Fixing the Disconnected Grind and Avoiding the Lazy AI Trap

David Cantu has deep empathy for restaurant operators because he walked in their shoes for almost two decades. He knew they often lacked the right tools to meet the needs of the business and wanted to change that. 

“Running a multimillion-dollar operation well takes a tremendous amount of insight, that usually only comes with tenure, years of scars, and lessons learned the hard way. Restaurateurs wear that like a badge of honor. We feel that same need to serve, at all costs, and too often that cost gets paid in relationships outside the restaurant.”

The CEO of Craftable, a hospitality platform that brings siloed operational data together, previously co-founded HotSchedules, served as CEO of BlackBox Intelligence. He started his career in restaurant operations at P.F. Changs, and currently co-owns a restaurant. Cantu believes everything Craftable does starts with the idea of whether it's useful to an operator standing on the floor at 6 p.m. on a Friday, not whether it looks impressive in a slide deck. 

“Operator first isn't just a tagline here. It's the filter that every decision runs through before a single line of code gets written. There are so many talented, smart people running these businesses who deserve better tools than what they've had, and that's what we're dedicated to at Craftable, making their lives less stressful, constantly looking for ways to improve their day to day and clear the obstacles in front of them.”

Crafti, the AI layer inside Craftable, was designed to provide real-time visibility into food costs, vendor spend, and margins, in one system, giving operators the ability to control  the controllable and accurately forecast to make better calls faster instead of simply guessing under pressure.

In this conversation with Modern Restaurant Management (MRM) magazine, Cantu discusses how restaurants are using AI to enhance the guest experience, AI red flags, practical strategies for reducing food waste, and data integration across systems.

What are current trends in AI use in restaurants?

Everyone's talking about the visible stuff – voice ordering, chatbots, AI answering the phone, etc. That's cool, but it's not where the money shows up. The bigger shift right now is happening in the back office – think forecasting, ordering, labor, and invoicing. It’s the unglamorous “plumbing” that decides whether a restaurant makes money. For most operators, the information lives in tools that were not built to talk to each other, so while pieces are automated, it’s not truly built for the operator’s needs. 

The story that matters in 2026 isn't AI replacing staff, it's AI cutting the burnout and chaos that drives turnover in the first place, and a lot of that burnout comes from managers being the human glue holding disconnected systems together.

Forecasting is also getting sharper. It's not just looking at last year's sales anymore, it's pulling in weather, local events, holidays, etc., and that forecast is starting to connect directly into ordering, prep, and staffing, instead of sitting in its own report. Some operators using predictive analytics this way have cut food waste by up to 35 percent, while making sure their best sellers never run out. That's the gap between reacting to a bad week and seeing it coming ahead of time.

The other misread is  negativity around staff retention. The story that matters in 2026 isn't AI replacing staff, it's AI cutting the burnout and chaos that drives turnover in the first place, and a lot of that burnout comes from managers being the human glue holding disconnected systems together. Give a GM their time back, give them a forecast they can trust, put it all in one connected system instead of five, and you fix what's been driving people out of this industry. Not headcount. Burnout. 

What misconceptions about AI do you hear from operators?

This might not even be a misconception so much as operators not having seen the difference yet.. There's an assumption that all LLMs are created equal. They're not. A lot of operators we speak to have tried an LLM on their own, without really vetting the platform, and got lackluster results because the model wasn't personalized to their business or their actual problems A generic model can't see across your operational data, it doesn't know your vendors, your recipe costs, your labor patterns, so it can't hand you an answer that's actually specific to your restaurant. It takes refinement, training, and agents that get tuned and QA'ed over time rather than just switched on and not updated. We've seen what happens when that tuning doesn't happen, and the trust just isn't there yet, understandably so. You earn that trust by showing the data not by asking people to take your word for it., 

There's a well-known study out of MIT's NANDA initiative from 2025 that found 95 percent of generative AI pilots at companies failed to show any measurable return. That number got a lot of attention, but the reasons behind it matter more than the number itself. Mostly it came down to generic tools that were never built for a specific workflow, data that was messy or disconnected from the start, and teams launching a pilot without ever defining what success would look like. Closing that gap is exactly what we're trying to do. Not by pretending AI is magic, but by doing the unglamorous work, building it specific to this industry, testing it against real restaurant data, and being honest when it's not there yet. It's not an easy road. But a team that's genuinely dedicated and locked in on its customers can get there, and we're seeing it happen.

In what ways does AI help managers and staff focus on guest experience?

