A guest asks for the wifi password at 11pm. Another wants to know if the building has a lift. A third says the shower isn't draining. On a portfolio of 30 properties, that's a normal evening, and most of it can be handled without anyone reaching for their phone.
That's the job of an AI autopilot. It sits on top of your property management system, reads every incoming guest message, and deals with the routine ones from start to finish. The ones that need a person go to a person, with the context already attached.
Auto-reply, drafts and autopilot
You'll see three kinds of AI in short-term rental software, and they're easy to confuse.
An auto-reply answers simple, repeated questions on its own. Check-in time, parking, wifi. Most PMS platforms now include one (Guesty has ReplyAI, Hostaway has Auto-Reply), and it's a sensible first step.
Draft replies write a suggested answer for every message, and someone on your team reads it, edits it and presses send. That saves typing, though a person still opens every conversation.
An autopilot takes the whole routine operation. It replies, but it also acts on what the guest said. A blocked drain becomes a maintenance task with the address, the access details and the guest's message. A request for a 1pm arrival gets checked against the turnover before anyone promises it. A guest asking about one more night gets an answer that reflects your calendar and your pricing.
The difference matters most at volume. A typical property gets somewhere between 10 and 20 guest messages a week. At five properties that's comfortable. At 40 it's a part-time job, and most of it is the same questions on repeat.
You'll also hear autopilots described as agentic AI or AI agents. The terms overlap. "Autopilot" is the outcome, routine guest operations running without a person. "Agentic" is how it gets there: an AI that reads the context, decides which of your policies applies, and takes actions across your PMS and your team rather than producing a single reply. We've broken that down step by step in agentic AI for short-term rental operations.
What it handles on its own
A well-configured autopilot covers the questions that have a right answer somewhere in your own information. House rules, listing details, arrival instructions, local recommendations you've written down, your policies on pets or parties or luggage drop-off.
It also covers the small operational chains that hang off those messages. Briefing housekeeping that a cot is needed. Telling the guest the cleaner's running late and their new arrival time. Logging the broken kettle so it's replaced before the next stay. These are the jobs that quietly eat a team's day, and they're where an autopilot earns its keep over an auto-reply. The more capable systems now go further into operations, re-planning staff days when a cleaner calls in sick and scoring every stay so you know which property or process to fix first.
Language is rarely a barrier. Modern systems understand and reply in any language a guest writes in, including regional dialects, so the Italian family and the Korean couple get the same quality of answer as everyone else.
What should stay with your team
Some conversations shouldn't be automated, and a good autopilot knows which ones.
A guest whose flight was cancelled and who wants a refund. A neighbour complaining about noise. A guest who's upset, or unwell, or locked out at 2am in the rain. These need judgement, and often a bit of warmth that's hard to fake. The autopilot's job here is to recognise the situation quickly and hand it over cleanly, with the conversation summarised so your team isn't starting from scratch.
That's also why no serious tool promises 100%. One wrong promise in a hundred replies, like a late checkout agreed when the turnaround is tight, costs more than the other 99 save. Across a portfolio of a few hundred properties, one in a hundred is dozens of guests a week. So the sensible target is around 70 to 80% of guest operations on full autopilot, with refunds, complaints, upset guests and anything outside your rules going to a person. Escalating correctly matters more than replying.
The aim is to take the routine weight off your team so they've got time and patience for the moments that shape a guest's stay (and their review).
How to measure it
Vendors quote automation rates, and the numbers aren't always comparable. Before you trust one, ask what's being counted.
The useful definition is the share of guest messages and tasks the system handles completely on its own, where a person would otherwise have done the work. Drafts that someone reviews and sends shouldn't count, because a person still did the work. Neither should canned messages your PMS was already sending on a schedule.
It's also worth asking whether the figure is an average across all customers or a best case. A rate that holds across a range of portfolios, from city apartments to rural cabins, tells you far more than a headline figure from one account.
Keeping it up to date
An autopilot is only as good as what it knows about your properties, and that changes all the time: door codes, parking, cleaners, building works, seasonal rules. With a handful of properties, someone can update the AI by hand. With hundreds, nobody can keep up, and the AI starts giving out-of-date answers with full confidence.
That's why the better systems learn on their own. They build their knowledge from your past conversations, save the answer when they have to ask your team something, and remember corrections, so the same gap doesn't come back. We've covered what self-learning should mean, its risks and how to test it.
Where it fits with Guesty and Hostaway
An autopilot doesn't replace your PMS. It connects to it, usually through the PMS marketplace, reads the reservations and conversations, and writes its replies and changes back. Your team keeps working in the same inbox they already know. Nothing about your channel connections or your calendar changes.
That makes the choice less dramatic than it sounds. You can start with the AI your PMS includes, see where it runs out, and add an autopilot when the routine work is still taking more of your team's time than it should. We've written separate guides for Guesty and Hostaway, and a list of questions to ask before choosing a tool.
One example of an autopilot built this way is Aeve, which runs on both Guesty and Hostaway and handles 70 to 80% of guest communication across its customers. It's deliberately not 100%. The rest goes to people, which is how it should be.