Agents in hospitality: reservations + recovery

A hotel GM we worked with had it right: "Anyone can sell a clean room. The brand is whether we know how to fix the dirty one." Hospitality agents that focus on reservations are useful. Hospitality agents that focus on recovery — the moment between a guest experiencing a problem and that experience becoming a reputation issue — are differentiators.

The recovery use case is where the highest leverage sits. The reservations use case is table stakes.

Why recovery beats prevention

Most hospitality businesses already have the prevention systems. PMS software, channel managers, OTAs, point-of-sale, kitchen display systems. These are mature. Agents that try to "improve reservation booking" usually shave seconds off a process that already works.

Recovery is different. When something goes wrong — late check-in, cold meal, missed reservation, ignored request — the response window is short and the brand stakes are high. Hospitality businesses are bad at this not because they don't care, but because the staff who notice the problem are not the staff with the authority to fix it.

A working agent bridges this gap.

Front-desk to GM. Guest mentions an issue at check-in. Front-desk agent logs it, classifies severity, suggests recovery action (room change, comp, F&B credit), routes to the GM if it's above threshold. GM approves with one tap. Recovery happens within minutes instead of hours.

Server to manager. Server flags a slow ticket. Kitchen agent reads ticket history, the reservation context, and prior visits. Suggests a recovery (apology, comped course, manager visit). Manager confirms. Recovery happens at the right moment, not after the guest leaves.

Post-stay to brand. Negative review posted online. Brand agent drafts a response with the specific details from the reservation referenced. Brand manager reviews, edits, publishes. Response is personal, fast, and on-brand.

Personalisation without overreach

The privacy line in hospitality runs the same way as retail: lean toward first-party data, opt out of aggressive history use.

A guest's preference for a high-floor room, recorded by them on a previous stay, is fair game. The fact that they always order red wine on the second night, inferred from POS data without their explicit opt-in, is not. Even if the inference is correct, surfacing it feels surveilling, not solicitous.

The agent's design should default to guest-volunteered personalisation. Anything else needs an explicit opt-in.

Front-desk handoff

Hotels run on shifts. The agent should preserve continuity across them. A guest's special request at 11 PM should be visible to the morning shift; the morning shift's response should close the loop without the guest having to repeat the ask.

This is unsexy CRM work. It's also where the agent's compounding value sits. Hotels with good shift-handoff agents have measurably better guest satisfaction than ones without.

Loyalty data discipline

Loyalty programs are the centerpiece of hospitality data. They're also where privacy risk concentrates. The agent's loyalty discipline:

Use stated preferences (room type, bed type, pillow preference, late checkout).
Use stated restrictions (allergies, accessibility needs).
Don't use spending patterns as a personalisation lever (the upsell can feel predatory).
Don't share loyalty data across properties without the guest's clear consent.

Restaurants: the no-show problem

Restaurants have a different angle. The hot use case is no-show prediction and recovery. Reservation systems can flag likely no-shows; agents can craft confirmation messages that feel warm rather than nagging; the host stand can have an alternate-guest list ready when a no-show happens.

The agent doesn't punish no-shows; it routes around them. Recovery, again.

What we won't ship

Anything that prices differently per guest beyond standard loyalty pricing. Personalised pricing in hospitality has the same regulatory and reputation risk as in retail.

Anything that makes operational decisions without the manager's signature. The recovery action — comp the meal, move the room, add a credit — is a manager's call. The agent suggests; the manager approves.

Anything that auto-responds to negative reviews. Reviews are public. The response is brand. Always human.

The four metrics

Time-to-recovery for guest issues.
Recovery-to-resolution conversion (issues that result in a happy guest vs. ones that escalate).
Guest-volunteered preferences captured per stay.
Repeat-guest rate for guests who experienced (and recovered from) an issue vs. ones who didn't have one.

The fourth is the killer metric. Hospitality businesses with great recovery actually have better loyalty among guests-who-had-an-issue than guests-who-didn't. Recovery is a brand asset.

How to start

Pick one property, one shift, one workflow. Recovery agent at the front desk during evening shift. Run for a quarter. Measure. Expand to morning shift, then to a second property.

Close

Hospitality agents earn their keep on recovery, not reservations. The recovery moment is short, high-stakes, and currently under-served. Build for the moment something goes wrong. Build the front-desk-to-manager bridge. The reservation use case will follow.