The Tech Stack Is Dead: Building the Restaurant Operating System

Matt Wampler (Intro): On today’s episode of the Restaurant AI Podcast, we’re doing something we’ve never done before. Two guests, one conversation, and a question every operator is about to face: what happens when the tech stack dies and your restaurant runs on an operating system instead? I’m joined by Deric Rosenbaum, President and CTO of Groucho’s Deli, an 85 year old brand he’s rebuilding for the AI era, and David Corts, CEO of Fresh Technology, the company behind Fresh KDS. The three of us have been building a connected back of house system together, and this is the honest, behind the scenes version of that conversation. We dig into why dashboards are dead, what AI actually needs before it can run your kitchen, and how far away we really are from turning an agent loose on operations. But before that, a quick thanks to our sponsor. ClearCOGS helps restaurants turn the data they already have into practical daily guidance for prep, ordering, and labor, so operators can make better decisions before the day gets away from them. Because when you have a clearer view of what’s coming, you can run a more profitable restaurant. Now, here’s my conversation with Deric and David.

Matt Wampler (00:00): All right, I don’t know if you guys saw this, but my wife brought this up. Apparently IQs are going down, like measurably. Just did a big study, like four points down. Did is anybody talking about that with you guys?

David Corts (00:15): I’ve heard the same thing. Is it because and then is it attributed to AI? Is that the idea?

Matt Wampler (00:21): I can’t imagine it is. I would think it’s social media.

Deric Rosenbaum (00:22): You couldn’t measure that this early accurately.

Matt Wampler (00:25): No. I mean I gotta it’s gotta be social media. But I I I

Deric Rosenbaum (00:37): I mean, it could be an impact. I mean, COVID definitely had an impact, you know, around social skills and and soft skills, especially with the young generation that was, you know, quarantined to their home for however long and now they’re in the workforce. And so, I mean, maybe you attribute that to IQ? I I don’t know. I mean, is it intelligence or is it social IQ? Or is it you know, what IQ are we talking about?

Matt Wampler (00:59): Well that was

David Corts (00:59): Yeah, right.

Matt Wampler (00:59): the thing that the guy brought up with me. He was like, No, this isn’t their scores. Scores go down, but raw intelligence going down isn’t how well you did on a test, it’s actually your intelligence.

Deric Rosenbaum (01:16): I I I don’t have near enough data to even give you a remote informed opinion.

Matt Wampler (01:20): I think the real question is you gotta I mean, that’s a lot of your workforce. You see it at all?

Deric Rosenbaum (01:26): I mean, the younger generation is Pete Riggs and I had a conversation last week about this, and I think he surmised it perfectly. That, you know, w the younger generation now views us paying them as paying them for their time and not for their care and effort. Where I grew up, those were synonymous. Like if I’m paying you for your time, I’m paying you for your care and effort.

Matt Wampler (01:51): Do you think that’s the whole gig economy thing? ‘Cause I remember a a while back talking to my cousin, it was like everybody’s got three jobs now and they’re an entrepreneur and if I don’t pay enough, they’re gonna just go drive Uber.

Deric Rosenbaum (02:02): You talk to any Uber driver or Facebook, what is your real job? Well, I’m a content creator. I mean, what the hell does that even mean? Like, you know

Matt Wampler (02:10): Says the guy in the booth with the hashtags.

Deric Rosenbaum (02:15): Amen. Always be branded, baby.

Matt Wampler (02:18): love that. I love that. No, I I I keep struggling with this with the whole AI thing, you know, is it I I hear a lot of people say it’s making us all dumber. And on the flip side, I’ve got a nine, ten year old daughter and she wants to get on Chat GPT and build games and I’m like, you know, I I have to imagine that’s a good muscle. It’s a tool for the future. But am I gonna make her dumber?

David Corts (02:42): I was

Deric Rosenbaum (02:43): I think it it depends on the person. Like for me, like my brain is wired to work with artificial intelligence. Like if you’re a lateral thinker, artificial

David Corts (02:54): Yeah.

Deric Rosenbaum (02:54): intelligence is a godsend to you.

David Corts (02:57): I think there’s two types of people in the world. And I think I heard Bill Garlay talking about this the other day. It’s either the type of people that AI is gonna make ten times smarter. And I sort of put myself in that category. I’m probably like Derek. I have a million things and love to put puzzle pieces together and this helps me just dive into things that I want to learn about and it’s enabling and empowering. And then you have people that are gonna use it to cut corners and just be dumber. And so I think it’s kind of binary. Also just we’re talking about that younger generation. I have also read that on college campuses and looking at my children who are a little bit older, range seventeen to twenty four, like they’re not real into it. And on college campuses, it’s like particularly prestigious ones, like eighty percent backlash against it, which is really scary.

Matt Wampler (03:48): somebody was telling me the same thing. I think it was Josh from Big Chicken. He’s like, All of our younger people hate it ’cause it’s the thing that’s quote unquote taking their jobs and you know they don’t want anything to do with it. I thought that I hey, if I was in college and I had AI, college would have gotten a lot easier for me.

David Corts (04:03): Yeah, I might like dropped out of college and started a company.

Deric Rosenbaum (04:07): That’s what I did without AI, so

Matt Wampler (04:14): Turns out you didn’t need AI to drop out of college.

