With snow returning to the mountains, I thought it was time to switch from sailing to skiing analogies!
I was MC at ODSC West in San Francisco last week. Between sessions and conversations with vendors and practitioners, four things caught my attention.
1. AI Reasoning Over Databases. Who’s Actually Using It?
There’s a frenzy of activity around AI-powered data analysis—from context graphs to Claude dashboards. Vendors are racing downhill on a black diamond, building technology to make enterprise data accessible to AI.
Meanwhile, enterprise adoption seems to be taking the gentle green run. The gap between what vendors are building and what enterprises are actually using is striking.
2. BigQuery Wants to Be Your ML Platform Too
Google is pushing BigQuery beyond SQL analytics into Python-based machine learning.
The interesting question: Are we watching the boundaries between databases, analytics, and ML platforms disappear?
3. Why Rent AI When You Can Own It?
The enthusiasm for AI sovereignty was hard to miss. Open models are changing the conversation. Enterprises increasingly have a choice between renting intelligence from frontier model providers and running models they control.
The future may not be rent or own. It may be knowing when to do each.
4. We Built CI/CD for Software. Where’s CI/CD for AI Agents?
We spent decades figuring out how to make software reliable in production.
Now we’re deploying agents that reason, make decisions, and take actions—with potentially different outcomes each time.
Testing code is one thing. Testing reasoning is another.
Agent reliability could become the next major engineering discipline.
My Takeaway
We’re seeing a shift from what AI can do to what enterprises can actually trust, operate, and own.
And that may be where the most interesting innovation happens next.