By Matt Wampler, CEO of ClearCOGS
A restaurant operator with over two decades in the industry, a technology background layered on top of an operator background, recently described his experience at the National Restaurant Association show after years away.
His takeaway: nothing had changed. Same categories. Same promises. Updated logos.
That observation lands in a specific way in 2026, because it runs against the direction of nearly everything else in restaurant technology. According to Qu’s 2026 Restaurant Technology Benchmark Report, 73 percent of operators are actively investing in AI or planning to start. Margins are compressed. Traffic is down at 57 percent of brands surveyed. Technology is increasingly framed as the path to operational efficiency.
And yet: only 5 percent of operators in the same report say they have seen measurable operational value or guest impact from their AI investments to date.
That gap between intent and results is not a surprise to anyone building technology for restaurant back-of-house operations. It points directly at the problem the operator noticed at the show: the most fundamental back-of-house decision, what to prep, how much of each item, before the shift starts, is still largely made by hand. Not because the technology doesn’t exist. Because the technology hasn’t made it into the kitchen in a form operators can actually use.
What restaurant technology has gotten right
Over the last decade, meaningful progress happened in clearly defined areas.
Point-of-sale systems got faster, more integrated, and more useful as data sources. Online ordering and delivery platforms changed the revenue model for entire segments. Labor scheduling software improved. Payment processing got easier. Customer-facing tools multiplied.
Each of these solved a problem with a clear output that fit into an existing workflow. A better POS terminal processes orders. A scheduling tool builds a labor plan. A loyalty app sends a push notification.
Back-of-house technology, and specifically prep, never fit that pattern cleanly. The output is not an action the software takes. It is a decision a person makes, informed by data that has to come from multiple systems, resolved into a number a kitchen manager will actually trust at 6 in the morning.
That is a harder problem.
Why prep was always technically complex
Automating the prep decision requires solving several distinct problems simultaneously, and all of them have to connect.
Demand forecasting needs historical sales data at the item level, adjusted for day-of-week patterns, weather, local events, and seasonality. Recipe mapping needs to tie each menu item to its ingredients at the ounce, accounting for portioning variability and yield assumptions at each location. Shelf life logic needs to determine not just how much to prep today, but how to build ahead for items that hold across multiple shifts. And the output has to arrive in a format a kitchen manager already trusts, in time to be useful before the shift.
Most restaurant technology platforms have addressed one or two of these layers in isolation. POS systems capture item-level sales history. Inventory platforms manage on-hands and ordering. Recipe management tools store ingredient maps. The value of each layer depends on all of them working together, and that integration has been the persistent gap.
The National Restaurant Association’s 2024 Technology Landscape Report found that food and beverage operators cite back-of-house operational efficiency as a top investment priority. The same report found that three in four operators believe technology gives them a competitive edge, while only 13 percent describe themselves as technology leaders. The intent to solve this is present. The execution gap is real.
What has actually changed
The technical barriers to automated prep forecasting have not disappeared, but they have meaningfully lowered in the last few years.
Machine learning models for time series forecasting have improved in ways that directly affect prep accuracy. Specifically, models can now account for external signals at the location level: weather, local event calendars, school schedules, and channel mix between dine-in and third-party delivery, without requiring extensive manual configuration. That matters because a prep number that is accurate on average but wrong for a specific Tuesday at a specific store does not help. Location-level accuracy, continuously recalibrated from new data, is the technical threshold that actually changes the outcome.
POS integrations have also become more standardized. Toast, Square, NCR Voyix, and other major platforms have opened more data access for third-party systems, which means the historical sales data that forecasting models need is more accessible than it was three or four years ago.
And there is growing recognition that output delivery is as important as model accuracy. A forecast that lives in a dashboard is not a prep decision. A prep decision is a specific number, in the right format, on the right screen or printed sheet, available before the kitchen team starts work. That delivery layer is where most restaurant technology has the most ground still to make up.