Deric Rosenbaum (04:18): No. I I you know, I I agree with David a hundred percent.

Matt Wampler (04:23): And lateral

Deric Rosenbaum (04:23): Hundred percent.

Matt Wampler (04:23): thinking is just just you have lots of interests in lots of different places.

Deric Rosenbaum (04:27): Well, no, like if a problem comes to me, most people just look at it linear, like this is the problem, blah blah blah blah. I look at all the inputs. to make the best informed decision of how do we solve this, how do we reverse engineer this? How do we build for this XYZ?

Matt Wampler (04:47): I c kind of keep oscillating back and forth between AI is going to be like an insanely powerful thing and we’re in the first inning of it and it’s moving so fast and the AI companies are only showing us, you know, the the lowest models they have, and then in two years everything we’re talking about is going to be moot and there’s going to be AGI and it’s going to be changing everything. And then the other part of it that it’s just, you know. predicting the next word and it’s not really going to it’s gonna be the internet and it’s gonna take twenty years before it actually changes anything.

David Corts (05:22): It’s probably somewhere in between. I think the technology is probably there to be a lot more powerful, how it propagates through society and what society’s ready for. I mean the humans have a very strong immune system, right? And just because AI is capable of it, our own self interest and politics and regulation, there’s gonna be a lot of human forces that stop it dead in its tracks in a lot of ways before it can really deliver on the promises.

Matt Wampler (05:55): I I was watching CNBC the other day and they had all the Chinese robots doing like choreographed

David Corts (05:58): Right.

Matt Wampler (06:00): dances and I asked my ten year old to come in and watch and she looked at it and for the last three days all she said is, I think technology’s gonna take over everything. You’re driving through a field and she’s like, No, I think there’s gonna be like robots in that field before

David Corts (06:12): Yeah.

Matt Wampler (06:13): too long.

David Corts (06:13): Did you see the one that ran B2N Bolt’s speed record? Yeah,

Matt Wampler (06:18): No.

David Corts (06:19): that happened last week.

Matt Wampler (06:23): Yeah. Something new and weird. The moment I see one of those like you know how we have the little like carts that are going down the street today that are kind of idiots. They always get in your way, block the cars.

David Corts (06:31): Yeah. Yeah.

Matt Wampler (06:33): The moment I see one of the Terminators walking down the street like with a delivery in their hand,

Deric Rosenbaum (06:38): Yeah.

Matt Wampler (06:39): yeah, that’s that’s gonna scare the hell out of me.

Deric Rosenbaum (06:41): I mean, y you see

David Corts (06:42): Mm-hmm.

Deric Rosenbaum (06:42): some of these videos, especially with these military grade robots where they can go from bipedal to quadpedal and like jump across I mean, it’s just mind blowing that they can build these things now. Like

David Corts (06:52): It is. And there there’s gonna have to be a some regulation around it because they’re humanoid, but they can be three times the speed and ten times the strength of humans. Like it’s potentially really dangerous.

Matt Wampler (07:07): We wrote the Terminator movie. We we’ve seen

David Corts (07:10): And it’s

Matt Wampler (07:10): iRobot. We know what this future can look like. It’s it’s a little scary that we’ve decided to build it.

Deric Rosenbaum (07:17): Sir, y you do run a restaurant I mean a AI podcast. So I think you feel like you’re kinda in the weeds here.

Matt Wampler (07:24): Yeah, maybe a little bit. Who knows? Well take it back to restaurants. You you are, you know, running brick and mortar on a daily basis. Is AI even something that, you know, is on your radar or changes the way you plan?

Deric Rosenbaum (07:39): Thousand percent. Thousand percent. I mean we I I mean part of what we’ve been working on and working through is you know this whole re reenvisionment of of our our systems and we kind of put it into four buckets. So we have the transactional layer and that’s your source of truth. So for us that’s square. So it’s a singular omnichannel menu feeding everything everywhere, your nap information. all the things that need to go everywhere, right? And then the next layer above that is the brand layer. So everything gets benchmarked against your brand standards, your manuals, your KPI, all the things that you do, right? And that’s what I’ve spent all year rebuilding to make it AI compatible, digestible, and have tied agentic tools to it.

Matt Wampler (08:30): And you did it.

Deric Rosenbaum (08:31): Correct.

Matt Wampler (08:33): That’s that’s insane in itself.

Deric Rosenbaum (08:35): Well, Claude and I did it, but yes. you know,

Matt Wampler (08:37): Yeah.

David Corts (08:37): He’s a lateral thing.

Matt Wampler (08:39): Lateral thinker, there you go.

Deric Rosenbaum (08:41): and then the the third piece is the best and fit like targeted targeted vendor. So like you would fall into that bucket, ovation would fall into that bucket, marquee would fall into that bucket. They’re bringing the best of a specific niche thing that I see is the future of how we need to run our business. And then what I think is coming next. And you can’t do the next without the first two being like spot on. Like you have to have the source of truth. You can’t have six fucking menus. You to have one menu, you know. And you gotta have the brand layer that’s fully synthesized. But the I think the fourth layer and which is coming and the piece that we’re shopping actively is the synthesis layer. And that’s taking all of this stuff up and down and outputting three buckets marketing, operations, and supply chain management. We’re working on the supply chain management piece.

David Corts (09:37): Mm.

