Receipted AI for Hospitality: When a Guest Says Your AI Promised Them a Price

It is 2am on a Friday when your AI booking agent takes a call. The guest is booking a long weekend stay. The agent confirms availability, quotes the rate, takes the booking details, and creates the reservation in Seekom. Everything works as designed. The guest hangs up with a confirmed booking reference.

On Friday afternoon, the guest checks in. They tell you the rate at reception is different from what the AI quoted them. They have no written confirmation — the booking was by phone — but they are confident in what they heard. You have the Seekom booking showing the rate your system uses. But what did the AI actually quote? What exact rate did it communicate during that conversation? Without a receipt for that interaction, you cannot answer the question with precision. You are in a dispute with no paper trail.

The Air Canada Precedent Applies in New Zealand

In 2022, Air Canada's chatbot told a grieving passenger that he could apply for a bereavement discount retroactively. This was not Air Canada's actual policy. The chatbot invented it. Air Canada argued that the chatbot was “a separate legal entity” responsible for its own statements. The British Columbia Civil Resolution Tribunal rejected this argument entirely and held Air Canada liable for the discount the chatbot had described, plus tribunal costs.

The legal principle the tribunal applied is straightforward: a business is responsible for the representations made by its agents, including digital agents. When your AI system talks to a customer, it is acting on your behalf. What it says creates expectations that, under NZ consumer law, your business may be required to honour. If a guest can produce credible evidence that your AI quoted them a particular rate, the burden is on you to disprove it — or honour it.

New Zealand hospitality operators are not insulated from this. A similarly documented chatbot invention at a NZ hotel would face the same analysis under the Consumer Guarantees Act and the Fair Trading Act. “The AI said it, not us” is not a defence that has survived legal scrutiny anywhere it has been tested.

The 2am Booking Scenario in Practice

The 2am booking is not an edge case — it is one of the primary value propositions of AI for hospitality. Your property captures direct bookings at hours when your reception desk is unstaffed. The guest gets an immediate confirmation instead of an email in the morning. The booking goes directly into Seekom without requiring a staff member to manually enter it.

But a 2am booking handled by a black-box AI is also a 2am liability created without a witness. If the AI communicates the wrong rate — because your rates were misconfigured, because the AI misapplied a seasonal rule, or because there was any error in the interaction — you will not know until the guest arrives and the dispute begins. At that point, you have a Seekom booking showing one rate and a guest claiming another. You have no receipt showing what the AI actually quoted.

A Receipted AI agent changes this scenario completely. Alice creates a receipt for every booking interaction: the rate communicated, the room type confirmed, the Seekom booking reference created, the confirmation sent to the guest, and the timestamp. If a guest disputes the rate on Friday afternoon, you open the receipt from 2am Friday. It shows exactly what was quoted. The dispute resolves in minutes, one way or the other — either you have evidence that the correct rate was communicated, or you have evidence of an error in your rate configuration that needs to be fixed. Either outcome is better than a dispute with no records.

Seekom Integration: The Receipt and the PMS Record Connect

AliceHQ integrates directly with Seekom, the property management system used across a significant portion of NZ accommodation. The integration means that when Alice handles a booking call, the reservation is created directly in Seekom at the time of the call. The receipt captures the Seekom booking reference returned at the moment of creation, so the connection between the AI interaction and the PMS record is explicit and timestamped.

This matters for dispute resolution because it closes the possible argument that the AI said one thing and the system shows another. The receipt and the Seekom record are linked by the booking ID. If the receipt shows the AI quoted $140 and the Seekom booking was created at $140, the paper trail is complete. If there is a discrepancy, the receipt identifies exactly where it occurred.

Privacy Act 2020 and Guest Data Handling

Beyond booking disputes, NZ Privacy Act 2020 obligations apply to every interaction in which your AI handles a guest's personal information — their name, contact details, payment information, and stay preferences. Information Privacy Principle 5 requires you to protect this information from loss or unauthorised access or use. Principle 8 requires that information be accurate.

An AI booking agent that collects guest details and writes them to Seekom is handling personal information at the point of collection. The receipt of that interaction is your evidence that the information was collected with the guest's knowledge, for the purpose of the booking, and written accurately to your system. If a guest or the Privacy Commissioner ever asks what happened to their information during a booking call, you can answer that question from the receipt: what was collected, what was written to Seekom, and what the guest was told.

Australia's Privacy and Other Legislation Amendment Act 2024 introduces additional automated decision-making transparency requirements from December 2026. For NZ hospitality operators with Australian guests or Australian operations, the nine months before that activation date is the window to implement receipt-grade AI.

What Did Alice Quote Them? You Have the Answer.

The phrase that property managers, hotel owners, and trades operators come back to when evaluating AI accountability is always some version of “what did it book at what price?” This is the operational question that matters. Not whether the AI worked. Not whether it was polite. Whether the record of what it did — the price communicated, the booking created, the confirmation sent — is accessible and accurate.

Receipted AI answers that question by design. Every booking has a receipt. Every receipt shows the rate, the room, the system write, and the timestamp. If something goes wrong, you have the receipt. That is what operational accountability looks like when AI is running your front desk at 2am.