How hotel work will change as software tools become AI teammates

Having a human at the helm, not just in the loop, is key to setting AI agents' goals, limits, and escalation triggers, and then to being able to watch the audit trail for anomalies.

First, delegate commercial tasks that already operate on written rules, reserving human judgment for those that don't.

These tasks go first because the limits are already written. Every commercial team has a rate band, a response-time standard, and an upsell price list. An AI agent inherits those limits rather than setting its own. They go first because errors are detectable, exposure is limited, and actions can be corrected. A misrouted lead can be reassigned; a mispriced contract or mishandled key account may be harder to recover.

The human belongs at the helm. Before the agent runs, establish its goals, its boundaries, its escalation triggers, and its accountable owner. Audit trails are deeply important to monitor performance and exceptions. Start with a thorough human review and gradually decrease intervention as reliability is demonstrated. Keep humans in the loop whenever exceptions, commercial exposure, or guest impact exceed the agent’s authority.

Hotels have increasingly adopted AI for data analysis and recommendations, yet decision-making and execution remain largely manual. The next leap is moving from AI-powered insights to autonomous action, with humans focusing on governance, strategy and exceptions.

In Revenue Management, agents can monitor demand and adjust rates, inventory and restrictions autonomously. Individual bookings suit automation; complex group business still requires human judgment.

In Marketing, advertising platforms already automate bidding and optimization across placements. Hotels can delegate campaign execution and budget allocation while retaining strategic control, especially over upper-funnel investment.

In Customer Service, multilingual, 24/7 agents can answer questions and manage quotations, bookings, changes and cancellations. Humans step in when guests request them or exceptions arise.

In Knowledge & Content Operations, AI can consolidate fragmented hotel information into a canonical source of truth, continuously maintaining it without burdening hotel staff. It identifies knowledge gaps and asks employees only for missing answers or necessary validation.

The biggest opportunity lies in connecting these agents. Falling demand could instantly trigger targeted marketing; high occupancy could reduce unnecessary ad spend. Shared information also enables real-time, cross-functional reporting for operational managers and executives.

This is coordinated, autonomous action that grows revenue, reduces costs and shifts people from repetitive execution to governance.

We've seen that significant revenue opportunities, both from groups and leisure guests, exist inside reservation teams' Outlook inboxes, but they often lack the time and capacity to systematically manage them.

I think one of the first things hotels should delegate to AI agents is managing these opportunities from the initial inquiry all the way to a booking, although that is very hard to do correctly and responsibly.

AI agents should be able to understand guest requests, check availability and rates, prepare relevant offers, and proactively follow up if the guest, travel agent, or whoever made the inquiry doesn't respond. And they should be able to take action in hotel systems rather than just suggest what an employee should do next.

Humans should remain involved where judgment, empathy, negotiation, or commercial discretion matter.

I don't think success should be measured by how many emails are answered, but by how much direct demand actually turns into revenue, and how much better the guest experience becomes.

Your revenue manager probably spent the first hour of today checking pace, scanning the comp set and nudging rates. Tomorrow they’ll do it again. That's not why anyone joined hospitality.

Pricing is the natural place to start. An agent can read demand and competitor rates around the clock and adjust prices at a pace no team can match by hand, within the limits your team sets.

Guest requests that carry revenue come next. A late checkout means checking availability, taking payment, updating the booking and telling housekeeping. An agent can run all of it, from the guest’s message to the finished task. Back office work follows, from invoicing to matching payments to the ledger.

My rule of thumb is simple. Delegate the work that repeats every day and follows clear rules. Keep people on decisions that call for judgment, local knowledge or a relationship. When an agent can't resolve a request, it hands it to a person.

Treat agents like a new team member. Teach them, review their work and give them more responsibility as trust grows. Every task an agent takes on gives your team more time with guests.

Unless the plan is to be a capsule hotel, a human must always remain in the loop.

Not because the human can perform the role or task more effectively. In some cases, they cannot. But because the humans translate the cold hard data and process into the human experience. They decode this to a fellow human.

As we experience more agents in the workflow of hospitality, the emerging challenge will be the output and outcomes of the data processes, and how to grasp and manage the information for a better overall hospitality product and service.

Ironically, today we often claim we don't have enough information. We also claim we have too much information to leverage.

In the future I foresee we will have an even greater level of information to interpret and leverage, but we will not be able to maintain the pace. The human in the loop may very well end up being the human roadblock in the loop. Not able to process the presented information swiftly enough to allow other automation to continue.

Put simply, let the technology handle things that a human brings little value to. We already do this today, but the efficiencies will be even greater as we move forward.