Deric Rosenbaum (09:38): I’m working with PerDiem on the marketing automation piece. They’re building CDP within their tools. So I mean, and like that is taking, you know, forecasting or sentiment analysis or dispute recovery and then correlating that directly to sales along with other KPIs that you’re measuring, whatever those KPIs. Are important for your brand, right? Whether it’s AUV, LTV, pick up pick a P, right? and that that’s what I see. So yes, it’s an integral part of what we do. It’s an integral part of what I’m building for the future. And I could not build these things, I couldn’t rebuild our toolbox without AI. I I no human could have the capacity for that.

Matt Wampler (10:18): I I want to get to the whole like synthesis orchestration layer, but before that, David, you know, I know you guys are doing stuff internally for, you know, your AI operating system. I I

David Corts (10:32): Yeah.

Matt Wampler (10:32): I deal with the same thing. I’ve got one set up for finance and one for marketing and one for investors. And they kind of work well in their own little sandbox, but the moment the picture becomes too broad, the whole thing just starts to collapse. What What are you doing and how are you getting it all to work together?

David Corts (10:51): Great question. And I’m I’m excited to go back to Derek’s think too because I got a million questions on that. But I think building an operating system for a tech company, for a restaurant, it’s all probably the same key components. And the way Derek laid it out makes a lot of sense. But yeah, we have the way we think about it, right, is we have systems of record. And for a while everybody thought, you’re gonna code your own HubSpot, right? Just the purpose of HubSpot, we have It it’s just additional context. The whole game is context engineering. So HubSpot still holds all of our prospects and it still holds the stages for our sales cycle and all of those things. And we have another product called FinCorp that has all of our customer revenue subscription data, but highly curated. But these are all just forms of context that still exist in systems of record. What’s missing is all of the context engineering. To make that data more useful. So on the other side of the equation, we have GitHub repos, not for our code, those are obviously in GitHub repos too, but GitHub repos, one that we call canonical, meaning this is the truth. This is what our pricing is, what our discount is, probably similar to what you’re saying when you can’t have 12 fucking menus, right? This is things that are highly curated, controlled by only a couple of people in the company that say, This is our brand messaging. This is our core values. This is the process we follow to onboard a partner, et cetera, et cetera. That’s PR gated and can’t be changed. Underneath that is a different repo, which is what we call the ops repo, that probably gets 200 commits to it a day. It’s the synthesis of every conversation that happens around the company, right? It’s the questions that people are asking of the canonical brain. It’s every update. And then when you have the infrastructure there with all of this context that’s managed and you know what’s true and relevant and what is additional context to augment that, then building agents on top of that, and we’ve built it all model agnostic, so you can use Claud or Codex or an open weight model or whatever you want to, those agents become incredibly powerful. They can do things. So I mean, Derek, what I think you’re describing is you need highly curated data in your kitchen about what happened, right? And you can hopefully get that from us, which

Deric Rosenbaum (13:20): Yeah.

David Corts (13:20): is not just a bunch of raw ingredients, but it is like the analogy would be it’s not just a whole bunch of raw ingredients in a restaurant. The data collected from the fire hose of information that comes out of every tap on your screen gets curated into a semantic layer that actually makes sense for the goals that you’re trying to accomplish. Which are probably get every order across every channel to the customer and keep that brand promise, right? You have the same thing with ovation data, you have the same thing with your forecasting, all of that, but you have to have these systems that you know and trust. When you have that with the additional context about what you’re trying to do as a brand and a company, building agents on that, you can use Mythos or you can use a low-level Chinese open white model, and the result is great. Because it’s all just shooting

Deric Rosenbaum (14:10): Phenomenal.

David Corts (14:11): vectors through it. So what we’re trying to do for you is say, how do we take all of that fire hose and noise that comes out of every order that comes in, hits a screen, goes to another screen, gets tapped, etc.? How to we how do we synthesize that, which is another key word you said, into something that has meaning? Like we’re big on we sell meaning, not metrics, right? ‘Cause you pe I hurts me to watch us build more and more dashboards because that’s what people think they want to buy, those things are gone. It doesn’t make any difference. You have to have curated data. Yeah, exactly. I got a hundred

Deric Rosenbaum (14:44): I I’m dashboard rich, so yeah.

David Corts (14:48): of them and I do nothing with any of them. So anyway, that was a long answer to a simple question, but that’s the way we think about it. And I think it’s very similar to the way that you’re thinking about it, just with a different

Deric Rosenbaum (15:00): I I I think you you’re spot on and you said two key words and I will add one more. So I mean the hard part for me in in in transforming all of this institutional knowledge that we’ve collected over the decades i is you have to bring this ontology layer, the semantic layer, and the the conical

David Corts (15:19): Yes.

Deric Rosenbaum (15:20): glossary for us is how we built it. So that if that word comes up and it shows up in the conical glossary,

David Corts (15:28): Yeah.

Deric Rosenbaum (15:29): that What is the definition of this word? Period.

David Corts (15:32): You just said yeah. I’ve been talking about an an ontology for the kitchen for a year and a half, and you’re the first operator I’ve met who said it before I said it. And that makes me so happy. That is the whole

Deric Rosenbaum (15:46): And I

David Corts (15:47): thing. You’re making a a a digital representation of your business, which is the ontology. And we have a graph index and machine readable front matter on every file. So that when you deploy an agent it knows exactly what it’s looking for, where to get it and what to do next.

