I built an AI Data Analyst for SaaS companies to make better decisions based on their data
I spent the last month building a side project that solves a problem I've encountered myself in my work. I work as a software engineer at a short-rent marketplace that charges owners to list their place.
To make better decisions we had to use AI, and to get better decisions and insights we had to feed the AI our data that we extract from the DB, Google Analytics, and/or Stripe. It was a painful process to get to have answers for questions like: Can you compare churn rates across all plan types for the last three completed months? or What was the churn rate by revenue, not just customer count? etc.
So, we built a tool for it, and then we decided to make it available for others with the name DecidexAI. The idea is simple, you connect your Stripe account and/or Google Analytics, we get all your historic data, and then you can use AI to build dashboards with just plain English, and ask your data questions and get answers. Plus there are more than 7 prebuilt dashboards with 68 metrics computed automatically every 5 minutes. The good thing is that you get notifications for your saas metrics via email and/or via slack. We also built a Churn Analyzer that work with your real Stripe data to detect possible churns and tell you how to avoid it.
There was a critical part though, how do we get to understand the SaaS business of the customer so when we provide a suggestion or an answer it would be super beneficial? We came up with a pretty simple solution, we ask the customer to provide his website, and then we read his site, and make the content available for the AI to digest and understand the business and the saas pricing etc. And then when you ask a question to get answers based on your data we take this into account. For example, the tool can suggest based on your data that you need to add more features to a particular plan that you have, why does it know that? that's based on churns on that plan from your stripe + the pricing of your site.
Here is a demo video:
https://youtu.be/Agnb0LAA-y8?si=vgwb2qz71x72P6kU
But a question arises: I can connect Stripe MCP to Claude — why do I need this app?
If you're technical enough to use Stripe MCP with Claude Code, you can absolutely query your Stripe data in natural language. That works. But here's what it can't do:
- Remember yesterday. Stripe's API is point-in-time — it tells you what's true right now. Ask "what was my MRR three months ago?" and there's no answer, because nobody was snapshotting it. The app records your metrics daily, so you get trends, comparisons, and cohort retention over months — not just a live snapshot.
- Give you consistent numbers. When Claude calculates MRR from raw Stripe objects, it's doing math in-context every time — and may get different answers between sessions depending on how it interprets subscription objects. DecidexAI pre-computes 68 SaaS metrics with precise definitions that produce the same answer every time. For board reporting, consistency isn't optional.
- Watch your data while you sleep. MCP tools are reactive — nothing happens until you open a terminal and ask. The app monitors your metrics 24/7 with anomaly detection, sends Slack alerts when churn spikes, delivers a daily metrics briefing every morning, and writes an AI executive report every week. The insights come to you.
- Connect more than Stripe. "How does paid traffic correlate with trial conversions?" requires Stripe + GA4 in the same query. MCP tools are single-source. DecidexAI connects Stripe, Google Analytics, and your website content into one semantic layer the AI reasons across.
- Work for your whole team. Your co-founder, head of sales, or investor isn't opening a terminal. They need a browser-based dashboard where they type a question and see charts. The app gives them that — MCP gives them nothing.
Think of it this way: Stripe MCP is a developer tool. The app is the SaaS data analyst you'd hire at $100K/year.
I'd love to hear what you think about this project!