A firsthand experiment in why CRM is becoming infrastructure.
The field of AI-native CRM is moving fast. Most new products promise productivity gains. Fewer actually change how you think about what CRM is.
I’ve been spending time with several of these tools, testing them in real work: managing deals, running parallel sales motions, and keeping up with customers while context constantly shifts. Most tools feel additive. A few feel structural.
I did not come to that conclusion because of a feature checklist. I came to it because of how the product behaves when you stop using it as a CRM.
At the surface level, Day AI does what you would expect: it records calls, captures emails, integrates deeply with Slack, surfaces deal context before meetings, and nudges you to close loops you forgot to close. I used it exactly that way while running several live opportunities. It worked well enough that I kept it on.
Then I tried something it was clearly not designed to sell me.
After a few weeks of accumulated emails and call recordings, instead of asking for deal status or next steps, I asked the agent whether it could help me think through a propensity model: what kind of companies should I pursue next, and why.
The result was not impressive in a classic data-science sense. The model was basic. It relied only on the data it already had. It was not something I would deploy as-is.
But it changed how I think about the stack.
The important realization was this: the call recorder, the pipeline view, the reminders — those are not the product. They are data capture mechanisms. Once you accept that, CRM stops being the thing you buy and starts being the place where high-value business interaction data accumulates.
In the past, this data mainly created lock-in. You stayed because moving was painful. That logic no longer holds. With a cognitive layer on top, the value shifts from storage to interpretation. Suddenly, I can explore models, questions, and analyses that were never scoped into the original CRM purchase.
For example, I can look at how deals progress not just by stage changes, but by how conversations evolve. I can inspect my own language patterns versus buyer responses. That kind of analysis used to require separate tools, integrations, and manual effort. Here, it emerged naturally from the same data I was already generating by doing my job.
This is where I believe the revenue stack is heading, even if most vendors will resist saying it out loud. The stack is collapsing downward. At the base, it becomes a database of interactions. On top of it sit agents that continuously reinterpret that data to create new value — often value you did not know you would want when you bought the system.
Day AI did not invent propensity modeling. What they did — intentionally or not — is expose how easily CRM turns into infrastructure once you put a cognitive layer on top of real business communication.
That shift is not theoretical. I felt it while using the product.
monday.com lost roughly 70% of its value in nine months. The debate that followed was predictable: AI will replace SaaS camp vs. the no it won’t, evRead more...
A firsthand experiment in why CRM is becoming infrastructure.The field of AI-native CRM is moving fast. Most new products promise productivity gains. Read more...
Jon MillerI’ve been following Jon Miller for years.In 2006, he co-founded Adobe Marketo, a company that revolutionized the industry with its marketRead more...
What roles does creativity have in a world increasingly dependent on the use of AI technology, especially in content creation?
Elad Ben David, HeaRead more...
CRM Isn’t the Revenue System. It’s the Data Layer
March 16, 2026
Kfir Pravda
A firsthand experiment in why CRM is becoming infrastructure.
The field of AI-native CRM is moving fast. Most new products promise productivity gains. Fewer actually change how you think about what CRM is.
I’ve been spending time with several of these tools, testing them in real work: managing deals, running parallel sales motions, and keeping up with customers while context constantly shifts. Most tools feel additive. A few feel structural.
One that stood out for me is Day AI
I did not come to that conclusion because of a feature checklist. I came to it because of how the product behaves when you stop using it as a CRM.
At the surface level, Day AI does what you would expect: it records calls, captures emails, integrates deeply with Slack, surfaces deal context before meetings, and nudges you to close loops you forgot to close. I used it exactly that way while running several live opportunities. It worked well enough that I kept it on.
Then I tried something it was clearly not designed to sell me.
After a few weeks of accumulated emails and call recordings, instead of asking for deal status or next steps, I asked the agent whether it could help me think through a propensity model: what kind of companies should I pursue next, and why.
The result was not impressive in a classic data-science sense. The model was basic. It relied only on the data it already had. It was not something I would deploy as-is.
But it changed how I think about the stack.
The important realization was this: the call recorder, the pipeline view, the reminders — those are not the product. They are data capture mechanisms. Once you accept that, CRM stops being the thing you buy and starts being the place where high-value business interaction data accumulates.
In the past, this data mainly created lock-in. You stayed because moving was painful. That logic no longer holds. With a cognitive layer on top, the value shifts from storage to interpretation. Suddenly, I can explore models, questions, and analyses that were never scoped into the original CRM purchase.
For example, I can look at how deals progress not just by stage changes, but by how conversations evolve. I can inspect my own language patterns versus buyer responses. That kind of analysis used to require separate tools, integrations, and manual effort. Here, it emerged naturally from the same data I was already generating by doing my job.
This is where I believe the revenue stack is heading, even if most vendors will resist saying it out loud. The stack is collapsing downward. At the base, it becomes a database of interactions. On top of it sit agents that continuously reinterpret that data to create new value — often value you did not know you would want when you bought the system.
Day AI did not invent propensity modeling. What they did — intentionally or not — is expose how easily CRM turns into infrastructure once you put a cognitive layer on top of real business communication.
That shift is not theoretical. I felt it while using the product.
And once you see it, it is hard to unsee.
CRM Isn’t the Revenue System. It’s the Data Layer Beneath It. was originally published in RevenueFlows on Medium, where people are continuing the conversation by highlighting and responding to this story.