As AI assistants like ChatGPT and Claude plug directly into marketing systems, a false claim is spreading with them: that AI like this will soon make the CRM marketing platform unnecessary.
While we run a platform company, this isn’t about preferring the platform. It’s about what each piece can and can’t do, which is easiest to see if you’re precise about who does what.
The AI assistant is a visitor. It shows up, and it knocks.
The connection that lets it in has a name, the Model Context Protocol, or MCP. Think of it as the door.
The marketing platform is the house. It’s where the value lives: the customer data, the decisioning, the years of accumulated knowledge about what works.
Three things, three jobs.
The visitor and the door are new and genuinely useful. But neither is the house, and the house is where the value has always been.
The AI knocks because that’s where the answers are
When someone asks an assistant which customers are about to churn, or which offer to send whom, the assistant does the reasoning. It interprets the question, works out what it needs, and shapes the reply into something useful. What it cannot do is invent the underlying facts.
It reaches through the MCP into the platform, where the customer data, the decisioning, and the record of what works all live. The assistant is genuinely smart, smart enough to know what it does not hold, and to go get it. The intelligence in the answer is the assistant’s reasoning working on the platform’s substance. Neither produces it alone.
This is what the exuberance tends to obscure. When an AI hands you something sharp, it feels like the AI is the intelligence. It isn’t. It knocked on the right door. Take the platform out from behind the connection and the visitor is standing in an empty lot. In an empty lot there is nothing to ask and no one to answer.
“We’ll just build it ourselves” is harder than it looks
Because the connection is so easy, marketing can start to feel like something you assemble yourself. If an AI can plug into anything, why not wire one up to your data and skip the platform? I hear this most from technical teams, who look at an assistant and see a build-versus-buy decision they can win.
They’re not wrong to ask. For some use cases, in-house will get there, and the teams pushing on this are early rather than mistaken. But I’d be honest about how far away “there” still is.
Take building an audience. It looks like a query. It isn’t. It’s a chain of judgment calls: who genuinely belongs and who only looks like they do, which offer fits which person, how to set the goal, how to structure test and control so you actually learn something. Every one of those carries years of practice a general-purpose model hasn’t lived through. It can produce an audience that looks reasonable. It can’t yet produce the right one, and the gap between reasonable and right is measured in revenue.
That will change eventually. But “a model has absorbed enough marketing judgment to be trusted with it unsupervised” is a claim about years from now, not months. The platform isn’t just software you could rebuild from parts. It’s the accumulated discipline of doing this well, so the person asking isn’t starting from a blank prompt and hoping.
The relationship, not the rivalry
The error in the “AI replaces the platform” story is that it casts the assistant and the platform as competitors. They’re not. They’re parts of one system, each doing what the other can’t.
The assistant is built for reasoning and conversation. It meets people where they already work, takes a request in plain language, works out what it needs, and turns what comes back into something useful, with no new interface to learn. The platform is built for a different kind of intelligence. It holds the customer data, the decisioning, and the accumulated record of what works, and it turns a decision into the right message on the right channel at the right moment. The MCP is the connection that lets the reasoning reach the substance.
Remove any one of the three and the value collapses. An assistant with no platform behind it has nothing to answer with. A platform with no way in stays locked behind its own screen. The connection on its own is an empty hallway. What works is the relationship: a visitor that asks well, a door that opens cleanly, and a house worth entering.
This is already happening and not hypothetical
Here are three applications in use right now:
Campaign drafting. A team hands an AI assistant its weekly brief and gets back a full week of campaigns, eight across two brands in a single ninety-minute session. Planning document in, campaign calendar out.
Always-on QA. A scheduled agent checks every morning for journeys, templates, and campaigns expiring in the week ahead, flags what is missing or broken, and opens tasks before anyone arrives.
End-to-end retention. A framework researches retention topics, blends that with the company’s own platform data, recommends approaches, helps build the campaigns, runs the post-campaign analysis, and loops what it learns back in.
In each case the assistant reasons and orchestrates on top of the platform’s data and decisioning. Small teams run like much larger ones, not because AI replaced the platform, but because it put the platform’s intelligence within everyone’s reach, all the time.
The honest part
There’s a real catch, and it’s better to name it than to pretend it away. As assistants handle more of the asking, the platform gets less visible. Someone gets a great answer through their AI and never registers where the intelligence came from. Value goes up; credit can go down.
That’s a genuine shift, but less visible is not less important. The platform is still doing the work, whether or not its name is on the screen when it happens. A utility doesn’t stop mattering because you’ve stopped seeing the wiring. If anything, assistants knocking on that same door, quietly and constantly, is the clearest evidence of where the value actually sits.
The tools change. The intelligence lasts.
Every wave of technology arrives with the same promise: this is the one that finally makes everything before it unnecessary. And every time, the reality is quieter. The way people reach the work changes. The work itself, knowing what to say to which customer, when, and why, stays where it was, and grows more valuable for being easier to reach.
AI knocking on the door is a real advance, and there will be more doors, from more assistants, every year. But a door has always meant there’s a house behind it. The assistant asks. The connection opens. The platform answers. The intelligence is, and always has been, the house.
Pini Yakuel is the founder and CEO of Optimove.
The post AI can knock on the door. Your CRM marketing platform is still the house. appeared first on MarTech.