Guest requests get written down at booking, then have to survive a chain of people before anyone acts on them. A cot request left in a reservation comment has to be noticed at the front desk, logged somewhere, logged in the right place, seen by housekeeping, and actioned before arrival. Every link in that chain is a point where the request can quietly disappear, which is why a guest sometimes arrives to find the cot is not there.

Tracey is Sweeply’s AI agent for exactly that failure. It reads the comments attached to a reservation in the PMS, works out what the guest asked for, and creates the task in Sweeply for the team that has to deliver it. Nobody retypes anything, and nobody has to remember.

What does an AI agent for reservation traces actually do?

A reservation trace is the free text attached to a booking: the note a guest types into the booking engine, the request an agent pastes in from an email, the preference a returning guest has on file. Most property management systems store that text well. What they do not do is turn it into work for a named person on a specific day.

That is the gap Tracey closes. It reads the trace, classifies the request, and creates the corresponding task in Sweeply, routed to housekeeping, maintenance or the service team depending on how the property has set up its categories. From there Sweeply behaves as it always does: the task lands on the right person’s phone, in their language, with the room and the deadline attached. Sweeply supports 20+ languages, so a note written by a supervisor in one language is read by a housekeeper in another.

Sweeply is an operational layer on top of the PMS rather than a replacement for it, so the PMS stays the system of record. You can see how the agents sit alongside the rest of the platform on the AI Agents page, and how the resulting work is run day to day under Housekeeping.

How does a reservation comment become a housekeeping task?

1. The guest writes the request at booking. Owner: the guest, through the booking engine, an OTA, or an email your team pastes onto the reservation. Duration: seconds.

2. The comment is stored on the reservation in the PMS. Owner: the PMS. Duration: immediate.

3. Tracey reads the trace and classifies it. Owner: the agent. Duration: seconds after the comment appears.

4. The task is created in Sweeply and routed to the right team. Owner: the agent. Duration: seconds.

5. The team delivers it and closes the task. Owner: housekeeping, maintenance or reception. Duration: within the normal task cycle, before the guest arrives.

6. A manager reviews what was created. Owner: duty manager. Duration: a few minutes a day, and rather more in the first fortnight while categories are tuned.

Routing depends on the task categories a property configures. A typical set-up looks like this:

What changed at THE FLAG in Zurich?

THE FLAG, a hotel and serviced apartment group in Zurich running on Apaleo, was one of the first operators to put the agent into live operation. Guest requests were arriving in reservation comments and reaching the team through a slow manual chain that was error prone and tiring to maintain.

After rollout, THE FLAG reported around 30 tasks created automatically every day and 3.5 hours a week saved on task creation alone, with fewer missed requests and less back and forth between reception and housekeeping. Sweeply’s published THE FLAG case study records the longer picture across the six properties: 1,800 reservation traces processed with 5 misses, more than 8 hours a week saved, and 80% fewer complaint-related tasks.

“Sweeply is more than a housekeeping app. It’s a fulfilment platform for AI agents. By opening our APIs to autonomous systems, we enable agents not just to understand guest needs, but to act on them. When an agent identifies a guest preference, Sweeply makes sure it actually happens,” says Frans Veigar Garðarsson, VP Product at Sweeply.

The agent runs inside the stack THE FLAG already had. It connects through Sweeply’s open APIs and operates within the Apaleo ecosystem, which is possible because Apaleo publishes its API openly in its developer documentation. No new infrastructure, and no second system for the team to learn. The connection itself is described on the Apaleo integration page.

Do you have to change your PMS to run an AI agent?

No. Sweeply connects to 20+ property management systems, including Mews, Apaleo, Oracle OPERA Cloud, Guestline, Cloudbeds and Guesty, and a property on a supported PMS can be live within 24 hours.

What matters is not the brand of PMS but whether guest requests are actually being written onto reservations inside it. If your team already captures requests as reservation comments, an agent has something to read. If those requests live in a staff WhatsApp group, the first fix is upstream in the booking and front-desk flow, not in an agent. That distinction is worth settling before you buy anything, and the case for wiring the two systems together is set out in more detail in the benefits of integrating your PMS with task management. If you want third-party reviews before that conversation, Sweeply holds a 4.8 rating on Hotel Tech Report.

Where an AI agent will not help

Being straight about this saves everyone a wasted trial.

• It only sees what is written on the reservation. A request made by phone, at the desk, or in an OTA message thread that never reaches the PMS is invisible to the agent. Those still need a person to log them.

• Ambiguous free text still needs judgement. “Something nice for the little one” can be classified, but not perfectly. Review the created tasks daily for the first couple of weeks and tune your categories as you go.

• Low request volume means low return. A small property with a handful of reservation comments a week will not get hours back. The value scales with the number of traces you handle, which is why groups and serviced apartment operators feel it first.

• It does not replace the conversation. Knowing about the cot is not the same as noticing that a guest looks exhausted at check-in. Removing the admin is worth doing because it frees people for the part software cannot do.

• Sweeply is not a PMS. If reservation data is thin or requests are never captured at booking, the problem sits in the booking flow, and no operational layer will fix it.

Get an exact quote for your property

Email your room count, the PMS you run, and the modules you want (housekeeping, maintenance, Guest Connect, AI agents) to hello@getsweeply.com. You will get back a per-room monthly price for each module, the one-off onboarding fee per property, and confirmation of whether your PMS exposes reservation comments to the agent today. Sweeply prices per sellable room, per month, per module, and non-sellable space is free.