Hi everyone,

3 days. 1 booth. And the same question from almost every leader who stopped by: “We have agents now. How do we trust them?” I spent last week with Boomi at Gartner IT Symposium/Xpo in Orlando. Boomi World was just the start. Orlando was where we got to see it in action.

Here’s what happened, and what I’m taking home.

Booth #221 was home base

The Ravit Show was live from the Boomi booth all week. Cameras rolling, hallway conversations, and a lot of honest talk about what it really takes to run AI agents inside a business.

The questions were different from last year. Fewer people asked what agents can do. More people asked how to control them.

A live build on the show floor

The moment I keep thinking about: Michael Bachman building an agentic integration live, in front of the room.

He took requirements from a North American 3PL provider and, in about 5 minutes, built an EDI 204 load tender integration with error handling, test data, and docs in 3 languages. Claude Code and Boomi Companion did the building. The Boomi platform ran it.

A few ideas from the session stuck with me:

• Agents and agentic systems are different things. Agents are probabilistic. Traditional software is deterministic. Agentic systems combine both. Michael’s way of putting it: a little silver (agents) and a lot of gold (automation) go a long way.

• Skills and context matter more than the model. Models will keep improving on their own. What sets you apart is the domain context, tools and security around them.

• AI authors the integration. The platform runs it. Generative AI stays out of the runtime path, so every component stays visible, governed and auditable.

For that 3PL use case, the goal is partner onboarding going from 90 to 120 days down to 3 to 4 weeks. That’s the kind of number that gets a CIO’s attention.

On The Ravit Show

Right after the session, Michael and I sat down to talk about why so few enterprises fully trust their agents. We covered the #1 mistake companies make when agents get access to systems of record, what’s at stake for the CISO when agents aren’t visible or attributable, and how CIOs should think about bringing their own model.

I also sat down with Erica Reuter from Snowflake. We talked about what joint Boomi and Snowflake customers ask for as they move agents into production, and what happens if the governance layer breaks down while agents are querying governed data.

Both interviews are live now on The Ravit Show.

3 things I’m taking home from Orlando

1. Building an agent is the easy part. Proving what it touched and who it acted for is the hard part.

2. Skills and context matter more than which model you pick.

3. Governance has to sit between the agent and the data from day 1.

If you’re working through any of this right now, hit reply and tell me where you’re stuck. I read every one.

Thank you to the Boomi team for having me in Orlando. See you at the next one.

Ravit Jain