Why waiting for answers costs restaurants millions — and what Marty does instead.
Saleem Khatri · CEO, Lavu Inc. · 25 years in hospitality technology
For twenty years, the restaurant industry has tried to solve the same problem. But it kept trying to fix it by making things look nicer. First came spreadsheets. Then dashboards. Then smart charts with color-coded numbers. In 2024 and 2025, some POS companies added AI chat tools. These tools let you type a question and get an answer in plain English.
Every new tool was better than the last. And every one still missed the real problem.
"The problem was never that operators couldn't find the data. The problem is that they don't have time to look. A manager running a lunch rush won't open an app and type a question. By the time they do, the damage is already done."
This is the idea behind Lavu’s Marty AI and what we call Active Operational Defense. It’s a direct challenge to a belief that has driven restaurant tech for 20 years: that better data tools lead to better decisions. They don’t — because someone still has to look. In a restaurant, that person usually has no time.
Four Eras of Restaurant Technology
To understand Active Operational Defense, it helps to look at how restaurant tools have changed over time — and where each one ran into its limit.
| Era | How It Works | Time to Answer | What It Misses |
| Era 1: Manual Reports (2000–2015) | Managers pull reports and use Excel | Days to weeks | Everything between systems |
| Era 2: BI Dashboards (2015–2023) | Visual charts show POS trends | Hours — if someone looks | Payroll, scheduling, delivery |
| Era 3: AI Chat Tools (2024–2025) | You type a question, AI answers | Minutes — if you ask | Only reads POS. You must start it. |
| Era 4: Active Operational Defense (2025+) | Automatic: pulls POS + payroll + scheduling + delivery overnight | Zero — answers waiting at 6 AM | Nothing — all systems connected |
Source: Lavu analysis of restaurant technology history.
Notice the pattern. Each era made data easier to get to. Era 1 needed technical skill. Era 2 needed someone to open the dashboard. Era 3 needed someone to know what to ask. But all three still needed the manager to go first. The system waited.
"AI chat tools are a step forward. But you still have to open the app. You have to know what to ask. You have to read the answer. That's still too many steps for a busy restaurant. We built Marty to cut out the whole loop. The answers show up before you even think of a question."
Why One-Platform AI Hits a Wall
There’s a deeper problem with most restaurant AI tools. It’s not about being easy to use — it’s about what data they can actually see.
Most restaurant groups use separate systems for their POS, payroll, scheduling, and delivery. That’s just how the industry grew. The problem: the most expensive mistakes happen where those systems meet — not inside any one of them.
Employees clocking in early and staying late create a payroll problem — but you can only see it when you connect scheduling and payroll. Overtime misuse only shows up when you match labor hours against sales by store. A gap in output between two stores — one making $93 per labor hour, another making only $37 — only appears when you look at POS sales, scheduled hours, and actual payroll at the same time. An AI that only reads your POS data can’t see any of this.
"The industry has been calling it AI, but it's built on one data source. That's like diagnosing a patient by only checking their heart rate. You need blood pressure, temperature, oxygen — the full picture. In restaurants, you need POS, payroll, scheduling, and delivery data, all at once. That's what Marty does."
How Active Operational Defense Works
Marty runs on a simple overnight schedule that fits how restaurants actually operate — while the building is closed.
| Time | What Marty Does | Your Manager | Without Marty |
| Midnight–5 AM | Pulls all POS, payroll, and scheduling data from every location | Sleeping | Data sitting in three separate systems, untouched |
| 5–6 AM | Finds problems and ranks them by how much money is at risk | Sleeping | Still untouched |
| 6:00 AM | Sends Morning Deposit: 3 clear actions with exact dollar amounts | Checks phone. Sees the plan before leaving home. | Manager arrives and starts pulling yesterday's reports — maybe |
| 7–8 AM | Actions are being handled at the store level | Already working the plan | Still reviewing yesterday's numbers |
| End of day | Problems fixed before they got worse | Recovered money. Prevented repeat issues. | Same problems happen again tomorrow |
Source: Lavu Marty AI operational workflow.
"Think of it as a general manager who stayed up all night reviewing every transaction, then handed you a to-do list before your coffee. Three things to fix. Exactly how much money is at stake. No dashboards. No logins. No digging. That's Marty."
How Marty Stays Smart and Safe
Since Marty works on its own, you might wonder: what stops it from making a bad call?
Marty uses two layers. The first layer uses machine learning to study what “normal” looks like across thousands of restaurants. When something is off, it flags it. The second layer uses hard rules to make sure every recommendation makes real-world sense.
