AI Cash Flow Forecasting Dashboard for Solopreneurs

For Episode 14 (September 26, 2025), Jordan built a finance dashboard for solopreneurs with our guest's company, Collective, in mind. You upload a profit and loss statement and a balance sheet, and Gemini produces a cash-flow forecast, an insurance compliance check, a lending readiness report and overall AI insights. He started it in Bolt and finished it in Cursor with OpenAI Codex. He built it in time for the episode but didn't give a build time.

Why we built it

Jordan has known Hooman Radfar, Collective's co-founder and CEO, for a long time. Collective is a back office for solopreneurs. Jordan isn't a Collective user, so he asked Hooman for a few features they were thinking about and built prototypes of them as a fun extra for the episode.

The stack

• Bolt: the quick first version.
• Cursor: the IDE he moved into.
• OpenAI Codex: coded the front end, wired things up and made sure the APIs worked.
• Vite, React, shadcn/ui and Tailwind CSS, plus a table-filtering library: the UI.
• Gemini: every forecast, check and report.

How we built it, step by step

1. Get something up in Bolt.
2. Move to Cursor and Codex. At 08:00:

But I quickly moved it into, like, my IDE, which is cursor, and I used OpenAI's codex to just do some coding for the front end and wire everything up and make sure, like, all my APIs work.

— Jordan Metzner, Episode 14

3. Use a familiar UI kit. Vite and React with shadcn/ui and Tailwind, because they're responsive and look good.
4. Make sample data. Jordan generated P&Ls and balance sheets for three fictional businesses under "Built This Week Holdings": Sam's sewing business, Jordan's design business and Hooman's consulting business.
5. Pick a model with room for the data. He started with a roughly 120B-parameter model he had access to, but the financial statements were too much for it:

I needed a bigger model, and so in fact, I moved to Gemini, and I'm using Gemini here because I was able to take the data in.

— Jordan Metzner, Episode 14

6. Forecast cash flow. Upload the CSVs, set revenue, expense and owner-salary outlooks (up, flat or down), and Gemini returns next-month, next-quarter and next-year projections with commentary.
7. Reuse the pattern. The same approach runs the insurance compliance check, the lending risk report with loan eligibility, and the AI insights panel.
8. Add themes. Light, dark, and a "Make It Rain" mode with a custom song made for Collective.

How it turned out

The forecast ran live on Hooman's fictional consulting business and called out healthy revenue growth with fluctuating margins. The insurance check rated compliance as moderate, flagged cyber liability, and caught a lapsed workers' comp policy. The lending view produced a risk report and an eligibility read. Jordan admitted the overall AI insights panel might not be "working that well."

Hooman joked that Jordan would now run Collective's marketing, and said he'd take the demo back to his product team. His broader point: people still underestimate how fast you can get to a prototype, but getting from prototype to production is the hard last mile.

What we'd do differently

• Size the model to the documents. Financial statements can overflow a smaller model; start with one that can take the whole input.
• Use real data before trusting outputs. Everything here ran on generated sample statements, and real lending decisions would use outside lending tools.

Jordan's bonus build that episode was the Make It Rain mini-game. For the tools, see Claude Code vs Cursor. More: AI in Finance.

FAQ

How do you build an AI cash flow forecasting tool for a small business?

Jordan's dashboard takes an uploaded P&L and balance sheet as CSV, lets you say whether revenue, expenses and owner salary will go up, stay flat or go down, and sends it all to Gemini, which returns next-month, next-quarter and next-year projections with commentary.

Which AI model works for cash flow forecasting from financial statements?

Jordan started with a roughly 120B-parameter model he had access to, but the P&L and balance sheet data was too much for it. He switched to Gemini because it could take the data in.

Can AI check small business insurance compliance?

In Jordan's prototype, Gemini reviews your insurance policies and flags gaps. On sample data it rated compliance as moderate, flagged cyber liability, and caught a lapsed workers' comp policy.

What is a good Bolt to Cursor workflow?

Jordan built the first version in Bolt to get something up quickly, then moved it into Cursor and used OpenAI's Codex to code the front end, wire everything up and make sure the APIs worked.