Here’s a simple test for whether you’re overpaying for software. Open the main menu of your CRM or marketing automation platform and count the buttons. Then count how many of them you actually click in a normal month.
For us the score was roughly three out of twenty. We were paying for the whole cake and eating one slice. And every year the price of the cake went up, while every new feature was just another decorative layer on top. Nothing for the core.
So at the change of the year we made the decision: we would replace our CRM, sales funnels, and marketing automation with a tool we build ourselves. Three months later the old platform was switched off.
The actual build took a week.
What we paid for and what we used
The 10% we actually used was the obvious stuff: CRM, sales funnels, marketing automation in a few formats, and the analysis and reporting that comes with them.
The 90% we never touched? SMS messaging, WhatsApp messaging, cart abandonment features built for e-commerce, and more. I won’t name the platform, but go look at the feature list of any major marketing automation suite and you’ll get a pretty good idea of what we weren’t using.
And here’s the thing nobody says out loud: every bought B2B CRM is really built for SaaS companies. That’s fine, there’s a lot of SaaS in the world. But if your business isn’t a SaaS sales motion, you end up bending your processes to fit the tool instead of the tool fitting you. That’s was the constant struggle of a stack that was bought, not built. Nobody ever really likes using it.
Yes, it was vibe coding – supervised by experts
Let me get this out of the way, because it’s the part people raise eyebrows at. The build was done with AI. What made it work was not the AI. It was that we had people involved who consult on these systems, who have developer experience, and who understand cybersecurity. People who know what to expect from the other end, what the result is supposed to look like, and what to look for when checking it.
That’s the difference between vibe coding and cowboy coding. We could quality assure and evaluate what was being built at every step. And critically, we knew what to ask for. The AI cannot know what you’ll need in the future. You have to know it yourself and ask for it.
Same with GDPR. We’ve done consulting work on it, so we knew how to keep the risk vectors low and build mitigations in. There’s never zero risk, but you can build a strong setup, and you should know what one looks like before you start.
We built iteratively rather than writing one giant spec and hitting go. Some people prefer the full-spec approach, and the iterative way sometimes costs a bit more in tokens, but you always know what’s happening in the build, and you can steer. For a system your business runs on, I’ll take that trade every time.
The biggest surprise? It was much easier than expected. My one moment of doubt was before the build started. It faded fast, because the thing simply worked.
What the tool actually does
It does what a CRM, sales funnels, and marketing automation do. Nothing exotic. The difference is that it’s light, without the bloat, and it’s shaped around how we work instead of the other way around.
One principle guided the design: a tool is only useful if it’s being used. So we made it as interactive as possible. It reaches out to our people when they need to know something. The tool works for the people, not the people for the tool.
The extras we didn’t plan
Then something interesting happened. Once the customer management side was in good shape, the natural question was: what’s the next stage?
That’s how lead acquisition and lead intelligence started forming. They were never in the original scope. Our thinking on leads is simple: the key is finding the right companies, and the right people in them. Not spamming everyone with more volume. AI should be used for better aim, not more shots.
What it costs
Hard to give an exact number, but here’s the honest comparison: the entire stack we used to build and run this costs less than what we previously paid for just the CRM, sales funnels, and marketing automation. And we use that stack for plenty of other things too.
So depending on how you want to allocate the cost, our CRM now costs somewhere between zero euros and still-less-than-before. I can live with either math.
Who should not do this
Now the honest caveats, because this is not for everyone.
Don’t build enterprise systems this way. Keep it small enough. Developers and architects will tell you the same: past a certain scale, vibe coding will not suffice, especially if you don’t know what you’re doing. At that point the SaaS route is the safer option.
And don’t build critical systems if you don’t know what the end result should look like. If you can’t tell a good system from a bad one when you see it, you’ll either be disappointed (low criticality) or in trouble (high criticality). You need to know what you’re doing regardless of who or what writes the code. Because you are responsible for it either way.
“But that’s developer territory”
If you’re a marketing leader thinking this is developer territory, I’d ask how you arrived at that. Because it has code in it? By that logic, anything with pictures in it is marketing territory, and there are pictures everywhere.
This is a collaboration. You’ll build it as a team.
And about the developers: from what I’ve seen, their problem is rarely the idea of using AI for coding. Their problem is the hype at the other extreme, the claim that everything can be built from thin air with one sentence and it’ll be perfect immediately. So developers do what politicians do and take the completely opposite stand: nothing can be done. The truth, as usual, has people spread around the middle.
Talk it through and meet in the middle. You will need those developer guys and girls for quality assurance and architecture, at minimum as consultants. And they’ll need you to know what the business actually requires.
That part, knowing what you want and recognizing it when you see it, no one can outsource. Not to a platform, and not to an AI.
Also read
Age of AI Is Age of Marketing Tech Transferability
Self-Hosting Your Marketing Tools: A Strategic Decision
One More Thing to Maintain: The Technical Debt Hiding in Your Marketing
Three Jobs To Clean Up Your Marketing Stack