Anything data-heavy and pattern-driven, and expected to produce an utility-driven value, is increasingly likely to be outsourced to AI due to better economics. For a hotel, such utility functions can include: (a) agentic booking (b) revenue management (demand forecast and pricing) (c) digital concierge.

But on the other side, human connection is becoming a LUXURY, not commodity. Such functions include:

(a) Group sales: Agentic AI’s efficiency in driving transient demand is allowing hotels to reallocate resources toward higher-value, relationship-led group business.

(b) Revenue management (channel-mix decision): Hotels have both profitability vs liquidity, return vs risk, and long-term vs short-term considerations that go deeper than numbers on paper, especially when the risk, leverage, and goodwill (brand) sides of the story aren't reflected on P&L.

(c) Tour operators: TO is for more complex trips that require dynamic, "in the weeds" operational coordination, and assume operational responsibility or risks.

(d) Guest experience: F&B, Front desk, human concierge, etc. AI won’t reduce, and may even elevate, customer expectations for human-led touchpoints in any guest-facing functionality that goes beyond pure utility and delivers rich meaning, emotion, and trust. Oh, and guest feedback management (responding to reviews) belongs to such functions too.

Start with the work nobody enjoys. Every commercial team has tasks that eat hours and need little to no judgement. Pulling last week’s numbers into a report, running the same forecast scenarios, summarising information to share. AI can do all of that continuously and have it ready before anyone logs in. That’s the first thing to delegate, because it’s low risk and the time saved is immediate.

The second layer is monitoring and analysis. Let’s say demand shifts, a competitor drops rate, a group enquiry comes in that may displace transient business. AI should watch for these around the clock, analyze the scenario and bring a recommendation, because it can take in vastly more information than a person scanning a dashboard.

Where the human stays is on decisions that impact performance. In one word: trade-offs. Do I take the group booking or hold the rooms? The AI’s job is to give them the right information to make the best possible call, and the responsibility stays with the team.

Hand over the repetitive work today, move to insights and recommendations, then hand over actions only once the recommendations have earned your trust.

I think a lot of effort will go into automating frequent commercial tasks. The knee-jerk reaction will be to cut costs, understandably: savings drop straight to the bottom line. But expenses have a floor. Revenue has no equivalent ceiling, and growth usually comes from making guests happier or winning new ones.

How to do that? Personal agents will increasingly represent travelers’ interests. Hotels can respond with more compelling offers grounded in detailed product knowledge and reliable execution. A quiet room on a high floor, ready by noon with my loyalty entitlements delivered, beats another generic deluxe-room description, provided the hotel can deliver it.

That takes agents coordinating across the entire workflow: understanding the request, constructing the offer, allocating the room, coordinating housekeeping, and verifying fulfillment. Otherwise, we’ve just taught the departmental silos to type faster.

Humans should own the objectives, policies, and difficult exceptions. Agents should execute within those boundaries, supported by a shared, current picture of the property, audit trails, and controlled overrides. The aim should be to create more value for guests and capture a fair share of it, rather than simply operate the same hotel with lower costs and fewer people.

Hotel work won’t change by replacing human judgement. It will shift where that judgement is applied.

Today, commercial teams lose hours pulling pickup reports, monitoring parity and making individual price tweaks. As AI takes on more of that, the job moves from approving tasks to setting direction.

The smartest approach is delegating the watching while keeping the deciding. One revenue team we work with sells late, so every rate change waiting for approval meant bookings coming in at the old price. Their plan is to approve every change for a few weeks, then let the agent act inside a floor and ceiling and report back what moved.

That’s how human-in-the-loop grows. Instead of approving every click, hoteliers stay in control by setting the boundaries and handling what the system can’t see, like local relationships or a promised group block, and putting the time they get back into strategy and guests.

The human still owns the outcome. AI just gives them the time to act on it.

Hotel work becomes conversational. Guests will talk to AI, and so will staff.

Talking to AI becomes practical for staff the moment a hotel connects its systems of record to the AI model of its choice, with permission to read and to act. Then a front office manager can say: “Show me today’s arrivals, best guests first, and block our best rooms in that order.” A marketer can say: “Send a WhatsApp to every guest ten minutes after check-in.” A general manager can say: “Respond to all new reviews and tell me which team should improve on which topic.”

The value grows with every function the AI can reach. One instruction that spans reservations, messaging and reputation does what no single tool could.

Where to start? Delegate the reading first: monitoring pickup, pace and reviews, preparing reports and spotting patterns. Then hand over routine actions: review responses, pre-arrival and in-stay messages, room assignments. For anything that acts, keep a human in the loop, approving every action at first and later spot-checking. Price changes, offers to top guests and anything contractual should stay with people longest.

Automation should follow trust, not the other way round. Every reliable output earns the next level of autonomy.