This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady.
🦥 Sloth Collector — Less chasing. More doing.
What I Built
Small businesses can do the work, send the invoice, and still spend weeks waiting for the money they've already earned.
And the problem isn't just waiting. It's the repeated follow-ups, checking due dates, calculating delays, and finding the right words to ask a customer for payment without making the conversation awkward.
A few numbers made this problem hard to ignore:
- 29.3 days: The average time US small businesses waited to be paid in Q2 2026, with invoices arriving about 8.5 days past their due dates, according to Xero data. (Source)
- 59%: Nearly three in five small businesses reported having invoices overdue by 30 days or more in 2026, up from 47% the previous year. (Source: QuickBooks 2026 Small Business Late Payments Report)
- 51% vs. 21%: Businesses with overdue invoices were more likely to report internal delays in paying their own contractors, suppliers, vendors, or creditors than businesses without overdue invoices. (Source: QuickBooks)
These figures describe different parts of the problem, but they point to the same frustrating cycle: money arrives late, cash flow gets tighter, and the pressure can spread to other businesses.
That is where Sloth Collector comes in.
I built a local AI-powered invoice reminder app that helps users identify overdue invoices and prepare polite payment reminders without doing every repetitive step manually.
Upload a CSV containing invoice details, review which invoices are overdue, select an invoice, and generate a reminder with a locally running AI model. You can edit the message and then open it in your email app or WhatsApp to send it yourself.
The idea fits the challenge theme, Slow and Steady: let a small helper handle the repetitive preparation so people can spend less energy chasing and more time doing their actual work.
It doesn't promise to make customers pay faster. It makes the follow-up process easier to start.
Demo
Watch Sloth Collector in action: [https://youtu.be/xwwbwu_9C7U]
The demo shows the invoice workflow, overdue detection, AI-generated reminder, and contact handoff.
Code
GitHub repository: [https://github.com/Flockstock/Sloth-Collector]
How I Built It
I used Python and Streamlit for the application interface, Pandas for invoice processing, and Ollama to run the qwen2.5-coder:3b model locally.
The app validates uploaded CSV data, checks invoice due dates and payment status, calculates overdue amounts and days, and passes the selected invoice details to the local model to draft a polite reminder.
The user remains in control: the generated message can be reviewed and edited before opening the email or WhatsApp handoff. The app does not send messages automatically.
I chose local inference instead of a paid AI API because I wanted to build something useful without adding an ongoing API cost. The invoice processing and model inference run locally, so invoice details used for generation don't need to be sent to a third-party AI provider.
Why Does Open Innovation Matter?
Building with an open-weight model and free local tooling made this project accessible to me as a student working with limited resources.
I could experiment with AI without paying for API calls, and I could run the model on my own laptop instead of relying on a hosted AI service.
That matters for tools handling business information. Local inference can help reduce unnecessary sharing of invoice details with external AI providers, although users still need to consider how their files and other connected services are handled.
Open innovation made it possible to experiment, learn, and build a practical tool without needing a paid AI infrastructure budget.
The Lazy Payoff
Sloth Collector targets the small, repeated tasks around invoice follow-ups: finding overdue payments, checking how late they are, and drafting a message.
I haven't measured a reliable time-saving figure or tested whether the app helps businesses collect payments faster, so I won't claim a number.
The payoff I'm aiming for is simpler: less manual preparation, fewer reasons to put off a follow-up, and more control over a repetitive task.
The reminders are drafts, not automatic messages. A person reviews and chooses whether to send them.
Prize Categories
Overall Hacktoberfest Open-Source AI Challenge submission.
Built for Hacktoberfest 2026 — Week 3: Slow and Steady 🦥
Top comments (0)