How AI agents fit into workflows

"I might trust an agent to check my bank balance, but not to move $200. We need finer-grained ways to talk about permissions... The best way to think about an agent is as an elite human teammate, highly capable but needing the right tools and guardrails." — Adam Seligman

Business processes rarely live in one app — they span tools, teams, and departments. Workato has spent the last decade making those processes easier to orchestrate, and now bringing AI agents into the mix without losing reliability. In this episode, Ravi Madabhushi talks with Adam Seligman, CTO at Workato, about the future of workflow automation in the agentic era.

You’ll hear:

• Why speed is the new macro force driving business automation.
• The pattern of marrying deterministic workflows with non-deterministic LLMs.
• Lessons learned from $150M of R&D on agentic workflows with Fortune 500 customers.
• Why robust evals, guardrails, and observability are non-negotiable for production agents.
• How to treat agents as superhuman teammates with scoped permissions and KPIs.
• The emerging need for finer-grained access controls and contextual policies for AI agents.

The episode closes with a reminder to developers: focus on business context first, not frameworks — happy users matter more than clever demos.