Pega’s contrarian take on Agentic AI

Two contrarian points of view from PegaWorld last week stuck with me

In his keynote, Alan Trefler, founder and CEO of Pegasystems shared two strong beliefs.

For processes like service—where consistent, predictable execution is paramount and best practices aren’t always reflected in generic datasets—there’s a better alternative to “one-shot” agents.

Trefler argues that you should design the process first—using GenAI on curated best practices and figuring the right AI techniques for each step—and then execute. I think of it as a two-step model.

He also contended that agents can radically simplify data access—eliminating the need to first build a data fabric or unified repository, and instead using data agents to access data wherever it resides.

Lots to ponder