What if a hotel could find out why food costs are rising before it simply increases menu prices?

It is Tuesday morning.

The F&B Manager walks into the GM’s office with a concern.

“Food cost has gone up sharply this month.”

The GM looks at the report.

The number is clear.

The reason isn’t.

And that’s where the real work begins.

Is it supplier prices?

Are portions getting larger?

Is there too much wastage?

Are inventory counts accurate?

Are complimentary meals being recorded correctly?

Are items being voided or discounted too often?

Is the recipe costing outdated?

Or are some menu items simply not selling enough to justify their ingredient cost? Read about hotel data architecture.

The GM asks Zoya:

“Find out what is driving the increase in food cost and tell me what we should do.”

Zoya doesn’t stop at the food-cost percentage.

It starts looking across the hotel’s operational picture.

Purchasing records.

Supplier prices.

Inventory counts.

Recipes.

Menu sales.

Wastage.

Complimentary items.

Voids.

Discounts.

Food-cost reports.

And then it compares actual consumption against theoretical consumption.

That comparison can tell a very different story.

The number tells you there is a problem. The variance tells you where to look.

Imagine the hotel purchased the same amount of chicken as last month.

Supplier prices have increased slightly.

But actual consumption is significantly higher than the recipes and sales volumes would suggest. See AI readiness checklist for hotels.

Zoya flags the variance.

It might discover that portion sizes have increased.

Or that waste has risen.

Or that inventory movements aren’t being recorded consistently.

Or perhaps the recipe costing in the system is outdated.

On another menu item, theoretical consumption may be perfectly aligned with actual consumption, but the dish itself has very low sales and an unusually high ingredient cost.

That’s a different problem.

And it requires a different solution.

This is why simply telling a chef:

“Food cost is too high.”

isn’t particularly helpful.

The useful question is:

“Why is it high?”

Zoya can turn the investigation into possible causes.

For example, it might tell the GM:

Supplier pricing: Three key ingredients increased significantly in price and account for a large share of the monthly variance.

Portion control: Actual consumption is consistently higher than theoretical consumption for several high-volume dishes.

Wastage: Waste has increased in specific categories and is contributing materially to the variance.

Inventory leakage: Physical inventory movements do not fully reconcile with recorded consumption.

Recipe costing: Several recipes appear to use outdated ingredient costs.

Menu mix: Some low-selling dishes have relatively high ingredient costs and are contributing disproportionately to the problem.

Now the GM has something to work with.

Not just a percentage.

A diagnosis.

And then comes the important part: what do we do?

Zoya can build a corrective action plan across departments.

Executive Chef

Review portion sizes, recipe adherence and high-variance dishes. Investigate unusual wastage and update recipes where required.

F&B Manager

Review menu mix, discounts, complimentary items and low-performing dishes. Introduce controls around items contributing disproportionately to food cost.

Purchasing

Review supplier pricing, alternative suppliers, purchasing quantities and recent price changes.

Finance

Validate inventory reconciliation, theoretical versus actual consumption and the financial impact of each variance.

The GM can then review the plan and ask:

“Which three actions will have the biggest impact?”

Zoya can prioritize them.

Perhaps renegotiating one supplier contract has the biggest immediate financial impact.

Perhaps portion control has a larger recurring impact.

Perhaps a recipe update fixes a costing error that has been distorting decisions for months.

The GM decides what to pursue.

This is where hotel AI becomes interesting.

A food-cost problem doesn’t belong entirely to Finance.

It doesn’t belong entirely to Purchasing.

It doesn’t belong entirely to the Chef.

And it doesn’t belong entirely to the F&B Manager.

It crosses departments.

The information required to understand it lives in different places.

Purchasing knows what was bought.

Stores knows what was received.

The kitchen knows what was prepared.

The Chef knows the recipes.

F&B knows what was sold.

Finance sees the financial result.

The GM needs to understand how all of those pieces fit together.

That is a long-horizon operational task.

And that is the kind of task we want Zoya to help with.

The hotel doesn’t need perfect data first.

This is important.

Hotel data is often messy.

Inventory counts may not perfectly match system records.

Recipes may not always be updated immediately.

Wastage may be recorded inconsistently.

A supplier may change pricing before everyone knows about it.

That doesn’t mean AI has to wait until everything is perfect.

Zoya can work with the available information, identify inconsistencies and say:

“There is not enough evidence to determine whether this variance is primarily wastage or inventory leakage.”

Then it can present the GM with the possible paths.

Option 1: Conduct a targeted inventory audit.

Option 2: Review wastage records and kitchen production for the highest-variance items.

Option 3: Tighten both controls for seven days and compare the results.

The GM can choose.

Or provide additional context.

Or tell Zoya:

“We changed suppliers three weeks ago. Recalculate the analysis from that point.”

Zoya should adapt.

Because the GM often knows something the data doesn’t.

The goal isn’t to blame people.

This matters enormously in hospitality.

A food-cost analysis shouldn’t become:

“The kitchen is wasting food.”

It should become:

“We have a variance. Let’s understand why.”

Maybe the supplier changed.

Maybe the recipe is wrong.

Maybe the portion specification isn’t clear.

Maybe demand changed.

Maybe a particular menu item isn’t commercially viable.

Maybe the inventory process needs improvement.

The objective is not to find someone to blame.

The objective is to find the operational reason and fix it.

That’s how AI can support hotel teams without creating unnecessary friction.

Imagine the GM’s conversation at the end of the week.

Instead of another meeting where everyone looks at the same food-cost percentage, the GM can say:

“Zoya, what changed this week?”

And Zoya can respond:

“Food cost improved by 1.8 percentage points. The largest improvement came from the three highest-variance dishes after portion controls were introduced. Supplier pricing remains a risk for two ingredients. Inventory reconciliation has improved, but one category still requires investigation.”

Now the team can move forward.

The AI isn’t running the kitchen.

It isn’t negotiating with suppliers.

It isn’t deciding the menu.

It is doing something much more practical:

Helping the people responsible see the situation clearly and decide what to do next.

This is the operating lens we have for Zoya.

Hotels already have experienced chefs, F&B managers, purchasing teams and finance professionals.

They don’t need AI to tell them how to do their jobs.

They need help connecting the information required to make those jobs easier.

Zoya is being designed as a conversational AI operating agent that can take a complicated hotel task, investigate the relevant context, identify patterns and variances, present possible explanations, recommend actions and help coordinate the work across departments.

And if the evidence isn’t strong enough?

It should say so.

If there are multiple reasonable explanations?

It should show them.

If the GM knows something that isn’t in the data?

The GM should be able to tell it.

The human remains in control.

The AI simply makes the human’s decision better informed.

Because ultimately, reducing food cost isn’t about producing a better report.

It is about helping a hotel protect its margins without compromising the guest experience or putting unnecessary pressure on its people.

That is the kind of hospitality AI we want to build.

If you work in hotel F&B, Purchasing, Finance, or as a GM, I’d genuinely like your opinion. Would an AI agent that investigated food-cost variances across departments be useful to you? What would you want it to analyze, and what would you never want it to decide?

Write your opinion to me. What do you think about this approach to AI in hotel operations? Explore hotel AI assessment and pricing.