What an operations agent for delivery orders costs
The price of an operations agent for delivery orders, from a Proof on your own order feed to a governed agent live in your systems, and what makes it cost more.
An operations agent for delivery orders starts with a Proof: from 1,500 euro, 5 working days, one detector family on your own order feed in shadow mode. A Pilot live in your systems is from 6,000 euro for 30 days. The full agent, with a rulebook and a ledger, is a Build from 20,000 euro over 3 to 6 weeks.
It is for ordering, POS and delivery platforms, and for restaurant groups that sell on several apps.
What the agent does
- Stuck orders. It finds orders the apps accepted that never reached your POS or the kitchen.
- Menu drift. It finds a dish the kitchen ran out of that an app still sells.
- Written authority tiers. Some actions it takes alone. Some it takes and tells someone. Some wait for a person. Some it never takes.
- A log with an undo. Every step is written down with the evidence, the rule that allowed it, the call it made and the undo.
- Self-graded accuracy. It scores itself per detector, per day, and names the weakest.
More in the stuck orders write-up and the audit trail write-up.
What each step costs
| Step | Price | Time | What you get |
|---|---|---|---|
| Proof | from 1,500 euro | 5 working days | One detector family on your feed, in shadow mode: it reads and reports, and writes nothing. Graded by 3 to 5 pass/fail tests written with you on day 1. |
| Pilot | from 6,000 euro | 30 days | Live in your systems against a baseline you agree first. |
| Build | from 20,000 euro | 3 to 6 weeks | Several detector families, or the governed agent with a rulebook, a ledger and its own accuracy page. |
| Care | from 500 euro a month per live system | Monthly | Monitoring, fixes, small changes, a monthly accuracy report. Hosting in the EU or UK region included when you have no cloud team. |
The Proof is credited in full against a Pilot booked within 30 days. 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. Prices are in euro, excluding VAT.
What moves the price
- How many channels. Each delivery app is its own feed. Deliveroo's developer documentation, for example, lists separate APIs for orders, menus and sites. Each channel needs its own reader and tests.
- How many POS vendors. A stuck order is found by comparing the app's side with the till's side. One POS brand needs one till reader. A platform that connects many brands needs one for each.
- Whether writes are allowed. Reading and reporting costs least, and is all the Proof does. Each write, such as re-sending an order or switching an item off, needs an adapter, a test, an undo and a line in the rulebook.
- Where you want a person. You choose the tier of every action. Actions it takes alone need more tests before they go live, because nobody checks them first. A person in the loop costs less to build and more of your team's time.
A worked example, on synthetic data
Our demo runs on synthetic data, made up to look like real orders. On a busy evening a till's login token expires. Accepted orders never reach the kitchen. The agent finds them, refreshes the token and re-sends the waiting orders, which its rulebook lets it do alone. The log shows what it saw, the rule, the call and the undo. Taking a store offline on every app waits for a person. Refunds, prices and contracts it never touches.
On 7 replayed days, 249 planted incidents, it scored precision 0.98 and recall 0.99. It handled 95% without a person and handed 5% to a person, with the evidence attached. These scores are against faults we planted. Your Proof is graded by your tests, on your data. See the demo run or read its accuracy page.
Replies within 4 hours, 08:00 to 22:00 Dubai time.
Send us the job in the form below and a plan with a fixed price comes back within 24 hours.
Contact
Send us the problem or email jayron@geurs.com