Salesforce: Generative AI

I was the lead strategist and design manager for Sales Cloud AI after ChatGPT launched in 2022. I created a collaboration with my design strategist peers to create a framework for designing Generative AI across Salesforce products. Our efforts led to a Salesforce re-org and the creation of a team dedicated to the development of Einstein Copilot.

My Role (2023)

I was the design manager and lead strategist for the Generative AI work stream. I designed demos, ran weekly design crits, mentored design thinking, and collaborated on AI strategy across Salesforce teams.

SErvices

UX/UI Design, Design Mentorship, Workshop Facilitation

Salesforce Announces Einstein GPT

Salesforce announced Einstein GPT in March 2023. Teams across Salesforce were asked to integrate the service into their products in one engineering release. However it was a brand new technology and there were no existing standards or best practices yet.

Balancing Innovation with Cooperation

Product and engineering teams were in launch mode. There was a lot of pressure to ship fast and compete internally. However, without standards or best practices, the risk was high that the UX would be poor and disjointed.

My goal was to make the case that cooperating at a higher level would accelerate our ability to:

Reduce time to market by reducing one-off solutions
Design higher quality, innovative solutions with a shared understanding of the technology’s capabilities
Reduce technical debt by designing a good foundation for future experiences

Workshop and Cross-Cloud Collaboration

I organized a kick-off workshop and then a weekly group of interdisciplinary leads to collaborate on creating a shared framework. We did a deep-dive UX audit, invited internal subject matter experts to consult, compared engineering roadmaps, and started prioritizing questions that needed to be answered.

Confusion about Generative AI + Disjointed UX

UX audit showed disjointed UX that mixed and matched conversational turn-taking experiences with prompt generation.

Modes of engagement

In the UX audit, we saw 6 underlying modes of engagement with with Generative AI. We created a framework considering some key criteria for each mode:

What kind of data is involved, is this data was already structured or unstructured?
What are the potential risks? Creating user trust is like walking up stairs, slow to go up, fast to go down
What are design principles or “best practices” for each one?
How could these modes of engagement flow together?

Screen Shot 2024-09-25 at 3.17.15 PM.png

Screen Shot 2024-09-25 at 3.18.24 PM.png

Screen Shot 2024-09-25 at 3.21.07 PM.png

Screen Shot 2024-09-25 at 3.21.27 PM.png

Screen Shot 2024-09-25 at 3.21.40 PM.png

Screen Shot 2024-09-25 at 3.23.07 PM.png

Designing for AI Responsibly

We also worked closely with the VP of Ethics at Salesforce to develop a framework for risk mitigation.

Making Recommendations to our CDO

We presented to our Chief Design Officer and highlighted questions that needed to be answered ASAP based on team roadmaps. Our two major questions were:

What is the consistent entry-point for Einstein GPT in Salesforce?
What kind of “generative” experience is it? Is it always a a conversational turn-taking experience?

Einstein Copilot Beta UX

After presenting to our CDO, she advocated for creating a consistent, dedicated sidecar experience that was conversational. A re-org created a new team dedicated to designing the Einstein Copilot UX instead of each team designing for Einstein GPT on their own. This demo shows the aligned sidecar UX for Einstein Copilot across Sales, Marketing, Service, Commerce Clouds, Slack and Mobile.