Case study · Founder, via Zild · Live in production

JustPayMe: collections run from a chat.

I lead the technology at Zild, the company behind JustPayMe. It gives a plumber, a tutor, anyone running a business on their own, a back office in their pocket: overdue invoices chased professionally from WhatsApp, with the owner approving every message before it goes out.

WhatsApp chat mockup showing a business owner sending a voice note to the JustPayMe assistant, which replies confirming it will bill a client £850 and chase payment if not settled in seven days.

The problem

Chasing an unpaid invoice is the job nobody wants. It's awkward, it competes with the work that actually pays, and so it slips. The reminder that should go at day seven goes at day thirty, or never, and the business eats the gap in its cash flow. Both sides of the problem already live in WhatsApp; collections just didn't.

What we built

The owner runs everything from one thread, by text or voice note: "chase the Henderson invoice", "Rachel paid", "add a new client and bill him £50". The assistant works out what they mean, drafts the action, and waits. One tap approves it. The client gets a professional, well-timed reminder from a separate business number, can reply in plain language, and can pay straight from the message. The moment a bill is paid, the remaining reminders stop on their own. The same flow runs over SMS, email and Telegram, and it syncs with Xero, Sage, QuickBooks and FreshBooks so it only ever chases real invoices.

Where the real work was

Sending a reminder is easy. Making a messaging channel dependable enough to run money through is not. The conversation has to survive the messy middle of real usage: corrections, disputes, promises to pay that don't materialise, "I already told you". The tone has to grade from friendly to firm based on where the client actually is in the cycle, not a fixed calendar. And the production discipline has to be airtight: never chase twice, never chase the paid, respect STOP rules, retry delivery sensibly. That judgement layer is the product.

The owner stays in control of every send. The assistant just makes the easy thing the professional thing.

What it means

This is what "AI-native fractional CTO" looks like in practice: not a strategy deck about agents, but an agent in production, handling real payments for real businesses, built to be boring and reliable. If your company needs someone who can both call the architecture and ship it, this is the reference.

Built with
WhatsApp Business Platform voice-note transcription LLM intent and reply understanding human-in-the-loop approvals multi-channel delivery accounting integrations

Tell me what's next.

Fifteen minutes, no pitch. If you need a fractional CTO, we'll figure out the shape together. If you need something built, we'll scope it. If it's not me, I'll say so and point you the right way.

You work with me directly. Not an agency, not a sales rep, not a junior.