Synopsis: In this Techstrong.ai Leadership Insights interview, SnapLogic CTO Jeremiah Stone explains why legacy systems are emerging as a major barrier to deploying AI agents at scale. He discusses the integration, data access, and architectural challenges enterprises must address to successfully operationalize AI agents across existing IT environments.
Stone’s core point is practical: no one is ripping out decades of systems overnight, because that is where the data lives. Some of it is on premises, some in the cloud, and most of the new AI tooling assumes cloud-first access. That mismatch forces teams to rethink integration, data access, and architecture before they can trust agents to do real work.
He outlines a pattern he is seeing across customers: stop treating everything as “legacy.” Instead, separate systems that should be retired or refactored from “heritage” systems that hold critical business logic and domain expertise worth preserving. The goal is to modernize around them, not bulldoze them for the sake of shiny new architecture.
The conversation also draws a clean line between deterministic workflows that must run the same way every time and probabilistic work where humans already make judgment calls. Stone argues AI delivers faster value when it augments those human, context-heavy processes. He shares SnapLogic’s own example, a system called Sonar that summarizes customer context across internal systems and tools like Salesforce and Zendesk, reducing toil and improving customer conversations.
Finally, Stone pushes back on narratives that knowledge work will disappear overnight. Instead, he sees Jevons paradox at play: as productivity rises, demand for building and governing these systems grows. His advice is straightforward: invest in people, training, and workforce planning now, because the organizations that operationalize agents responsibly will be the ones that turn AI from hype into durable advantage.
