Somebody should really check that. Now somebody does.
Every property has a list of things someone ought to check. Is the deep clean done before the VIP arrives? Is the balance paid? Nobody gets to it until the month it costs you. An Apex agent goes on that list for good.
Animation: the Apex agents dashboard over an afternoon. Arrival Readiness flags a cabin whose balance is unpaid, Late Task Escalation texts the housekeeping lead about an overdue deep clean, Maintenance Follow-Up drafts a guest message that waits for approval, and a custom Cleaning Fee Review agent flags two reservations missing a cleaning fee. A manager approves the guest message and it is sent.
A missed cleaning fee is a rounding error on one reservation and a real number by the end of the season. Nobody can review every booking every day. An Apex agent reviews them on a schedule and flags the ones where the fee never landed.
Animation: a Cleaning Fee Review agent scans six upcoming reservations one by one and flags the two with no cleaning fee for the team.
A general manager checks their phone at nine at night because they are the backup for everything that might not have happened. Tell an agent to flag the deep clean on Unit 7 when it runs late. It holds that watch.
Animation over a photo of a housekeeper at work: the deep clean on Unit 7 passes its 3:00 pm due time, Late Task Escalation notices, and texts Maria, the housekeeping lead.
An agent that acts across your property is not something everyone should be able to create. In ApexOS, building agents is a permission you grant. And wherever money moves or a guest gets messaged, an agent stops and asks.
Animation: Maintenance Follow-Up proposes a text to a guest after a repair, with its reasoning. Because it messages a guest it waits for approval, until a manager approves it and it is sent.
Most automation means a rules builder, so most operations have three automations and a backlog. In ApexOS you describe what to watch and what should happen, review the agent Apex proposes, and save it.
Animation: an operator types a plain-language description of an agent, presses Propose an agent, reviews the proposed name, goal, mode and scheduled trigger, and creates it.
Watches today’s and tomorrow’s confirmed arrivals and flags the ones that aren’t ready: unit not clean, open maintenance, unpaid balance, unsigned agreement. Before the guest finds out.
When a task runs late, the agent tells the people you name, in-app, by email or by text. The phone alarm and the sticky note can retire.
In Watch mode an agent only records what it would do. Assist handles low-risk internal work and asks about the rest. Autopilot handles more, and still asks about anything high risk.
Anywhere money moves or a guest gets messaged, an agent always stops and asks. You can add that gate to any other skill. You can tighten the rules. Nobody can loosen them.
Every action records the reasoning behind it. When an agent gets something wrong, you can see why it decided what it decided. That’s the difference between fixing it and distrusting it.
Start from a preset for arrival readiness, maintenance follow-up or late tasks, and copy it into an agent you can edit. Or describe a new one from scratch.
Agents search your knowledgebase of policies and SOPs, and read the guest memories on file, before they recommend anything. So they follow your procedures, not a guess.
Arrival readiness
Anything that messages a guest or moves money waits for a person, in every mode. Approve it, edit it or reject it with a note. Every run lands in the log.
See how Skills workAn AI agent in a property management system watches for a condition in your operation, such as an arrival that isn’t ready or a task running late, and acts on it inside rules you set. It can recommend an action, carry out an approved one, or escalate to a person. The useful test is whether it does work or only sends you a notification about work.
Common jobs for Apex agents: flag arrivals that aren’t ready, such as a unit not clean or a balance unpaid; assign open tasks based on who is on shift and their workload; decide whether a guest should hear back after maintenance is done; escalate late tasks to the people you name; review reservations on a schedule and flag exceptions; and stop and ask before money moves or a guest gets messaged.
Not without someone approving it where it matters. Anything that contacts a guest, moves money or is marked restricted always waits for a person, in every mode, and you can add that gate to any other skill. Every action is logged with the reasoning behind it, and building agents is a permission you grant.
No. You describe what should be watched and what should happen, review the agent Apex proposes, and save it. What you need is someone who knows your procedures, which is why building agents is a permission you grant rather than a technical skill.
A skill is one action an agent can take, chosen from a catalog and tuned with your own guidance. An agent decides when to use it, following the procedures in your knowledgebase. You describe the outcome you want, and the agent picks the skills.
