AI for operations: how travel and hospitality groups can turn hours into margin

In travel and hospitality, operations are what guests actually buy. Payroll is rising faster than revenue, and much of the coordination behind the service is still manual. Sidney Zeder, AI Strategy Director at Artefact, explains where AI creates value in operations, why the gains come from reshaping whole processes, and how freed hours become margin.

Operations is the product

A hotel group can win the booking and lose the guest at the front desk. A room not ready at 3 pm, an air conditioning unit that breaks down the night the guest arrives, a queue at check-in because two receptionists called in sick: these are the moments guests remember, and most of them can be seen coming in data the hotel already holds. Operations shape the guest experience and carry most of a hotel’s costs. The same decision, which room to clean first or which technician to send, sets how long a guest waits and what the shift costs. AI in operations works on both.

Why the pressure is on operations now

Demand is still growing, but payroll is growing faster than revenue (respectively +5.7% in 2025, against 3.3% in Europe.¹). Labour is also the hardest cost to manage: in North America, 65% of hotels reported staffing shortages in 2025.² When costs grow faster than revenue, growth alone no longer restores the margin.

At the same time, site teams describe the relevant challenge on labour: coordination is still largely manual, with heavy labour required. Housekeeping managers often learn on the morning itself how many rooms they have to turn, room status passes by phone or radio, and gaps in the roster are filled with overtime. The teams work hard, but the gap is not closing.

Three ways AI creates value in operations

Elevate the guest experience

AI works best behind the scenes, so that staff spend more time with guests: a room assigned before arrival according to the guest’s preferences, a failing air conditioning unit detected before check-in, a returning guest recognised with their usual requests ready. The human relationship stays at the centre; AI removes what keeps staff away from it.

Anticipate the unknowns

A storm, a strike, or a trade fair moved by a week changes a hotel’s needs within hours. Models that read bookings, weather, and events adjust rosters, purchases, and prices before the guest feels it.

Simplify the work

Scheduling, room assignment, housekeeping, maintenance dispatch and purchasing repeat every day and follow known rules. More automation of the manual work and synchronization of the tools can ease the operational work.

Where to start: where costs and guest impact meet

We rank processes on their weight in operating costs, how much the guest feels them, how well AI fits the work, and whether the data is ready. In hotels, four site processes come out on top. Being internal, they are also a safer first ground for AI agents.³

Priority process	Cost at stake	What the guest notices	What AI does
1. Workforce scheduling	First cost line	Queues at check-in, slow service at peak times	Forecasts needs by hour, drafts a compliant roster, fills open shifts
2. Housekeeping and task dispatch	The largest department by headcount	A room not ready on arrival	Ranks and assigns cleaning tasks as check-outs and arrivals change
3. F&B procurement and stock	Low profitability on F&B due to waste	Dishes missing from the menu	Sizes orders on forecast covers and sends them for approval
4. Maintenance and repairs	Critical impact on topline and guest experience	Broken air conditioning, a room change at midnight	Triages alerts, dispatches a technician, blocks the room in the PMS

The stakes are concrete. Labour accounts for more than half of operated department expenses in US hotels.⁴ In early hotel deployments, AI has made room preparation about 20% faster and halved food waste within eight months.² In manufacturing, predictive maintenance typically cuts machine downtime by 30 to 50%.⁵

Why we reshape the whole process

Giving employees AI tools for their own tasks is quick and builds familiarity, but the gains stay at the level of isolated hours, and the process remains sequential and driven by people. Redesigning the process end to end changes the flow, the roles, and who decides what. In our estimates, the first approach saves 5 to 15% of a process’s time; the second can save 60 to 90%. McKinsey reaches the same conclusion: the value of agents comes from rethinking how the work is done.⁶

“A tool added to an old process saves a few minutes here and there, too diffuse to show in the P&L. When we redesign the process with the teams who run it, the agent becomes part of how the job is done. That is where the value lies, and why teams keep using it.” – Sidney Zeder, AI Strategy Director, Artefact

Take rostering. Today a manager forecasts needs, builds the roster in a spreadsheet, checks labour rules by hand over several rounds, then fills gaps by phone. Reshaped, a model forecasts demand by hour, an optimiser drafts the roster, a rules engine checks compliance, and an agent offers open shifts to eligible staff. The manager arbitrates the exceptions and handles the conversations with the team.

How we build an agent: more autonomy at each sprint

Take a housekeeping agent, built in four sprints.

It first sees: it alerts the supervisor on rooms at risk for check-in.
It then understands: it drafts the morning room board.
Next it advises: after a late check-out, it proposes a new plan for approval.
Finally, it acts: it re-plans alone within rules set by the site manager, and every action is logged.

Each sprint ends with a live test with the team, and the key signal is what the supervisor actually does with each proposal.

“We never give an agent full autonomy on day one. It earns each level by proving, on live data and in front of the team that uses it, that its proposals are right.” – Bruce Delattre, Director, Data Science Lead, Artefact

What it is worth

Labour is the part of the value we can size with public data:

In Europe, operating costs absorb about two-thirds of hotel revenue: €194 of revenue per available room in 2025, for €69.3 of gross operating profit.¹
Payroll is about half of those costs, at €60.5 per available room.¹
In our estimates, the four priority processes account for around 40% of labour hours, and reshaping them saves 20 to 25% of their time: 8 to 10% of hours are freed.
That is €4.8 to €6.1 per available room: about 4 to 5% less operating cost, and 7 to 9% more gross operating profit.

Rooms back on sale and less waste come on top. But hours only become savings when rosters follow: the general manager or the owner decides whether to redeploy them to guest service or reduce labour cost, starting with overtime and agency hours.

Our convictions

Prioritise on the P&L. Start with the processes that touch the most hours, the largest cost lines and the moments guests notice.
Keep people at the centre of the guest experience. AI belongs in the back office.
Design for teams on the move. Site teams are less digital and rarely sit at a desk: train them, and build the agent into the tools they already use.
Plan for a patchwork of systems. A hotel group can run several PMS, POS, and ERP systems across its hotels, often in legacy versions. Integration belongs in the business case from day one.
Expect most of the effort outside the technology. About a fifth is technology, half is process redesign, and the rest is change management and adoption.