Take a limited-service restaurant turning over $1.5 million a year. In 2019 an operation that size kept roughly $75,000 of it. To break even on the same four walls today it has to push about $1.93 million through them. The old arithmetic at least fits in an operator’s head: a third of every sales dollar to food, a third to labor, most of the rest to occupancy and supplies, and about a nickel left at the bottom. Then costs climbed and the sales dollar did not.[2]
Payroll is where the gap opens widest. Limited-service operators who lost money in fiscal 2024 spent 34.1 % of sales on wages and benefits. The ones who turned a profit spent 30.0 %.[3] Four cents in every dollar separates them, and on a business that keeps a nickel, four cents is nearly the whole thing. Almost nothing that goes wrong in a restaurant today gets fixed by hiring somebody to fix it.
A thin margin is also an easy one to move. When a restaurant keeps five cents on the dollar, finding one more cent does not make it 1 % better off. It makes it 20 % more profitable. Two cents, and that becomes 40 %. Most of the software a restaurant pays for every month does not add a cent to that margin.
Restaurants are not short of software. They are short of anyone to act on it.
A single restaurant already pays for a point-of-sale system, scheduling, payroll, inventory, online ordering, reservations and accounting. All of them are good at telling the operator what happened. None of them does anything about it. The unanswered catering inquiry shows up in a report. So does the ad spend that brought in nothing, and the entrée that loses money on every ticket. Somebody still has to fix each one, and in most restaurants that somebody was let go years ago and never replaced.
Kutlerri.ai is built for the part that comes after the report. It builds one AI agent per problem, runs it on the restaurant’s behalf, and draws its fee from what the agent produces rather than a monthly license: a %age of the revenue it creates, or of the cost it strips out.
Houston-based Narahari Investments has put $4 million behind that model. The round scales the Catering Sales Agent and the agents behind it. Every agent is delivered as a managed service today, with a mobile-first app in development for operators who would rather run them in-house.
“Every restaurant software company sells the same two things: a tool, and the assumption that somebody inside will use it,” said Trisha Narahari, Managing Director at Narahari Investments. “That somebody left the industry years ago. Kutlerri.ai sends the worker, not just the tool, and takes its fee out of what that worker brings in. They ran it that way on a real restaurant’s books for 10 months before selling it to anybody. That is what we backed.”
“An agent has to earn its own line on the P&L, or it does not ship,” said Payal Tripathy, Co-Founder & COO of Kutlerri.ai. “That is the rule we build to, and our fee comes out of the line it earns. The bar we hold ourselves to is 20 % more revenue and 10 % less cost. No single agent gets a restaurant there, which is exactly why there is a team of them and why they go out one at a time.”
Why the first agent was built to sell catering
Catering is the largest single order a restaurant writes, and the one with the least selling behind it. Smaller operators have nobody assigned to it. Larger brands have a catering team that spends its day quoting, confirming and servicing the orders already coming in. In both cases the same work goes undone: going out and finding new business. The kitchen can already make the food. Finding the buyer is sales work, not kitchen work.
It is also unusually winnable, because catering demand does not spread evenly. Across more than 8,000 locations and 2 million catering orders, 5 % of customers placed 60 % of the orders.[4] A restaurant does not need a thousand catering accounts. It needs the few dozen offices, schools and job sites in its trade area that order every week. Most restaurants never get to that list, and route catering through third-party marketplaces that publish commissions of 15 to 30 % and keep the customer.[5]
The catering products on the market each take one half of that job. Some help a restaurant handle the inquiries that already come in, so fewer go unanswered. Others send campaigns to customers it has already served. Both only work with people who already know the restaurant exists. The other half is going out and finding the companies that have never ordered, and that has always needed a salesperson.
Kutlerri.ai’s agent is built for both halves. On the way in, it picks up the phone, the chat, the email, the web form and the text, carries the back and forth, sends the proposal and closes the order. On the way out, it builds the trade area from scratch, identifying catering-fit businesses within 10 to 30 miles and finding the decision-maker at each, then working them through multi-touch sequences over email, text and phone, alongside accounts quiet for 90 days. That is demand the restaurant never had. Only decisions come back for approval, and the customer stays with the restaurant.
Kutlerri.ai proved that inside one restaurant group before selling it anywhere else. RASA, a five-location fast-casual group in the DMV, took the design-partner seat and paid commercial rates from month one rather than taking a free trial. Across 10 months, and still running, the agent has booked RASA $450,000 of new catering revenue, better than 10 times what the group has spent, with no one added to payroll.
“Catering was revenue sitting right in front of us that we could never get to,” said Trey Moats, CEO of RASA. “Quoting, chasing and following up is a full-time job, and nobody running five restaurants has a spare one. Kutlerri.ai took the entire channel, and my team never had to leave the kitchen to do it.”
The agents are split the way the P&L is
Every agent sits on one side of the P&L or the other. One group brings revenue in. The other takes cost out. The Catering Sales Agent is the only one on general release, three more are in beta inside working restaurants, and two are in build.
Revenue expansion. The Catering Sales Agent is live and in market. The Third-Party Sales Agent is in beta, growing what a restaurant earns on the delivery platforms. The Online Sales Agent is in build, aimed at the restaurant’s own online ordering revenue.
Cost control. The Third-Party Marketing Spend Optimization Agent is in beta, cutting the advertising spend behind that delivery revenue. The Prime Cost Optimization Agent is in beta, working the two biggest lines on any restaurant P&L, food and labor. The Daily Prep and Waste Control Agent is in build, matching each day’s prep to demand so less food goes to waste.
Why Kutlerri.ai does the work instead of selling the tool
It would be easier to sell the software and let the restaurant run it. Kutlerri.ai runs it instead, for two reasons.
The first is that the agent keeps getting better. It sells catering for many restaurants at once, so it learns which businesses order, what a winning quote looks like and which follow-up gets a reply. A restaurant running software by itself only learns from its own results. An agent Kutlerri.ai runs arrives at the next restaurant already knowing what worked at the last one.
The second reason is how most software vendors get paid. A vendor that hands over a login has no control over what happens next, so it charges a monthly fee and hopes for the best. Kutlerri.ai does the work itself, so it can be paid on the outcome instead: a share of the revenue an agent brings in, or a share of the cost it takes out. If the agent delivers nothing, the restaurant pays nothing.
Kutlerri.ai serves limited-service and full-service operators nationwide. Access requests are open at kutlerri.ai.