Deric Rosenbaum (16:06): And so when I work with a vendor partner and and so maybe they need access to part of our institutional knowledge. Now I can share with them the structure, the ontology, all of this, so they know how it it’s almost like an MCP, but not really an I don’t have an MCP into my data. But you know, it it’s the same thing. I mean, an MCP is basically a set of instructions to tell an agent how to interact with another LLM, right? And it it’s kind of the same construct. Like we as brands. I think we have a responsibility to get our houses in order first. So the the layer one, which is the source of truth, layer

David Corts (16:40): Mm-hmm.

Deric Rosenbaum (16:41): two is the brand piece, because if you don’t get those, the other pieces will never effectively communicate up and down and you’ll never be able to synthesize the data. Because if you have three menus, one on DoorDash, one natively, and one on your own first party, they’re all selling the same bacon and turkey club, but they’re all named slightly differently or have, you know, then What is the source of truth, right? How does that get consolidated on the back end? Or how does how does ClearCogs break down those three different versions of the same thing? It’s the same make, it’s the same everything, but it’s the wrong

David Corts (17:16): Yeah.

Deric Rosenbaum (17:16): name. It’s a naming issue. So naming conventions kind of fall into that whole that whole part of I mean, the taxonomy of the ontology

David Corts (17:24): They do it it

Deric Rosenbaum (17:26): matters.

David Corts (17:27): has we have ontologies, right? And then we have vocabulary files. So say for this entity, you know, called a customer in our ontology that has tons of instances underneath that, there are properties that can be applied to that. But there’s a strict vocabulary file that it has to look at to go, I know that these are my choices for how I can define that. It’s it’s literally building a digital representation of your whole business. And

Matt Wampler (17:54): Yeah, but what what do

Deric Rosenbaum (17:55): When you when

Matt Wampler (17:55): you do with it, right? I get it I

Deric Rosenbaum (17:55): you

Matt Wampler (17:57): get if you’re doing a marketing campaign or you need to answer questions for an RFP, but like what’s the day to day practical use?

Deric Rosenbaum (18:06): So w w we built

David Corts (18:07): gosh.

Deric Rosenbaum (18:09): for for our brand, I’ll I’ll specifically speak about the work we’ve done, is you know, that whole layer and building that toolbox and then attaching our agent, Frank, which stands for franchise resource and network knowledge, allows our operators through a a dashboard that we built is kind of a one stop shop. We call it Grouch.os, so Grouchos. But with that, with that agent and this dashboard, like our operators can go in and get any SOP, they can get any video, they can get any make instructions, they can get anything related to their FDD, they can get I mean, literally whatever the thing is, it resides. And then we monitor it, we monitor the questions coming in. And if something’s not there, we go build it if it if it makes sense to have it within within our structures and our systems. So it it becomes a factual one stop knowledge base for our quick reference. Like You can’t talk to the director of operations. You can’t talk to me. You can’t talk to a field agent on demand, right? But you can chat with his agent on demand. And, you know, our job is to make sure that the results that we’re giving are accurate. And the only way to do that is through what we just discussed. And that is the hard part. And there’s no other way to do that but to roll up your sleeves and do it. Like, period. Like you just got to get this shit done.

David Corts (19:24): Yeah.

Matt Wampler (19:25): And Derek, do you see a lot of other restaurant brands doing this? Is this common at this stage?

Deric Rosenbaum (19:29): No, no, no. I I am definitely a a challenge thinker on this front, I’m pretty sure.

David Corts (19:37): it was exactly one of the first things that we realized we could do with this, which is if if you could be in every restaurant with your experience, your knowledge, your access to information, your, you know, long history with doing what you’re doing, if you could stand there in every restaurant and coach somebody every day and ask every question, identify every problem and solve it right there on the spot, you could do that, right? Can you do that across 30 stores? No, it’s absolutely impossible. And the further you get away from you, it’s like ripples in a pond, right? And the less time you spend. And we notice the same thing. I can have all hands here and say, like, this is what we’re doing, this is why it matters, this is how we do it, et cetera, et cetera. But the further you get from the core strategy part of the business, again, it’s the ripples in the pond. And we would look at what people were saying to customers and look at the presentations that they were making to prospects, and you’re just like, How how is it so degraded? You know? But when you make things canonical, like you’re saying, right? And inject them into the work stream, I look at presentations that an account manager makes to a customer, and it’s like, Whoa, I couldn’t have said that better myself. Like And it’s tied to a design system. So it always you know, we used to get I’m so sorry, I can’t make my thing go into do not disturb. You can

Matt Wampler (21:02): Are you sure you an attack company over there?

David Corts (21:04): Yeah, I don’t know how let me ask Claude how to make that happen. Anyway, you’re basically like cloning yourself and the other leaders in the company and then injecting them into the work that everybody else is doing. And it’s a really powerful effect.

Matt Wampler (21:24): Omnipresent.

David Corts (21:26): Yeah.

Matt Wampler (21:27): Well, Derek, let’s get back to the back of house. You know, former operator, you know, my heart is in the within the four walls. It’s tough being a general manager today. You know, this is all great, but most of the work in a restaurant is hands-on. It’s training, it’s guests, it’s slicing meats. You know, how does this actually affect their world and how do you look at it?

