For POS companies

We build an operations agent for POS companies. It watches the path from the delivery apps to your tills, catches orders that never reach the kitchen, and fixes what your rulebook allows. It starts as a Proof on your real data: from 1,500 euro, 5 working days, in shadow mode, graded by 3 to 5 tests written with you on day 1.

The problems on your side

• Tokens that expire. Uber's developer docs say client credentials tokens for its Eats APIs are valid for 30 days, and that past 100 tokens a new one invalidates the oldest. When one lapses unnoticed, orders stop reaching the till.
• Orders that time out. Deliveroo's API docs say an order in its tablet-less flow that is not accepted or rejected within 10 minutes is auto-rejected, and that the timeout can differ by region.
• Menus and availability drifting. The till says an item is sold out, the app still sells it. Deliveroo's menu docs say each item's PLU must find a match on the restaurant's POS, and that an item marked unavailable comes back at the next morning stock reset. A missing PLU means orders the kitchen can't make.
• Printers going silent. The kitchen printer stops, and the till keeps accepting orders.
• Release regressions. A new till version ships, and some merchants start dropping a modifier or an order type. Support hears about it one ticket at a time.

What the agent does, under a rulebook

The agent reads your integration logs, order events and till status for every merchant, all day.

• It checks every app order against what reached the till and the kitchen.
• It warns before a token lapses, and flags a merchant whose orders stopped arriving.
• It compares each merchant's menu and item availability between the apps and the till, PLU by PLU.
• It notices a printer that has gone quiet while orders keep coming.
• After a release, it compares errors by till version against the version before.
• It fixes what your written rulebook allows, for example re-sending an order or marking an item sold out on the app. Everything else goes to a person, with the evidence attached.
• It logs every action with its evidence and an undo.

What your partners and merchants get

Merchants get orders that reach the kitchen and menus that match what they can cook. Your support team gets a case with the evidence attached, not a phone call. Your delivery partners get a record of where each order stopped.

The path: Proof, Pilot, Build, Care

• Proof, from 1,500 euro, 5 working days. One workflow, such as expiring tokens or stuck orders, on your real data, in shadow mode: it reads and reports, it writes nothing. Credited in full against a Pilot booked within 30 days.
• Pilot, from 6,000 euro, 30 days. The workflow live in your systems, measured against a baseline you agree first.
• Build, from 20,000 euro, 3 to 6 weeks. Several workflows, or a governed agent with a rulebook, a ledger and its own accuracy page.
• Care, from 500 euro a month per live system. Monitoring, fixes, small changes, a monthly accuracy report. Hosting in the EU or UK region is included when you have no cloud team.

Half up front, the other half when it passes the tests. If it fails and we can't fix it within 5 working days, the first half comes back. You own the code.

A worked example, on synthetic data

Our delivery operations agent is live to click. Every order, store and failure in it is synthetic. The stuck-orders write-up explains what it checks.

On 7 replayed days, 249 planted incidents, it scored precision 0.98 and recall 0.99. The result: 95% handled without a person, 5% handed to a person, each hand-off with the evidence attached.

Those numbers are against failures we planted, not your merchants. Your Proof is graded by the tests you write with us, on your data.

Send us the job in the form below and a plan with a fixed price comes back within 24 hours.