Inspiration: 82% of SME failures come from cash flow — not unprofitability. SMEs face 30–90 day payment terms while bills hit immediately. Existing tools (accounting, neobanks, dashboards) only show the past — none predict, decide, or act. We built Cash OS to close that gap.
What it does An AI-native agent in five layers:
SEE — real-time cash visibility via Open Banking PREDICT — alerts 14 days before you hit zero ACT — autonomous collections across email, WhatsApp, SMS DECIDE — conversational CFO copilot FINANCE — embedded credit the moment a gap appears
What we learned : Visualizing a problem is not solving it: the leap from insight to action is where value compounds. Trust must be earned in basis points — every prediction has to be explainable in plain English. The forecast core is simple at heart:
- Cash(t) = Cash(0) + Σ (probability × inflow) − Σ outflows
- An alert fires when the probability that Cash(t) < SafetyFloor reaches 50% within the next 14 days.s.
How we built it : TypeScript, React 19, TanStack Start, Vite, Tailwind CSS, shadcn/ui — deployed on Cloudflare Workers via Lovable. Editorial design: cream paper, forest ink, serif display, structured like an investment memo.
Challenges: Trust in autonomy — solved with a tiered dial: suggest, draft, send-with-approval, fully autonomous. Cold-start prediction — Bayesian priors from sector and size, refined per client over time. Multi-bank EU reality — Plaid, Bridge and Swan unified behind one normalized schema. Brand discipline — refusing to ship the 47th blue fintech dashboard.
Built With
- and-shadcn/ui
- cloudflare
- deployed
- on
- react
- tailwind-css
- tanstack-start
- via
- vite
- we-use-typescript
- workers
Log in or sign up for Devpost to join the conversation.