FollowUpInvoice — AI Invoice Follow-Up Automation

The problem we set out to solve

Freelancers and small agencies do not have a collections department. They have one person who is also the designer, the developer and the account manager, and who has to decide — on a Tuesday morning, about a client they want to keep — how to word the third email asking to be paid for work delivered six weeks ago.

The result is predictable. Follow-ups get postponed because they are uncomfortable, then sent all at once when cash gets tight, in a tone that has drifted from friendly to annoyed without passing through firm. Studies of freelance invoicing put a third or more of invoices past due at any given moment, and the people we spoke to during discovery were losing about three hours a week to chasing money they had already earned.

The tooling that exists mostly solves the wrong half. Invoicing apps are very good at producing an invoice and almost entirely silent afterwards — the part where the money actually arrives is left to the user.

What we built

FollowUpInvoice is a full SaaS product, built end to end by ByteStackLab and running as one of our own products. It covers the whole cycle from client record to collected payment.

• AI reminders that escalate properly. GPT-4 drafts each follow-up with the invoice context — amount, age, client history, prior correspondence — and shifts register as the invoice ages: light at seven days, direct at thirty, unambiguous at sixty. Users can approve before sending or let it run.
• Payment inside the email. Every reminder carries a one-click Stripe link. The gap between "I should pay this" and paying it is where most late payment actually lives, so we removed it.
• Branded invoices in under two minutes. Line items, taxes, due dates, logo and colours, exported as a clean PDF and sent from the user's own domain over their SMTP, so it lands as their email rather than a vendor's.
• Recurring invoicing for retainers. Weekly, monthly, quarterly or yearly schedules that generate and send on their own, which for retainer-heavy freelancers removes the single most repetitive task of the month.
• Revenue analytics. Collection rate, overdue ageing, monthly revenue and per-client payment behaviour — the numbers needed to answer "can I take next month off" honestly.
• Slack integration. Daily digest, one-tap reminder approval and payment notifications, so the whole loop can be run without opening the dashboard.

Engineering decisions worth calling out

Tone escalation is a state machine, not a prompt. The model is given an explicit stage and a set of constraints for that stage rather than being asked to "sound firmer." Left to improvise, an LLM produces a plausible email that occasionally insults a client — which is exactly the failure a freelancer cannot afford. Constraining the ladder made output boring and reliable, which is what this job needs.

Delivery goes out over the user's own SMTP by default. Sending from a shared vendor domain would have been far simpler, and it would have put every customer's deliverability at the mercy of the worst-behaved account on the platform. For email whose entire purpose is to be read and acted on, that trade was not worth making.

Sending limits are enforced per plan at the queue layer, so a runaway schedule degrades into a backlog rather than a spam complaint. Everything reminder-related runs through queued jobs — a slow AI call or a Stripe timeout never blocks a user's request.

Results

Customers report getting paid around 40% faster and recovering roughly three hours a week previously spent on follow-ups, with one consultancy reporting its overdue rate falling from about 35% to under 8%. Setup to first invoice sent takes under two minutes, and the product holds a 5.0-star average rating from its reviewers. It ships on a three-tier plan — $49 Starter, $199 Professional, and custom Enterprise pricing.

The wider point for our clients: this is a product we built, launched, priced and now operate ourselves — the same discovery, architecture, AI integration and payments work we do on client engagements, applied where we carry the outcome.