Restaurants Are Finding AI’s Next Job: Running the Business Behind the Kitchen

For years, restaurants have embraced technology to take orders, process payments, manage schedules, and track inventory. Yet despite an abundance of software, many operators still struggle to answer deceptively simple questions: Why did profits fall last week? Where is labor overspending? Is theft occurring? Which location needs attention before tonight’s dinner rush?

A new generation of artificial intelligence platforms is attempting to answer those questions—not by replacing point-of-sale systems or generating reports, but by acting as an operational intelligence layer that sits across every system a restaurant already uses.

Among the companies pushing that shift is restaurant technology provider Lavu, whose AI platform, Marty, is designed to function like an AI general manager for restaurants.

“Most restaurant software tells you what happened,” Khatri said. “Marty tells you what to do next. That’s the difference between information and decisions.”

The distinction reflects a broader evolution in enterprise AI. While early applications focused on generating content or summarizing data, operators across industries are increasingly looking for systems that continuously monitor businesses, identify anomalies, and recommend actions before small problems become expensive ones.

Restaurants may be one of the industries where that evolution has the greatest financial impact.

Margins remain notoriously thin, labor costs continue to rise, and operators often manage dozens of moving parts simultaneously across multiple locations. A few percentage points of excess overtime, unnoticed discounting, inventory shrinkage, or declining guest traffic can quietly erode profitability for months before anyone recognizes a trend.

According to Lavu, one recent analysis across a multi-location restaurant group uncovered approximately $2.9 million in hidden operational exposure. The AI identified roughly $1.24 million tied to labor inefficiencies such as overtime creep and scheduling gaps, nearly $1 million related to pricing, discounts and guest retention, and another $680,000 stemming from operational controls and compliance issues including suspicious void patterns and shared employee logins.

None of those restaurants were considered poorly run.

“They weren’t failing,” Khatri said. “They were simply bleeding money in ways nobody could actually see.”

Rather than replacing existing restaurant technology, Marty connects to systems restaurants already operate—including point-of-sale platforms, payroll software, labor scheduling applications, inventory management systems, delivery services, and spreadsheets—and analyzes more than 50,000 operational signals for each location.

The system first verifies transactional data before using generative AI to translate those findings into recommendations managers can immediately understand.

Instead of presenting another collection of charts, the software delivers alerts such as: labor is running significantly above historical norms for a Tuesday afternoon; a particular employee’s void rate is dramatically higher than peers; or an online ordering platform appears to have stopped receiving orders during peak dinner service.

Each morning, operators receive what the company calls a “Morning Receipt” summarizing where profits were lost, where labor drifted, and which operational issues deserve immediate attention.

The objective, Khatri said, is reducing decision fatigue.

“Restaurant owners don’t wake up wanting another dashboard,” he said. “They want someone to tell them, in under a minute, where the money leaked yesterday and what they should fix before tonight.”

In one high-end steakhouse group operating in the Southeast, the platform analyzed nearly 68,000 guest visits and more than $8 million in sales before identifying an estimated $95,000 to $175,000 in annual recovery opportunities.

The findings included overtime costs exceeding $26,000 annually at one location, a sharp increase in discretionary meal comps, and an unexpected decline in late-night revenue despite guest traffic increasing. In another instance, the software linked declining private event bookings to behavioral changes following an internal management dispute—an operational anomaly that had not been apparent to ownership.

Another two-unit fast casual operator discovered more than $10,000 per month in recoverable value after the system highlighted inconsistent labor scheduling, unusually frequent void transactions from a single employee, recurring overtime tied to extended clock-outs, declining sales of signature menu items, and marketing dollars continuing to flow toward increasingly ineffective social media campaigns.

Perhaps the most striking example came from an international hotel group’s food-and-beverage operations in the Middle East, where the system identified what executives described as years of unnoticed financial leakage—including employee theft schemes, procurement fraud, improperly recorded food costs, and inaccurate revenue reporting—representing an estimated $20,000 in monthly recurring losses.

While each issue individually appeared manageable, collectively they represented what operators often describe as “death by a thousand cuts.”

Beyond financial savings, Khatri believes the larger value proposition is confidence.

“When owners aren’t physically in the restaurant, they’re often relying on instinct,” he said. “Our goal is to replace that uncertainty with evidence.”

That perspective is rooted in personal experience.

Khatri says his father spent years building a successful business before discovering that a trusted accountant had stolen millions of dollars over more than a decade—fraud that escaped detection by banks, accountants, and internal controls.

“My dad never had visibility into what was actually happening,” Khatri said. “A lot of restaurant owners are living that same blind spot today. Marty is the analyst I wish my father had.”

The technology also reflects a larger shift underway in artificial intelligence itself.

Rather than automating isolated tasks, AI is increasingly being deployed as continuous operational oversight—a digital manager that never stops monitoring businesses for emerging risks, changing customer behavior, operational inefficiencies, or financial anomalies.

That trend extends well beyond restaurants. Manufacturers are using AI to monitor production lines. Logistics companies are identifying shipping disruptions before they occur. Financial institutions increasingly rely on machine learning to detect fraud in real time.

Restaurants, however, generate thousands of operational signals every day while often operating on single-digit profit margins, making even modest improvements financially significant.

For operators managing multiple locations, recovering one or two percentage points of labor or food costs can translate into tens—or hundreds—of thousands of dollars annually.

“People assume AI is about replacing employees,” Khatri said. “What we’re seeing is AI making managers dramatically better at their jobs. The best operators still make the decisions. They just finally have something watching everything they can’t.”