Build it. Break it. Make it trustworthy.

The first rule of Data Agent Fight Club? You DO talk about Data Agents, especially when they’re wrong.

Building a Data Agent that can answer questions is easy. Building one that can answer the right questions, against the right data, with answers you can trust is where the fight begins.

In this hands-on workshop, we’ll turn the usual AI demo on its head. Instead of showing only what works, we’ll deliberately build, break, challenge and defend our Data Agents.

Across a series of “fights”, attendees will tackle the problems that emerge when Data Agents meet real-world enterprise data: ambiguous business language, competing definitions, complex relationships, different data grains, misleading but plausible answers, generated versus deterministic queries, permissions, security and changing models.

The objective isn’t to build an agent that answers everything. It’s to build an agent that knows what it knows, knows when it needs clarification, uses the right source and can produce answers you’re prepared to defend.