Deric Rosenbaum (21:52): At the end of the day Hospitality is human derived, period. Humans are fallible, and so is technology. so I I I think as society changes and morphs and adapts, and and we can bring in these intelligence layers into our operations and then automate some of these mundane and repetitive tasks that specifically our general managers and shift leads are doing, right? You know, the the the inventory counts or the or the placing truck orders. Or having to print this so they can go do par levels or fill in the blank, right? You know, how many hours a week are spent on actually placing truck orders? And, you know, yes, we have franchisees that have been with us for 20 years and their institutional knowledge cannot be replaced, right? Because I mean they just they have that instinctual gut. They can walk into a walk-in, not write anything down and know exactly what they need to order. How do you take that level of knowledge? And shorten the learning curve to a new franchisee. You do that through automation. So if you have a source of truth with a singular menu coming from all of your channels into your point of sale, and then taking that information and sending it to a clear cogs, for example, who has taken all of our recipe work and taken everything down to the slice, the ounce, the cut, the box, the whatever. And then convert that to the nomenclature that our brand uses and understands every single day. Because you know, what might be called something at this brand is called something else at Groucho’s, right? And so that what it’s called over there doesn’t make sense to us. So we take all that information, you synthesize that, and distill back, you know, a daily prep sheet, a daily pack sheet, a daily recipe sheet. And if you do this right and accurately, a truck order. You know, and then there’s automation that can happen after that. But I mean, that’s gonna require another third party to broad lines. And those are things that are happening. And that’s where I think where the synthesis layer comes in. And then like the the the big piece, you know, with Fresh KDS and you guys is you know, the friction point for adopting clear cogs for our franchisees was the physical act of having to print this long piece

Matt Wampler (24:14): Yeah.

Deric Rosenbaum (24:14): of paper with these checklists and k and and You know, clipboards and you know, nobody wants to deal with that. No nobody wants managers don’t want to print anything if they don’t have to. So I mean, you worked with another forward-looking brand like Fresh KDS, and now that information resides in our KDS. So our operators and team members are looking at those screens already every morning. So now when they walk in, they don’t have to print anything, they don’t have to pull out a clipboard, you know, there’s no more paper, there’s no more check boxes, it’s just there on a screen. Here is your execution list. for today around these deliverables, whether it’s making a recipe, cutting turkey, or packing salad dressing. Like

Matt Wampler (24:56): And David, I wanna I wanna kick it to you. so, you know, when Derek first brought this up, this idea of having centralized intelligence, you know, within the operations, within the back of house, you know, I’d always felt like forecasting, being able to know what’s gonna happen tomorrow, the answer to do I bake that extra cycle of bread or what do I order or what’s gonna you know, it’s two hours prior to close, do I need to grill that last batch of chicken? You know, these were answers, but I never really knew where those answers would live, right? I knew they were important. And you guys have really thought about how to make a connected system in the back of the house. So I don’t know. Derek had brought this up and brought it to me, and we all started talking. I’d love you to kind of shine some light on how you’re viewing the actual workflows, the operating system in the back of the house. How it comes together.

David Corts (25:57): Yeah, and I think it’s it’s a good question. And what I can’t what I’m not sure of yet is what that synthesis layer that Derek’s talking about really looks like. Because I think what we did together for Grouchos is indicative of the way is the the very beginning of the way I think a lot of these vendors are gonna work together, right? Which calls into question like, do you need a synthesis layer or is that synthesis layer just Claud or ChatGPT or some other model or an interface that Derek builds himself. Because I think if the component pieces like us and like you are doing their job very well, but then making that data accessible, whether that’s through an MCP or APIs, but more likely I think you’re gonna see agent-to-agent communication. So your agent, when we’re working together, is gonna call one of our agents. And say, I need this type of information to accomplish the goal that I’m working on. Same thing with the labor system, same thing with the back of house system. And I think the the vendors at this application layer that Derek’s talking about that do their job extremely well and then curate that data and make it accessible or create agents that can act on that data and then communicate with other agents are going to be the most successful vendors out there.

Matt Wampler (27:23): Yeah, that sounds good. But Derek, to you, I mean, I always view it almost like a school of fish. They move together in harmony, they react together, they’re in sync. And short

David Corts (27:33): Okay.

Matt Wampler (27:34): of chips being implanted in all of our employees’ brains or them all being on screens to get updates from their agents, how does it how do you actually view keeping everyone in sync and running on the same, you know, operational system and game plan?

Deric Rosenbaum (27:53): First, you need trust, trust in the systems and the processes, right? And proof provides trust. And I think we’re definitely still in the proving phase of all of this. Like, I mean, we’re we’re constantly working on refinements with clear cogs specifically around, you know, how we operate, what we do, and and right down to the store level, because this operator might operate just slightly different than this operator, right? I mean, there is no everybody learns and looks and thinks differently, right? And so h how do you standardize that? And it’s especially hard with a eighty-five year old legacy brand because I mean we’ve had operators that have been here forever and ever and ever and I’m in. And you know, they’re like, I don’t need to do that. I mean, I have a system. I’ve been yeah, yes, you do, but there’s a newer generation of workforce that’s coming that is not gonna adapt to your old ways. That they’re just not. I mean, they have no desire to adapt to a pen and paper model. Like it’s just not gonna happen. So I mean part of it is I I think part of it is Let me be clear. I I think most of this automation and these things has to be in the back of the house. It’s not a guest facing thing. We’re not putting robots out to talk to our guest or to take food to them or to you know, from from the cut guest facing perspective, it’s about How are we providing the best possible Grouchos experience and then the channel in which they choose to engage with our brand? and we’re agnostic when it comes to that. And so that part is out of the the equation. I I think I don’t have a great answer to this question, Matt, to be honest with you. I I think it’s it’s very brand dependent.

