms where the quality bar actually gets set.

Leading this effort reinforced one of my core beliefs as a design leader: excellence comes from consistent, incremental improvement. Built into every release, that improvement gave the organization a product it was proud to ship.

Lessons learned

• Alignment is a design problem. The organization’s biggest gaps sat between levels of the org chart. Treating the release process itself as something to research, prototype, and iterate on changed more outcomes than any single feature the team could have designed.
• Show the work early and often. Alignment decays between check-ins. Sharing work in progress throughout the release, rather than at the end, kept leadership, PM, and engineering solving the same problem the whole way through.
• Visibility compounds. The monthly reviews were built to reduce end-of-release surprises, but their best effect was one I never planned: PMs began defining requirements earlier to be ready for them, which bought designers two to three extra weeks every release.
• Quality is a schedule. A handful of bugs fixed in every sprint, sustained for a year, moved the App Store rating from 4.4 to 4.7. No heroic quality push would have gotten there; the cadence was the strategy.
• Design for the teams you will never meet. Copilot reached Setup, App Builder, and Experience Builder because my team treated extensibility as a design requirement from the first sketch. On a platform, the component library and its patterns are the product.

Results

Einstein Copilot became the foundation of Salesforce’s agentic AI platform. In the Spring ‘25 release, Salesforce formally renamed Einstein Copilot to Agentforce, and the assistant my team designed lives on as Agentforce Assistant. The figures below trace where the product line traveled after my tenure.

• $1.5 billion in Agentforce annual recurring revenue by mid-2026, up more than 240% year over year (Q2 FY27 earnings).
• More than half of Agentforce bookings in early 2026 came from existing customers expanding their usage (Q1 FY27 earnings).
• 29,000+ Agentforce deals closed in the sixteen months after launch (Q4 FY26 earnings).
• 2 million conversations handled by the assistant on Salesforce’s own help site, resolving roughly 70 to 75% of visitor issues (Salesforce).
• Reddit reported 46% of support cases deflected and first response times falling from 8.9 minutes to 1.4 (Dreamforce 2025).