Three moves in under a month made the same bet. Everyone is celebrating the part that was never the hard part.
By Jason Simon, Chief Growth Officer, Solvenna
On May 27, Snowflake bought Natoma. On June 16, Databricks launched CustomerLake. On June 23, Zeta moved its Data Cloud onto Palantir Foundry. Three moves, under a month. Most of the market read them as three product launches. I read them as one bet, placed three times.
The bet is simple. The data platform is swallowing the CDP and the marketing cloud. The vendors want to own the layer underneath all of it. Snowflake is extending its governance from data into the actions agents take. Databricks is putting the CDP inside the lakehouse so the data never gets copied out. Zeta is rebuilding its Data Cloud on the same infrastructure that already runs the rest of the enterprise. Different language. Same move. Own the foundation, and everything above it has to route through you.
They are right about the foundation. The data layer is the moat. I have no argument with the strategy.
My argument is with the timing of the celebration. Buying the platform does not make your data ready to run on it. That is the hard part, and it is the part the vendor cannot sell you, so it never makes the slide.
Here is what that gap looks like in production. This year, 74% of enterprises that put a customer-facing AI agent live pulled it back. Not paused. Shut down or reversed after it reached real customers. Read the next number twice. Among the organizations with the most mature governance, the rollback rate was not lower. It was higher, at 81%. The most prepared teams did not fail less. They saw the failure sooner, because they were the only ones with the visibility to catch it. And the leading trigger was not a bad model. It was customer data exposure. The researchers were blunt about where the problem lived. It lived below the governance layer, in the plumbing.
A separate read from earlier this year put a number on the same wall. Close to 80% of enterprises said their AI was held back by data access, not by the tools. Adoption is running well ahead of readiness. There is a name for that gap now. The AI readiness illusion.
The people who have shipped these migrations said it louder than I did
I recently posted about this. The responses that stayed with me did not come from other consultants. They came from the practitioners who have actually run these deployments, including people who have shipped Foundry work for a living. They were not debating the thesis. They were sharpening it.
The clearest one put it this way. The rollback is almost always upstream of the platform. The same customer shows up with three to five different definitions across CRM, ERP, marketing, and support. No team owns reconciling them. Move that mess onto Foundry or CustomerLake and you surface it faster. You do not solve it. The migrations that landed did the reconciliation work before the platform arrived, not after. That is the whole thing in one line. The platform surfaces the problem faster. It does not fix it.
Another framed it as plumbing, and I keep coming back to it. Teams buy tech to layer on tech and treat the wiring underneath as someone else’s job. It holds until it does not. Then the wiring is the only thing that matters. A third made the point that composable or packaged does not change the math. The readiness problem sits under both. It is the constant.
None of these people work for me. That is what makes them worth listening to. When the practitioners who have the most to gain from a new platform tell you the platform is not the hard part, the pattern is real.
The readiness work comes first
This is the problem Solvenna works on, and it is why QuickLift exists. QuickLift is a 90-day engagement that proves measurable revenue inside the stack a client already owns. No new platform required. It surfaces where customer definitions disagree, inside the current environment, before a migration turns those disagreements into a rollback. We did not build it to argue against the platforms. We built it because the readiness gap kept showing up first, in engagement after engagement.
None of this is an argument against the platform. Most of these are good products, and modernizing the data layer is a real decision. It is an argument for knowing what your data can actually do before you sign for the thing that assumes it is already done.
So, the same question, for anyone in the middle of one of these decisions right now. Before you signed, how ready was your data? Be honest with the answer. It is the one number the vendor deck will never show you.
