Why I decided to build a boring business
What it is
Recouped is a reconciliation as a service business. It checks that what a company was owed is what actually arrived, documents every gap line by line, and hands them the claim. This is the story of how the most boring idea I had became the thing I’m betting on.
Why
For the last two years I’ve run an AI automation business, spinning up random side projects along the way, hoping I would run into a problem enough times that I could productize it. That moment unfortunately hasn’t come yet, but I’ve really wanted to launch a more focused product or service. So I turned to what I spent the majority of my career doing: running recs.
What settled any questions I had was a YC video about service businesses in the age of AI. Services that used to need a team can now run lean.
The build
I spent a month or two building until I was comfortable with it. The concept is simple: make sure this side ties out with that side. Everything between is the work, and making sense of why things don’t line up is the cherry on top. Getting the platform to handle the different variations data can come in was a challenge, and so was making sure whatever is left over makes sense and is easy to raise. Right now I’m working through the reporting layer, making sure it tells the right story to the customer.
A major part of this build, and honestly any build these days, is deciding how to implement AI. What I landed on came straight from the automation work I do. If you look at an n8n workflow, 90% of the build doesn’t use AI. One or two nodes do, and those nodes matter. Recouped follows the same principle. The rec itself is deterministic end to end, because there’s no reason to have AI do math problems when formulas do them better anyway. AI steps in where it earns its place: pairing off whatever the formulas couldn’t match, and offering best guesses on why a gap happened. Nothing is tied to one model, and if AI is down for the day, the rec still runs.
The offer
The offer is a free 3-month review, and it’s twofold. It’s hard to say no to, and three months of data shows me the quality of your books. When it comes time to price the service, I’m not going in blind.
Closing
I avoided this idea for the longest time because running recs is boring, and it’s not something every business needs done professionally. Sometimes a simple Excel sheet can handle most of the job. But there are a ton of businesses with real volume, and because the work is a pain, it’s the perfect job to hand off.
In 12 months, success looks like 5 to 10 active clients and a clear answer on where this actually makes sense. I’m going after multiple verticals to start. The goal is to find proper product market fit.
I’ll post updates here as it goes, numbers included.