Four Takeaways from Snowflake Summit 2026

I just returned from an incredible Snowflake Summit. We came home with Snowflake's Product Partner of the Year award for Agentic Transformation, which was a good moment for the team. But what stuck with me had nothing to do with the award.

Last year, almost every head of technology I talked to wanted to know the same thing: does this stuff actually work? This year, nobody asked. Not once. They've moved past it. The question they're asking now is a much harder one.

Here are the four things I took away.

Takeaway 1: The question is no longer whether AI works

The new question is where's the value? — and the pressure to answer it is real. The shift is from try it to show me what it's worth.

You could feel it on the keynote stage. Two of our customers spoke this year — Emmanuel Frenehard, Sanofi's Chief Digital Officer, and Patrick Duroseau, Under Armour's Chief Data and AI Officer. Neither of them was up there debating whether AI is real. They were talking about what they're already running. The leaders who are furthest along stopped second-guessing the technology a while ago. They're being measured on outcomes now, and they know it.

That's the good news and the hard news in the same sentence. AI works, which means the grace period is over. Nobody gets credit for a pilot anymore.

Takeaway 2: If you've outsourced before, you already know how to do this

The companies that move fastest aren't the ones with the most exotic AI strategy. They're the ones who recognize the motion.

If you've offshored or outsourced at any point in your career, you've done this before. You take work that doesn't need to live in-house and you move it somewhere better suited to do it. That instinct — what to keep, what to hand off, how to manage the handoff — transfers directly.

The only thing that's changed is where "somewhere" is. You're not standing up a team in another time zone this time. You're handing specific tasks and workflows to AI agents. Same discipline, new destination. The leaders who frame it this way get moving quickly, because they're not waiting to invent a playbook. They already have one.

Takeaway 3: You're not automating people. You're automating tasks.

Here's where a lot of exec teams are about to make an expensive mistake.

Most of them have a consultant in the room telling them they can take 20, 30, 50% out of some function. And I understand why that lands — it's a clean number on a slide. But it leaves out the part that actually matters.

That percentage isn't people. It's tasks.

AI is going to remove an enormous number of tasks from back-office work. It is going to remove far fewer jobs. Those are not the same thing, and treating them as the same thing is how you end up running a reorg when what you needed was a result. The work gets reshaped. The headcount math the consultant promised mostly doesn't show up, because the unit being automated was never a person. It was a task.

Get the unit right and you'll set realistic targets. Get it wrong and you'll be explaining a miss to your board next year.

Takeaway 4: Legacy SaaS isn't just expensive. It's in the way.

If there's one takeaway I'd push every CIO to sit with, it's this one.

When technology leaders talk about ripping out legacy SaaS, the assumed reasons are cost and old architecture. Both are real. Nobody enjoys the renewal conversation, and nobody loves running on a stack designed in 2012.

But that's not the reason anymore.

The reason is that legacy SaaS is in the way. Your data is locked inside it. Your workflows are locked inside it. And you cannot pull AI value out of a system that was engineered, from day one, to keep everything in. The lock-in was the business model. It is now the thing standing between you and the outcomes you're under pressure to deliver.

So "goodbye SaaS" was never really a cost play. It's the prerequisite. You don't get the AI value until you clear the system that's blocking it — until your data and your workflows live somewhere your agents can actually reach them.

Where this leaves us

Put the four together and the picture is pretty clear. The question changed from does it work to where's the value. The fastest way to capture it is a motion most leaders already know. The value itself comes from automating tasks, not eliminating people. And the biggest thing in the way is the legacy software they've been paying to keep their own data hostage.

The companies that win in 2026 won't be the ones with the best agents. Everybody's going to have good agents. They'll be the ones who cleared the path first.

Thanks to everyone who stopped by, and to Emmanuel and Patrick for sharing the real version of this story from the stage. The grace period is over. That's a good thing.