Alex Bauer is the co-founder responsible for product at Upside, where he builds an AI-native go-to-market data layer for sales and marketing teams. His central concern is that agents cannot answer consequential business questions reliably when customer records conflict, company terminology is undefined, and generated conclusions cannot be traced to evidence.
From classical piano to enterprise data
Bauer earned a bachelor’s degree in piano performance from Mannes College and a master’s degree from the University of Kansas. After working as a product manager at Flyreel, he joined Branch in 2016 as a developer advocate and moved through developer relations, product marketing, and market strategy before becoming head of product.
At Branch, he tackled the measurement problems created when customer journeys span disconnected devices, applications, and marketing channels. His writing explored identity matching in mobile deep links and privacy-conscious mobile attribution, establishing a practical interest in whether business data accurately reflects customer behavior.
In 2024, Bauer co-founded Upside with Mada Seghete, Jonas Bauer, and Dan Ahmadi. The team initially developed Pipedash, a B2B attribution product, but found that reliable attribution first required reconciling activity across sales, marketing, partnerships, and customer-success systems. That foundation evolved into infrastructure for AI agents, internal applications, and business analysis. In an essay coauthored with Seghete and Ahmadi, Bauer makes the case that AI-enabled revenue operations depend on governed data and shared business definitions.
How he makes business agents more trustworthy
Organizational context determines whether answers are correct. A question about first-quarter pipeline requires knowing the company’s actual fiscal calendar, which sales stages qualify, and how employees define pipeline. Bauer treats those conventions as essential infrastructure for accurate analysis.
Librarian-agent architecture supplies company-specific knowledge. A specialized agent retrieves business definitions, product documentation, organizational knowledge, and earlier failed-query context before another agent answers. Citations make the resulting claims inspectable.
A jury-and-judge workflow makes ambiguous attribution accountable. Independent agents assess the evidence behind a deal, while a coordinating agent evaluates their reasoning and synthesizes their conclusions. If disagreement remains substantial, the system expands the jury or escalates the decision instead of manufacturing certainty.
Commander’s intent for AI agents improves execution. Bauer gives agents the purpose behind an assignment, alongside relevant context and oversight, while reserving consequential work for systems capable of planning, coordinating subagents, and using external tools. His AI Engineer World’s Fair talk illustrates these patterns through actual revenue workflows.
Bauer also sees AI-assisted development as a way for marketers and salespeople to build tools around problems they understand directly. His comparison of enterprise AI experimentation to pulling a slot-machine handle captures his impatience with workflows that substitute repeated guesses for reliable context, evidence, and human judgment.
