I Vibe-Coded Our Own Internal CRM

I vibe-coded our own internal CRM. It took 2 hours and $25 in tokens.

Context: I was debating between a few leading AI-powered CRM platforms. Some were more traditional (with AI "bolted-on"), some more AI-native, but they all came with a monthly price tag.

As someone who's never been in sales or GTM, I didn't want bells and whistles. I needed something simple to start out with, that could scale with me as I learned. Something that was flexible and simple, but also had AI features to make my life easier.

I often heard advice that for founder-led sales, you should probably be in an Excel sheet for a while, then upgrade when/if you need it. I totally agree - but there were a few helpful things like AI & automations that can't be easily added to a sheet.

So instead I had Claude build a basic CRM that connected directly to our customer data, our DB, and included AI-powered ICP prospecting and intelligent follow-ups.

Just as simple as an Excel sheet, but with some super helpful add ons. It's more customized to me and my needs than any off-the-shelf, general CRM. And if there's a feature missing or something that would make the way I work easier, I could just add it myself with a single prompt.

Build vs Buy

So, this leads to the bigger conversation around the platform shift going on and the debate between build vs buy. When AI can create a piece of software overnight with just few dollars in tokens, why spend hundreds of dollars per month on a SaaS product when you can build something that gets the job done yourself?

I definitely don't think AI will replace all software products and think there are many reasons why SaaS isn't "dead". Especially in domains where outputs aren't easily verifiable, inner-workings are extremely hard to replicate, and reliability guarantees are a must-have.

However, there is a lot of software that absolutely can be built in a night with some help with Claude, iterative prompting, and accepting that it might not be as "polished" as something you could buy. But it's the 80/20 rule.

If you can validate your "vibe-coded" software's correctness by just looking at the output, then the inner-workings don't matter. The code might be ugly and have a few bugs, but it works and was built and shaped exactly for you.

When software can be easily created and molded to your unique workflow, you move faster, learn quicker, and produce more. You're no longer constrained to the platform's capabilities and required to learn new behaviors to make the platform work for you.

It will be interesting to see how this plays out. I think there are a specific set of domains where personalized, AI-generated software will destroy existing SaaS, and another set where SaaS gets significantly better and defensible due to AI.

My gut tells me CRMs might be on the chopping block, but maybe I'm wrong. However my own CRM is working pretty well for me so far!