Where most restaurant AI investment is actually going
The 5 percent measurable impact figure from Qu’s 2026 benchmark is worth sitting with.
Among the operators surveyed, AI use cases attracting the most attention are customer-facing: voice ordering at the drive-thru, computer vision for kitchen throughput, loyalty personalization. These categories have visible outcomes and direct revenue ties, which makes them easier to fund and evaluate.
Operational AI, including forecasting and prep planning, ranks second at 40 percent of operator focus versus 53 percent for guest growth. That reflects an understandable priority order. Customer-facing AI produces outputs executives can see. Prep forecasting produces outcomes that appear in food cost percentages and waste reduction over time, which is a harder case to make in a budget conversation.
It is also a harder product to build. The Qu report identifies fragmented systems and operational execution as the top barriers to better guest experiences. That same fragmentation is exactly what makes automated prep forecasting difficult: the data needed to build a useful forecast is spread across POS, inventory, recipe, and labor systems that were not designed to work together.
The operators who are seeing results are the ones who have resolved that data fragmentation, and who have chosen vendors that deliver the output inside existing workflows rather than inside new interfaces.
What changes when the prep number actually works
Operators who have moved from manual prep processes to forecast-driven ones tend to describe the change in the same terms: something disappears from the day.
The 30 to 45 minutes a manager spent pulling the PMIX, reconciling it against the prep sheet, calculating on-hands, and second-guessing the output because it didn’t quite feel right: that time comes back. Not as a line item in a productivity report. As actual minutes in the morning before service that a manager can spend with their team.
Waste reduction tends to follow once the numbers are trusted. Over-prepping as a safety behavior mostly exists because managers lack confidence in the forecast. When the number is consistently accurate, the buffer gradually disappears. ReFED estimates that U.S. restaurants lose tens of billions of dollars annually from wasted food that better forecasting and purchasing decisions could prevent, with a meaningful share originating not from spoilage but from prep decisions made without reliable forward-looking data.
Across a multi-unit brand, consistency also becomes possible in a way it wasn’t before. The prep decision no longer depends on who is running the morning shift at each location. That matters in corporate stores and matters even more in franchise networks, where the person making that call each morning may have less brand-specific experience to draw from.
What operators should watch next
The direction of travel in operational technology is clear: predictive systems are moving from dashboards to workflows. A forecast that arrives as a recommendation inside an existing process is categorically different from a forecast that requires a manager to log into another interface and interpret a report. That delivery distinction determines whether the technology actually changes what happens in the kitchen.
The next meaningful shift is likely to come from tighter integration between forecasting outputs and the systems managers already use: inventory on-hands confirmation, holiday or limited-time menu adjustments, and delivery through whatever channel the kitchen team actually checks. That level of integration is not fully operational at scale across most restaurant brands today. But the forecasting model accuracy and data pipeline infrastructure are advancing to where it becomes achievable at the operator level, not just at enterprise chains.
The operator who left the NRA show unchanged may be right that the booths looked the same. The products that will actually change the prep decision are likely not the ones with the largest footprints on the show floor. They are the ones that have solved the last mile: getting a reliable number into the kitchen, before the shift starts, in a format the team will use without being asked to change how they work.
That is a narrower and more specific problem than “AI for restaurants.” It is also the one worth solving first.
At ClearCOGS, prep forecasting is the specific layer we have built toward: turning POS and sales data into daily prep recommendations that reach kitchen teams in the format they already use, before the shift starts. If your back-of-house is still running on manual PMIX pulls and spreadsheets, we would be glad to show you what it looks like when that decision gets made for your team instead of by them. Let’s Talk
Sources
- Qu. 2026 Restaurant Technology Benchmark Report: The Year of Serving Smarter. Business Wire, March 2026. businesswire.com
- National Restaurant Association. Restaurant Technology Landscape Report 2024. restaurant.org
- ReFED. Insights Engine: Food Waste Estimates for U.S. Restaurants. refed.org — verify specific figure before publishing
