Eduardo Gonzalez on Building AI Agents You Can Rely On
Click on the photo to Watch Full episode on Youtube
What does it actually take to build AI agents you can rely on — not ones that demo well, but ones that hold up when you put them in front of real customers and real operations?
In this episode, Olivia sits down with Eduardo Gonzalez, CEO and co-founder of Xpress.AI, to talk about how he's building an operating system for AI agents that enterprises can actually trust.
Whether you're a CEO, team leader, employee, or entrepreneur, this conversation will change how you think about building reliable AI agents in 2026.
What we cover:
How Eduardo modeled the brain's thalamus to stop hallucinations before they happen
Why SOPs are the single most important skill in the age of AI agents
How Xpress.AI runs more agents than people, and why that ratio matters
Why naming your agent "Bob" gets better results than naming it "DEV-10"
Why agents magnify bad leadership just as much as good leadership
The cultural gap between American "just try it" and Japanese "make it perfect first"
Who owns the agent when an employee leaves, and why that question is about to get real
How "legacy agents" could preserve company knowledge long after someone walks out the door
Xpress Claw: a local, open source version of Xpress.AI you can run on your own machine
Connect with Eduardo Gonzalez
Eduardo is the CEO and co-founder of Xpress.AI, building enterprise agent infrastructure out of Tokyo.
Website: https://xpress.ai Try the open source Xpress Claw for a local agent experience on your own machine, or reach out directly at hello@xpress.ai.
Click on the link below to Watch Full episode on Youtube: