Don't rebuild your CRM. Build the 10% it can't do.
AI has people talking about building their own CRM from scratch. In practice it rarely pays off. The real opportunity is getting everything out of your core system, then building only the piece it can't do.
The short version: rebuilding a core system like your CRM from scratch almost never pays off, even with AI making software cheaper to build. Small organizations don't save enough to justify it, and larger ones take on security and compliance risk that dwarfs the license cost. The better move is to get everything you can out of the system you have, then purpose-build only the piece that genuinely can't live inside it.
The noise about building your own
There's a lot of talk right now about building your own CRM, or any core system, from scratch. It usually gets framed as saving on licensing, with AI as the thing that finally makes it cheap enough to try.
In theory, sure. In practice, I've never once seen it pay off.
If you're small
It's the last thing you should touch. A free CRM already does the job, and your time, and your AI credits, belong in your actual product.
Saving a few hundred dollars a month sounds great until the "quick two-hour build" turns into days of real work to make something secure that a spreadsheet wasn't already handling. Then you own it forever: the bugs, the updates, the security.
If you're at scale
The money looks more real, mid five or six figures a year. But now security is the whole game.
Your internal systems may need to be SOC 2 compliant to satisfy your own customer contracts. That license fee is a rounding error next to a breach and the lawsuit behind it. The math quietly flips back.
The move: trust the core, build the 10%
So the move is almost always the same: trust the core system, get everything you can out of it, and purpose-build only the piece that genuinely can't live inside it.
That last part is where AI truly changes the math. The adjacent tool that used to take forever to build, cost a fortune, or mean a six-figure add-on is now something you can actually build.
Think of it as 90/10. Ninety percent is what your core system already does well, or could, if it were set up right. Ten percent is what it genuinely can't do. That ten percent is where building pays off.
Finding the line is the real skill
The hard part is knowing where that line actually is.
First, push the system to its limits, so you're getting every bit of value it can actually give. That's usually further than people expect, and it's the honest way to find the true edge of the platform. Most organizations are using a fraction of what they already pay for.
Then, and only then, build what lives beyond it. Sometimes that's an AI feature that couldn't have existed before. Sometimes it's software that finally got cheap enough to be worth it.
What the 10% looks like
The pieces worth building are usually adjacent to the core system, not replacements for it:
- A tool built around one workflow your CRM handles awkwardly.
- A database that brings your systems together for reporting and AI, without cluttering the CRM. More in A data warehouse isn't a big-company project anymore.
- An AI feature that works on your own data and fits how your team operates.
- A product your customers use that connects to the core system, rather than replacing it.
If you only do one thing
Before you build anything, list what you wish your core system did. Then check, honestly, how much of that list it can already do with better setup. What's left is your 10%.