Operators miss this a lot. When a manager isn't buried in a walk in counting inventory or triple checking a schedule, they're free to be present with guests. That's the gift AI gives back, time doing what they love. But time is only half of it. The other half is knowing where to focus attention, and that's where AI starts to help with staff too, not just ops.

It comes down to whether the forecast did its job in the back office, did it get the right people on the schedule for the demand we knew was coming, did it drive the right order and have the right prep to back that up. When it doesn't, guests feel it immediately; their favorite dish is 86'ed, or service drags because they were short staffed.

But guest experience isn't only a supply and staffing problem, it's also a people problem, and that's a different lever entirely. A kitchen can be fully stocked and a floor fully staffed, and a guest can still have a bad visit because their server seemed checked out or didn't know the menu. This is where AI is starting to do something new, picking up on signals that used to live in a manager's gut instead of any system and telling them where to look, who might need a coaching conversation this week, and who's quietly having a great month and deserves to hear it.

So, when restaurant leaders evaluate AI, the real question isn't whether it looks smart on a dashboard. It's whether it connects the back office, forecasting, labor, ordering, and prep, to what's happening on the floor with guests and staff, so managers spend less time firefighting and more time coaching the people who are the experience.

What are overlooked data points that operators should be tracking to be more profitable?

Purchasing, invoicing, inventory, recipe costing, labor planning, these are the things that determine profitability, and they're still mostly analogue in most operations. Worse, they're split across tools that were never built to talk to each other. That's why they get overlooked. This stuff isn't sitting on one dashboard anywhere; it's buried in invoices, spreadsheets, three different logins, and someone's head. As margins get tighter, you can't afford those five things to live in silos anymore. Tracking each one better in its own separate tool isn't the win. Tying them together is, so a change in food cost shows up automatically in recipe pricing, and a shift in demand shows up automatically in the labor plan, without a manager sitting there connecting the dots by hand.

How can AI be used to impact a restaurant's food waste?

Managers and chefs over-order because there's no tool telling them the right quantity, and that leads to waste. But food is only the most visible part of the problem. The same blind spot shows up with paper goods, to- go supplies, cleaning chemicals, dish soap, even utilities. Nobody watches those the way they watch food cost, so that waste goes completely unmeasured. And you can't fix what you're not tracking.

Ultimately, AI usage should follow the same rule as any restaurant technology. It has to add value to your operations, make your GM and staff more productive, and help your margins.

Even within food, most kitchens run on static pars, the same number every day regardless of what's happening that week. But pars were never built to be accurate. They were built to be simple, back when there were no real time data and no way for ordering, forecasting, and prep to talk to each other. A manager needed one number they could remember and act on without recalculating anything, so that's what par gave them. Fine, when it was the only option. Not fine now, because those systems are still disconnected in most restaurants, and the cost of that shows up as waste on one end and stockouts on the other, across every category.

This is where AI earns its keep, catching margin drift across all the controllables by finally connecting what's always lived apart. Craftable ingests every invoice in real time, across food and non-food categories alike, catching pack size changes or over ordering the moment they happen, so operators see where margin is drifting, on the walk-in shelf or the supply closet, instead of finding out weeks later on a P&L.

What excites you about increasing use of AI in restaurants?

The trust growing between AI and GMs. Like any restaurant tech, AI is only valuable if operators trust the answers, and Crafti, our AI assistant, was built around exactly that. It won't hand a GM a vague line like "your food waste looks high." It runs the investigation, queries the data directly, and hands over the actual numbers behind the conclusion. That transparency, on a platform we built in-house rather than skinning someone else's generic tool, with forecasting, ordering, and prep talking to each other instead of living apart, is what has me excited about where this industry is headed.

Here's the future I picture. A GM or chef driving to work gets an audio rundown of everything that matters informed by real-time data. Not a dashboard they have to dig through via four logins, but a teammate who already did the digging. That's what excites me, AI that doesn't just inform a GM, it has their back. 

Are there any cautions or red flags?

If the platform can't show you why it landed on a conclusion and gives a vague answer with nothing behind it, that's a trust problem waiting to happen. 

If it's a generic model wearing a restaurant skin instead of something built on restaurant data, it'll hit a ceiling fast. And if it only solves one piece of the puzzle, e.g. a forecasting tool that doesn't talk to your ordering system ora labor tool blind to your prep needs, then you haven't fixed anything. You've only added a sixth tool to the five your team is already stitching together by hand. Good AI shows its work, connects to the rest of your operation, gets used, and pays for itself in time or margin. If it's not doing that, it's not the right tool yet.

Ultimately, AI usage should follow the same rule as any restaurant technology. It has to add value to your operations, make your GM and staff more productive, and help your margins. If it's not doing those three things, tread carefully.