How American Airlines Puts AI Agents in Production Every 90 Days

Every enterprise is caught between twin pressures.

Leadership wants agents in customers' hands. Not next year. Now. The CEO is banging the table. The timeline is weeks and quarters. Yet every successful enterprise has a system of record holding the business together - and it wasn't built for agents. Not for the security problems they create, not for the quality problems, not for any of this.

American Airlines is no exception. Their CEO wants AI in production every 90 days. Yet, at the core of AA is SABRE - built with IBM over 60 years ago. It wasn't built with AI in mind. It wasn't even built in this century.

That's why I'm excited for this conversation with Pragadeesh Prakasam, architect at AA for the past 13 years. Years before anyone was talking about agents, Praga started building a "context layer" across customer, loyalty, reservations, check-in, and commerce APIs. When the AI moment arrived, it was no coincidence leadership tapped him to architect AA's commercial GenAI program on top of Praga's context layer.

And as always, we'll get into his personal champion's story, why he does this work, and his approach to servant leadership. Plus how he sees the travel industry coming full circle - from human travel agents to AI agents that finally bring personalized travel advice back at scale.

What You'll Hear

• What happens when one architect's context layer becomes the AI strategy - before the CEO has to ask

• What changes about security when the "bots" you've spent your career fighting are suddenly the ones you're building for

• "You're not going to become an AI engineer by just building MCP servers" - what most teams are getting wrong about the AI stack