How Does the Bot Actually Know What’s Available? (No Guessing, No Overbooking)

The real worry isn’t "will it sound robotic," it’s "will it lie"

Most first-time worry about an AI DM bot isn’t tone, it’s trust: what happens the day it tells a customer 7pm is open when it isn’t, or books two parties into the same table. That’s a fair thing to want answered before you connect a real Instagram account to it.

The short version: the model never invents availability

The language model that writes the reply is not the thing that decides what’s free. Every time it needs to check a slot, it calls a tool that queries your actual schedule — working hours, existing bookings, staff or table assignments, buffers, time off — and gets back real data. It can only offer what that query returns. There is no step where it pattern-matches a plausible-sounding time from training data; if the tool says nothing is open, the model has nothing to offer.

Why this is a hard boundary, not a policy

This isn’t "we told the model to be careful." Reservify’s agent has no direct database access at all — it can only act through a fixed, typed set of tools built for this specific job. That’s an architectural wall, not a prompt instruction, which matters because prompt instructions can be argued around by a clever enough message and a hard boundary can’t.

The same wall is what prevents overbooking: the availability check and the booking write both go through the same real-time source of truth, so two DMs racing for the same slot can’t both win it.

What this means day to day

If you close early on a Tuesday, mark a table under maintenance, or give a stylist the afternoon off, the bot sees that immediately on its next check — there’s no separate "bot calendar" to keep in sync, and nothing for you to remember to update in two places.

When the answer genuinely isn’t "yes" or "no"

Some questions aren’t availability questions at all — a price negotiation, a complaint, a request to speak to a person. The bot is built to recognize that and hand the conversation to you rather than guess at an answer it has no tool for. That escalation path is its own piece — see the next article.