Wells Fargo was the third major US bank to embrace bots/assistants to help service customers and invited me to their AI team as the second PM. I was tasked with consumer facing interactions, but that was destined to change.

I led a 9+ person multidisciplinary team to develop solutions for the bank’s 26 lines of business, using their homegrown NLU platform. My initial task was to glean lessons from a third party pilot they were running and then scale up the real 0-1 effort on the native platform.

I developed 14 high-level use cases that were functionally universal across most business domains. I then began work with my conversational designer on fleshing out dialog for those cases as well as working with a new vendor that was providing live agent software for the call center teams. Building relationships with stakeholders across departments and navigating enterprise politics was a given.

What came to light was that the WF platform and live agent application were incompatible when bot conversations needed to be elevated from bot to a real human.

My team pivoted to pick up the mountain sized blocker that issue represented and my thrust became routing. I learned the vendor’s routing system and shortcomings, and worked with them to determine necessary steps for the two platforms to communicate. I also addressed the vendor’s lack of in-queue user messaging, improving UX, happy path and sad path issues, edge cases, bot conversation transfer to agent’s screen, intent mapping to appropriate call center agent teams and fallbacks for system errors.