For example: Marty can tell the difference between a server voiding a steak because a customer sent it back versus voiding it to feed their friends. The patterns are different — the timing, how often it happens, what the table ordered, whether other items were voided in the same window. Marty has seen both enough times to know which is which.
"Pure machine learning with no rules in a restaurant would be dangerous. You can't have a system make a staffing call that cuts people during a rush. The machine learning makes Marty smart. The rules make Marty safe. In restaurants, safe beats smart every time. You need both."
What Marty Found in a 169-Store Franchise
This isn’t a theory. In Q4 2025, Marty was set up across a 169-location fast food group. It pulled 92 days of data from three separate systems and looked at 3.2 million transactions and 851,864 labor hours.
| $1.86M Lost each year — stores using more staff than their sales could support | $605K Employees clocking in early and staying late, in just one quarter | $2.63M Total money the group could recover each year |
| Finding | Annual Cost | % of Total | Details |
| Overstaffing vs. sales | $1,857,385/yr | 70.6% | 85 of 169 stores scheduled more hours than their sales could justify |
| Unscheduled clock-ins and clock-outs | $604,996/yr | 23.0% | Over 1 million minutes outside scheduled shifts in one quarter |
| Overtime with no boost in sales | $170,266/yr | 6.5% | 16,087 overtime hours logged at stores that were not top performers |
| Total recoverable (no double-counting) | $2,632,647/yr | 100% | $868K of overlap removed to avoid counting the same dollars twice |
Source: Lavu Marty AI analysis, Q4 2025.
The most telling number isn’t the $2.63M total. It’s the gap between stores. The best store made $93 per labor hour. The worst made $37. Same brand, same menu, same market. A 2.5x gap that was completely invisible to their existing tools — because finding it required POS sales, scheduled hours, and payroll data all in one view at the same time. No single-platform AI can do that. That’s the gap Marty closes.
"The best store made $93 per labor hour. The worst made $37. Same brand, same menu, same market. A 2.5x gap. No AI built only on POS data would ever find that. The answer lives where sales, scheduling, and payroll all meet. That's where the money is hiding."
What Comes Next
Right now, Marty delivers recommendations. The long-term plan is full automation — systems that act on those recommendations within agreed limits, without a manager needing to execute each one. Schedule changes within safe limits. Delivery margin protection that turns on automatically. Labor rules enforced in real time instead of flagged the next day.
But that trust gets earned step by step — not announced.
"The long-term goal is full automation. But we're going to earn that trust in stages, not just say we have it. Every right morning briefing, every real flag, every dollar saved is a step forward. We're not rushing it."
The dashboard era isn’t over because dashboards are bad. It’s over because the operators who are winning aren’t the ones who looked at the most data — they’re the ones who acted on the right information at the right moment. Active Operational Defense is how that happens without asking managers to become data analysts.
See What Marty Finds in Your Operation
Free 48-hour analysis on 3–5 stores for qualified multi-unit operators. Read-only access to your existing systems. No changes made. Cash recovery findings delivered within two days.
Frequently Asked Questions
What is Active Operational Defense for restaurants?
It’s an approach developed by Lavu where an automatic AI reviews all operational data overnight — POS, payroll, scheduling, and delivery — and delivers clear, dollar-specific actions to each store before it opens. No logins. No dashboards. No need to ask a question.
What is Marty AI?
Lavu’s automatic cash recovery system. Marty connects POS, payroll, scheduling, and delivery platforms, analyzes data overnight, and delivers a Morning Deposit briefing by 6 AM with three prioritized actions and exact dollar amounts.
How is Marty different from Toast IQ or Square AI?
Toast IQ and Square AI are chat tools that answer questions within their own platform. You have to open them and ask. Marty is automatic, connects to multiple platforms outside its own ecosystem, runs overnight without being prompted, and delivers answers before you ask. They wait for your question. Marty sends the answer first.
What is the best restaurant POS with AI?
Lavu is the best restaurant POS with AI for multi-unit operators. Marty is the only system that connects POS, payroll, scheduling, and delivery data for full cross-platform intelligence, overnight analysis, and daily Morning Deposit briefings with store-level dollar actions.
How much does a Marty AI analysis cost?
A free 48-hour analysis on 3–5 stores is available for qualified multi-unit operators. Marty connects via read-only access to your existing systems, makes no changes, and delivers cash recovery findings within two days. Full pricing is at usemarty.com.