Matt Wampler (29:33): Well it it

Deric Rosenbaum (29:35): And and part of it is like I’m a big fan of proving out the use case. I mean, I think you’ve known me long enough to know that and I and I prove it out in my own store. So like How do you do that? How do you get the adoption? Then you find the naysayers or or the progressive thinkers.

Matt Wampler (29:51): Well, I I think nobody’s figured it out, right? So this is all just speculation. This is how do we think about these things and how are they going to manifest? I’ve kind of arrived at it’s going to be one of two things. We’re wearing headsets and we’re having a voice AI agent that’s talking with the staff and saying, Hey, a big catering order just came in. I need you pulled off to go, you know, pull thaw some, you know, chicken for the order and you to go do a transfer order. Or it’s going to be on our KDS screens because that’s one of the things that we’re all interacting with. and so I would have thought, David, on your end, you guys would be starting to think about how do we, you know, become the communication layer that keeps everyone in harmony.

David Corts (30:31): I mean, we definitely are. And I think it’s it’s the completion of that flywheel where you have the intelligence, you have the monitoring systems. But we have right now that catering example you just gave, that’s an automation that you can build in like a Boolean way inside the interface and say, if this, you know, if you see this trigger under these conditions, take this action. So in our system today, that will pop up if you want it to, a notification that slides across the screen and says, Hey a big catering order came in. It stops there now. It doesn’t notify another back of house system, but there’s no reason that it couldn’t. It could turn on another fryer. It could do anything you wanted to, really.

Matt Wampler (31:14): Well, so so playing with that, I I guess the connection piece is you’ve built the infrastructure to create custom automations, you know,

David Corts (31:21): Yep.

Matt Wampler (31:22): within the KDS display so that there’s rules based logic. Derek, you’ve built out a, you know, semantic layer that’s pulling in all of your brand and information on operations. You know, the it sounds like the only bridge that’s left is, you know, Derek turning his AI loose to start managing what workflows get created and all of a sudden he’s got an agent that is literally telling the operation what to do and when and changing things that don’t work.

Deric Rosenbaum (31:51): I I I think we’re a ways away from that. this is definitely like crawl, walk, optimize, run. you know, at at best we’re walking right now and or in the early parts of optimization, right? And I mean I I I think we’re at least two years away before we start seeing some sort of like let the agent loose and run your operations. I mean, you gotta have lots of humans in the loop with specialty knowledge.

Matt Wampler (32:18): Dara, two years is a is is the snap of a finger. I mean, like there’s restaurants that are still getting online ordering on. If that’s happening in two years, that’s scary fast.

Deric Rosenbaum (32:29): But you can’t do any

David Corts (32:30): I I think

Deric Rosenbaum (32:30): of this if you if you don’t get the foundation right and

David Corts (32:33): That’s exact that’s the whole point, right? And you said my favorite word again with the ontology, because what we have an ontology of a kitchen in the in the product, but then we have a branch ontology for grouchos. And then we have a world model for each of your locations, right? That knows exactly how you have your screen set up. And it also knows what the the features that you have turned on are intended to do. And it knows your configuration and your store hours and everything because every one of these locations is idiosyncratic and different and has different staff and different demographics and all that stuff. It’s when you understand things at that level and you have an ontology and a world model associated with every one of his stores, yes, with human in the loop, but it the the thing that AI does so well is with the right context, it can analyze all the data under those dashboards that Derek probably won’t look at. And he doesn’t need to anymore because it will tell him you could turn on this feature and I will monitor this and see if it improves your on-time order.

Matt Wampler (33:41): So the reason I wanted to do this with the two of you guys is have an honest conversation and show what behind the scenes conversations are going on. You know, we all started working together to bring Groucho’s Delhi a true back-of-house operating system that can help inform what to do next, what to order, what to prep, integrate it and connect it into their workflows, an intelligent system. And It’s by no means done, right? This is this is an ongoing process. But I’d like to hear from you, Derek, you know, on the walk, for for all the operators out there that are like, what can I actually functionally do today with this technology? And what’s it gonna change? You know, paint me the picture of walk today. Or even better what the vision is and why you wanted it.

