What Hotel Teams Fear About AI, and What Settles It

Ask a hotel sales team why AI isn’t doing any of the work yet and you rarely hear “it doesn’t work.” You hear four other things:

1. What if it gets something wrong?

2. Where does our data go?

3. Will customers hate talking to it?

4. And, usually last and more quietly: will this take our jobs?

None of these is a silly question. Each is a question about control, and the only credible answer to a question about control is a boundary you can see, change, and audit. This post takes the four fears in turn, says what to ask any vendor, and shows where MeetingPackage, the AI-enhanced Operating System for Groups, Meetings & Events, draws its lines. Then it covers where to start, which is smaller than most teams expect.

What if it gets something wrong? Decide what it is allowed to decide

Be fair to the baseline first. Manual handling makes errors too; it just makes them quietly. Sally Booth, UK Group Strategy Director at Marriott International, said in an Edwards & Finn white paper: “It’s just very manual. It relies on that person understanding the situation and entering the data correctly. If they don’t, the decision-making becomes flawed.” We looked at what manual re-keying costs a desk in a separate post.

So the useful question isn’t whether AI can err. Anything that reads an inbox can. It’s what an error costs and who catches it. The fix is to shrink what the AI is allowed to decide, then make every handover to a person easy.

• Reading without pricing. The AI Email Agent sorts genuine RFPs from general enquiries and forwards everything else to a fallback address your team owns. It checks the request against the property’s configured capacity and amenities, and sends one follow-up asking only for what’s missing.

• Escalating on your thresholds. Rules alert the team when a high-value RFP arrives, in one of three modes: notify only, notify and continue, or notify and take over.

• Handing over on demand. An admin can switch any conversation to human takeover at any time, and every later message goes straight to your team.

• Quoting inside your rules. Instant Quote answers only what your decline and takeover rules allow, based today on group size and event duration. Anything else is declined politely or routed to a person. It also refuses to guess: enablement is blocked while a required rate code is missing, and a property without a connected PMS can’t go live until a named person has verified its inventory data.

Where does our data go?

Data is the fear with the shortest path to an answer, because the answer is a document. Ask every vendor the same five questions. Here is what MeetingPackage’s Data Processing Agreement says to each.

Two honest caveats. An agreement covers the platform as a whole, so ask every vendor, us included, to spell out in writing how each answer applies to the AI features: whether any data trains a model, which third parties process it, and where. And ask for the current sub-processor list and hosting regions before your security team signs off. They will want both.

Will customers like talking to AI? They like fast, accurate answers

Today’s baseline is not a delighted planner. Groups360 puts the average hotel RFP response rate at around 45%, so more than half of enquiries get no reply. “People are used to getting confirmations instantly. If they don’t get that experience, they drop off,” says Fredrik Hult, Head of eCommerce at Strawberry Hotels.

The customer-facing AI work is also narrower than a chatbot. The AI Email Agent sends one follow-up in the customer’s own language and never repeats a question. Once the request is in, the customer receives a branded confirmation carrying their order ID. When Instant Quote’s answer is no, you decide what the declined customer sees.

Speed doesn’t remove people from the picture. Amy Russell of PPHE Hotel Group, on online booking: “The customer is happy because they’re receiving a quick response. They can book instantly then and there.” She is also clear about the limit: “I don’t think RFPs will ever go away. There’s always going to be an event that people want to have that human connection with.” Keep a visible route to a person for those, and measure satisfaction rather than assuming it: compare reply times and win rates before and after each stage.

Will this take our jobs? The teams running it describe the work moving up

Amy Russell names the fear plainly: that online booking means people lose their jobs because human interaction is no longer needed. Her team’s experience is the opposite. “You still need the same number of staff to do that because our staff are now able to focus on the larger pieces of business.” Junior colleagues who used to work small meetings now work on larger events, she says, and are growing faster for it.

Hold one line when you brief your own team: the aim is more business handled by the same people, not the same business handled by fewer. Say it before rollout, not after. Bespoke, high-value events stay with people, because that is where a person changes the outcome.

Hand over control in stages

Bethan Stone, Applications Director at De Vere Hotels, describes how the handover feels from the desk: “It’s taken them a little time to get used to the idea that they don’t have overall control of their diary anymore.” Her team also worried that bookings would land that they wouldn’t have chosen themselves. Comfort grew as the bookings came through, and she says the team embraced it more as time went by.

De Vere went in stages. In a separate interview, a De Vere colleague described giving the system short-lead space first, with a typical lead time of about 28 days by their own estimate. Large events booked six, nine or twelve months out stayed in human hands. More availability went to the system over several months as trust built.

A four-stage ladder suits most desks:

Enablement is its own step. Saving rules never starts auto-quoting anywhere, and every enablement change needs an explicit confirmation.

Start with the list of what eats your week

The most practical first move needs no vendor. Write down every task and problem that eats your team’s time: retyping RFPs, chasing missing details, rebuilding the same proposal, sending reminders. Hand that list to an AI tool and ask two things of each item: could it take this on, and what goes wrong if it gets it wrong? Then pick the item that costs the most hours and needs the least judgement.

For many desks, that item is re-keying. MeetingPackage’s own estimate is 30 to 60 minutes per RFP email, which the AI Email Agent brings down to seconds. We mapped where RFP hours go in detail. For a stage-by-stage view of what to automate and what to keep human, see where AI actually earns its place in hotel event sales.

Bring the list

Fears about AI are really requests for control. Book a demo, bring your list, and we’ll show you where every switch sits.