Let me tell you something against my own interest. I can make our product do almost anything you ask for in a demo. Summarize the call, draft the follow-up, score the deal, answer the rep, write the brief. It’s genuinely impressive. And so is every competitor’s, doing the exact same things. not because we are all brilliant, but because we are all renting the same brain.
Nearly every serious tool in go-to-market now runs on the same small set of foundation models. When you watch a demo “do” something clever, you are mostly watching the model, dressed in our interface. The cleverness isn’t ours. It’s rented, and our competitors rent it from the same place. I think someone should just say that out loud, because almost the entire AI sales motion right now depends on you not noticing it.
Capability became a commodity, and nobody updated the demo
There was a time when “it can do that” was a real differentiator when building a capability was hard and having it meant something. That time is over. Capability is now a commodity input, like electricity. A building with power isn’t special; every building has power. A tool that “has AI” isn’t special; the AI is the grid, and everyone’s plugged in.
When every tool can do everything, “it can do that too” stops being an answer. It’s the one thing you can now assume about all of them.
But the demo hasn’t caught up to this, because the demo’s whole job is to make a commodity look like a differentiator. “Watch our AI draft this email” is theater - a magician showing you a trick every magician in the room can also do, while the spotlight makes it feel singular. I’m not above it; we could run that demo too, and the temptation is real. But you should know what you’re watching when you watch it. You’re watching the grid, not the building.
So what is actually different?
If the brain is rented and shared, the difference has to live somewhere the model doesn’t reach. It lives in two places, and they are the only two worth evaluating.
The first is guardrails: what the system was built to do well, and refuses to do badly. A general model will cheerfully do anything, including the wrong thing, confidently. The craft is in the constraint: the taste, the domain rules, the things a product was shaped to be excellent at and the things it was deliberately stopped from improvising. Two tools on the same model can behave completely differently, because one was given judgment and the other was given a chat box. The constraint is the product. The model is just the raw material.
The second is architecture: what the system does when you aren’t watching, and what it connects to. Does it wait to be asked, or does it act? Does it live in its own silo, or reach into the tools you already run? The model gives you a brain. Architecture is the body the nervous system that decides whether the intelligence does anything when no one is typing. Same brain, radically different bodies. That’s where the difference you’re paying for actually is.
The only question that survives
So stop asking the question every vendor wants you to ask, which is “can it do X.” The answer is yes. The answer is always yes now, for everyone, because the grid does X. Asking it tells you nothing, and the demo built around it is theater designed to run out your evaluation clock on a commodity.
Ask instead: what was this built, with real guardrails, to do excellently - and what does it do when I’m not in the room? Parity can’t fake the answer to that.
The vendor who responds with another flashy everything-demo is selling you the commodity and hoping you mistake it for the difference. The one who shows you a narrow, deep, specific thing, a hard problem handled with obvious judgment, work happening on your data while no one watches, is showing you the part the model didn’t give them. That’s the body, not the brain. In a market where everyone rented the same brain, the body is the only thing you’re actually choosing between. Trust it over the trick.
There’s a deeper reason this matters the shared brain isn’t just shared, it’s probabilistic, which quietly breaks the way enterprises have trusted software for forty years. That’s a larger argument than this one, and it’s the subject of its own piece. For here, the practical takeaway holds on its own: capability is the commodity; the body around it is the choice.