Deric Rosenbaum (34:39): You know, I I think the the vision is this. Like, you know, for years and years and years, especially just after post-COVID, you know, tech stack, tech stack, stack stack. Find a gap, plug a p plug a tool in. Find another gap, plug a tool in. Find another gap, plug a tool in. Did they all talk to each other? Very rarely. and so, yes, it solved the problem in the interim, but it just made that much more complexity in the long term, right? And so the the point of this is You you have to have this layered approach. And I I’m I’m of the opinion that the tech stack is over and we’re moving to an operating system model now, right? And so at that you have your core, your kernel, that is your layer one, that is your that is your source of truth, all the things about your business. Because if your menus and your nap, your name, address, phone, all all of your contact information, if none of if that is not accurate, you know. Everything should connect to the source of truth to then push out, right? So data discovery. But you can’t have the discovery if the data’s not accurate or if the data is Duplicative, right? So that’s step one. And I mean, we live in a world of where digital relevance matters more and more and more every single day. I mean, you can order Grouchos on Claude today if you have connected Cash App. So like that exists today. Like, no work on our part because we’ve already done the work on the menu piece. So like that’s just one more channel that it’s gonna allow us to be in. we’re working on another thing I’m not really gonna talk about right now, but you know. But we couldn’t do that if we hadn’t done these base layer components. So I I think you have to think about your relevance. Where do you want to be in 18 months, 36 months? You know, what what what is your timeline? What is your what is your growth plan? You know, are you corporate owned? Are you franchise owned? All of this will dictate, I think, your go to market around how you build your tech stack and you know. There there’s plenty of people that are still out there on these these legacy systems and God I get it. Change resistance is so real. It is so painful and it’s so natural. But I think we are not far away. If if you can’t compete digitally, you’re not gonna end up competing at all. Like I I just if you look at the sh just the the massive shift to off premise that’s happened since post COVID and you know. Yes, traffic’s down. Checks are I mean, well, I guess it depends on the brand, but you know, I just saw a crust report. You know, traffic was down in July and check average was slightly up, and you know, blah blah blah blah, fill in the blanks and all the economic things that are causing this, whether it’s oil prices or war or political uncertainty or, you know, just generalized inflation. and and that’s that’s noise, but if you can’t Find a way to get above that noise, especially in the digital realm, I I just don’t think you’re gonna find relevance in the future.

Matt Wampler (37:55): Well, I I think what I hear is it’s not like you’re, I don’t know, buying an AI subscription to a digital GM that you’re turning loose. You’re basically doing it decision by decision. My my team has to place orders. I want them to be able to place orders in the, you know, closest possible workflow on their KDS display. The team needs to figure out what to prep for the day. I want that displayed so that they’re not looking for the information or Managing Excel spreadsheets and almost taking it decision by decision and not necessarily just I’ve signed up for AI and now it works.

Deric Rosenbaum (38:34): No, it does not it d if you don’t do those other things that we talked about earlier, AI will never work ’cause it will always hallucinate. You cannot build a house on shifting sand. It will fall down. Period. you know, and to your point, like Forecasting, for example, your your your niche and your area of especially, it’s at the end of the day, it’s just math. I mean, it it’s really just math. It’s algorithms and math. So there’s no we live in a world where all of that can be, you know, you know what we sold, we know what went into that down to the unit level. So then you backfill your existing inventory. So you know that based on the last two years of machine learning around our historical sales on Tuesdays, this time of year. you know, this is projected what you’re going to sell within a realm of predictive accuracy. And, you know, so that allows our teams to be much better prepped for the lunch rush. Like as an operator, there is nothing that makes me more angry than to walk into a Groucho’s at eleven one and they are not ready for lunch rush. Like the

Matt Wampler (39:39): Yeah.

Deric Rosenbaum (39:39): nothing makes me angrier, except maybe putting an empty pickle bucket and the walk in. as a former sandwich guy, you know what I’m talking about.

Matt Wampler (39:48): Hey, I I used to have all those stacked up. That was actually my desk. My first desk was an extra door turned over and like six pickle buckets holding it up. It was it was fantastic.

Deric Rosenbaum (39:59): Yeah. It’s it’s you can always know that you always know the gardeners because they’re like, Hey, can I get your pickle buckets?

Matt Wampler (40:04): There you go.

Deric Rosenbaum (40:06): But I mean it it’s it goes back to what I was talking about earlier. It’s like if we can automate these mundane and repetitive tasks, especially if they’re algorithmic driven, it allows our operators to go our our managers to go spend more time with their teams refining the the the the presentation on the plate or the presentation on the table. and all the service components related to it.

David Corts (40:30): I I’ll give you the parallel here because back to the ribbles in the pond. The other thing is I looked around our organization and watched any given job and how much of that job was not really doing the thing that you were paid for? How many salespeople do you have to say, did you put your stuff into HubSpot? Go spend 30% of your time typing things into a system and then you had to have somebody that you paid to yell at them to make sure that they did that. So that we could build a dashboard and all theoretically look at it and change our behavior. Like if you can abstract away a lot of those mundane tasks, your people can do what you actually want them to be doing, right? Which is delivering great hospitality. So I think there’s a real value you’re in abstracting away the things that you traditionally had to do just in order to run the business. And the more of those that go away, The more of the people that you’re paying to do the job to deliver great hospitality actually get to spend time doing that. And I also completely agree with everything that you’ve said that any AI for AI’s sake without that foundation is destined to fail. It’s gonna be an experiment that maybe makes you feel smart for 10 minutes and then you’re gonna realize this doesn’t work. So everything we build internally here or that we put in our product related to AI serves a business goal. That the operator should be able to trace back to profit. Because at the end of the day, that’s what you’re doing.

