Bots or approved agents? Keeping bookings, stock and returns fair

Same request, two very different visitors

A customer asks their assistant: "Get me a table for two on Friday at eight." That's a good visitor: one customer, one table, a real dinner.

Now picture a script with fifty accounts asking the same thing for every Friday slot, then reselling them. In September 2026 a large US reservations platform found out how close those two look: it banned diners whose assistants kept checking for tables about 200 times an hour, then had to restore some of them. The line the industry is drawing is clear: approved, known agents in; anonymous bots out. Drawing it well takes more than a block list.

1. Know who is asking

An agent's name in a header proves nothing. Allowfy recognizes agents by their keys:

• Agents that connect to your Allowfy gateway sign in with their own key.
• Agents and shopping platforms that sign their requests the way Visa, Mastercard and Cloudflare are standardizing (Web Bot Auth) are checked against the keys they publish. You see each one, how often it comes, and you can block it in one click, or answer signed agents only.

2. Cap what one agent can take

Limits per customer already stop one person from booking ten tables. But a bot with many accounts is many "customers". So Allowfy now also caps one agent across all its customers: for example, no more than 20 bookings a day from any single agent. A busy, honest assistant stays well under it; a hoarder hits it fast, and what's over goes to your team instead of into your calendar.

The ready-made Fair bookings rules set this up in one click: one new slot per customer a day, 20 per agent, the rest to your team.

3. Say clearly why, so good agents don't give up

When an agent is refused, it needs to know why. Is the slot taken? Is the agent blocked? Does the customer need to approve? Should it try again in five minutes? Every refusal from Allowfy now says so, in a form the agent can act on: unavailable, not allowed, agent blocked, needs the customer, paused, try later. An assistant that can tell "sold out" from "blocked" won't send your customer to a competitor by mistake.

4. Watch what each agent does

The Agents page now shows, for each agent over 7, 30 or 90 days: what it did, how it ended, the returns and the amounts returned, cancellations, refusals by your rules and what went to your team. An agent whose actions are mostly returns is flagged, so you can look closer, or add a limit.

5. Cash on delivery deserves its own rule

In the Gulf and Egypt, many orders are still paid in cash on delivery. Returning cash, or cancelling an order a courier is already carrying, is something most stores want a person to handle. Allowfy now reads how each order was paid, on Salla, Zid, WooCommerce and Shopify, and one rule sends returns and cancellations of cash orders to your team while everything else stays automatic.

The result: say yes to agents, safely

None of this is about keeping AI out. Customers will keep handing tasks to assistants, and the businesses that welcome them will get the bookings and orders. The point is to welcome them on your terms: known agents, fair limits, clear answers, and a record of everything.

Start free in test mode, apply the ready-made rules, and watch the activity before you go live.