Inspiration

  • Freelancers lose money in two places: underpricing their work and chasing late or disputed payments.
  • Most invoices are priced by gut feeling, and disputes turn into long, awkward email threads.
  • We wanted an invoice that is defensible before it is sent, with PayPal as the backbone for getting paid.

What it does

  • Reads the client contract and the freelancer's work logs.
  • Extracts rates, scope, deliverables and payment terms.
  • Scores the invoice for fairness against market rates and work complexity.
  • Drafts line items with a plain-language justification for each.
  • Sends the invoice through PayPal Invoicing, and the client pays with a PayPal button.
  • Tracks disputes through webhooks and drafts evidence-backed replies.

Fairness score

$$ F = w_1 \cdot \frac{r_{\text{billed}}}{r_{\text{market}}} + w_2 \cdot \frac{h_{\text{billed}}}{h_{\text{expected}}}, \qquad w_1 + w_2 = 1 $$

  • F close to 1 means the invoice is in line with the market.
  • F well above 1 is flagged for review before sending.
  • F well below 1 tells the freelancer they are undercharging.

How we built it

  • AI layer: Claude/GPT API for contract parsing, rate benchmarking, invoice drafting and dispute replies.
  • Structured JSON outputs so every number can be validated before it reaches a client.
  • PayPal Invoicing API to create and send invoices.
  • PayPal Orders API for the pay-now checkout flow.
  • PayPal webhooks to track payment status and disputes, which update the dashboard and trigger the AI dispute drafter.
  • Frontend and backend: add your stack here (for example Next.js, FastAPI, Postgres).

Challenges we ran into

  • Hallucinated market rates: fixed by making the model cite its benchmark source and range, and clamping outputs to validated limits.
  • Messy inputs: contracts and work logs come in inconsistent formats and needed normalizing first.
  • Webhook reliability: handled retries, duplicate events and signature verification so payment status never went out of sync.
  • Dispute tone: replies had to be firm but professional, and a human approves them before sending.

Accomplishments that we're proud of

  • A full loop from contract upload to a paid invoice, with PayPal at the center.
  • A fairness score that explains why, not just a number.
  • Dispute drafts that cite the exact contract clause and work log entries.

What we learned

  • Designing AI outputs as verifiable structured data instead of free text.
  • The PayPal Invoicing lifecycle and webhook event model.
  • Trust is the product: both freelancer and client need to see the reasoning.

What's next

  • Live rate data from multiple market sources.
  • Time tracker integrations (Toggl, Harvest) to pull work logs automatically.
  • Recurring invoices and milestone-based billing.
  • Automatic late-payment reminders.

Built With

Share this project:

Updates

Submission history