Matt Wampler (42:00): I’m gonna throw I’m gonna throw an abstract one out at you guys and I I’m curious your thoughts. I feel like there are almost two schools of thought. One that I’m hearing from you guys, which is if you have a system of record, it becomes very easy. I know what I’m gonna sell tomorrow, system of record. I know what I have in stock, you know, therefore I know what to order. You don’t need to be an overly smart AI to figure that out. But the other school of thought is the context and reasoning power of these models is just getting better exponentially. And therefore, you’re almost wasting your time on systems of record because, you know, in six or twelve months, that shifting sands, though they’re going to be able to figure out what the right numbers are anyways, because you’re going to be able to throw enough enough hor AI horsepower at it. Curious, where are you guys? fall. I guess based on the fact that you’re big on systems of record, it probably is systems of record, but how do you think of the the the power and the the the way the models are changing?

Deric Rosenbaum (43:10): I I’ll let David take the mic on this first.

David Corts (43:14): Yeah, I I think some of that is ultimately gonna come down to cost. Like I don’t need Methos three to do what we’re trying to do here. And if I had access to it, it probably would not be cost effective to use it. And models are gonna get better and better and better, but I still think that take any given model, give it context. Perfectly engineered with systems of records and connections into them and rules on when to call that and what that data means, give it all of the context about what we’re trying to do. Any given model is going to perform better with that context, whether it’s just any given model. Mythos will perform better with that context than mythos without it. So if you want to get the most out of whatever you’re spending per token, you should build the infrastructure to deliver that context.

Matt Wampler (44:10): Also makes you less reliant on a single model.

David Corts (44:13): Completely. We that’s a big one for us. Like we we go to great lengths not to lock ourselves into the Clawed Harness or Tet G P T.

Deric Rosenbaum (44:22): I mean t today’s greatest is tomorrow’s not. So like I mean

David Corts (44:26): Yeah, or they can flip the switch and hold you hostage, right? I mean it changes so fast. So being agnostic and we we do a lot of experimentation with the open weight models. And I’m a big I’ve spent my whole career in text and I’m a just a big believer that long term open beats closed. And we’re not trying to fold proteins here or get to Mars, you know.

Deric Rosenbaum (44:52): Just shaving pink poodles,

Matt Wampler (44:52): Yeah.

Deric Rosenbaum (44:53): as Anthony would say.

David Corts (44:53): We’re just we’re just trying to help people get their orders out on time.

Matt Wampler (44:57): You know, it is funny, man. We we used to have conversations with people and they’d be like, Well, how are you gonna actually tell us, you know, what we’re gonna sell in the next hour? Like, you know, th Billy’s

David Corts (45:09): Yeah.

Matt Wampler (45:09): our best person. He’s really good at guessing. He’s got years of institutional knowledge, like, no way you can match that. And I’d always be like, Look, man, we’re putting computer chips in people’s brains. Rockets are landing themselves autonomously. Like, if you don’t think some of this tech could go in and actually Guess what food you need tomorrow? Like kind of living in the stone age.

David Corts (45:30): If if we can’t beat Billy with our technology, we’ve got bigger problems. We’re never getting

Matt Wampler (45:35): Yeah. Amen to that.

Deric Rosenbaum (45:37): I I agree a thousand percent with David. Like, I mean, yes, the models are gonna get better and better and better. But you know, let’s use that betterment to get the outputs to be better. And to use clear cogs as an example, so whatever the next model is, and so now you know what my Tuesday sales are probably gonna look like based on my historical averages. But hey, there’s a giant stormfront coming through. What impact is that storm front gonna have on my revenue? Because we all know that hyper-rainy days are bad for business, right? Or it’s the Sunday after the Georgia, South Carolina football game, it’s gonna get nuts, right? So like that’s where I I think the model piece comes in better. I mean, I I when I when we first when I first like put a model, I tied an agent to our existing toolbox and I was like, cool, it answered the questions. Three questions in, I was like, Yeah, this is not gonna work. So

Matt Wampler (46:28): Yeah, yeah, yeah.

Deric Rosenbaum (46:30): And and again, it’s taking me eight months to rebuild this thing. And I’m on version four because we just launched a new app. So I had to go in and rebuild all of our institutional knowledge around that. And you know, some other updating needed to happen. And but you know, the newer layers, like Fable, allows me to do that at scale. So now I can audit the entire toolbox at one time with eight agents going at once, each one monitoring a separate folder, right? So I I think that’s the value in in growth of the LLMs, not so much. I I I think the transactional, I mean the the data layer is the data layer. And you you have to have the transactional, the contextual in order to do these other things. So you need the transactional for what you do. I need the contextual for the marketing piece. Like so it it it all matters and I I would argue that. Y it has to be that way. And I I was on a call recently and I I said, I liken it. I I liken somebody that’s new to AI like this. You’re like a green cowboy trying to break a Mustang. You’re gonna get kicked in the teeth. Like it’s gonna happen. Like you’re like, this is cool. Let me go build this and then crash and burn. Right. So you read about all these articles, AHO is not performing, AI is not performing, but it’s because they built it on top of something that’s not structured or they bolted it onto something that’s very dated.

Matt Wampler (47:51): Hey, I don’t know if everybody else is doing this experiment, but I I used to do it every six months. I now have it as a continuous experiment experiment that’s always running. And that is, can Claude do what we do at Clear Cox? And it used to be I, you know, every new model would come out and I’d, you know, spend yeah, five, six hours and just let it build and throwing tokens at it. Now I’ve got it literally every day trying to figure out how to improve the models, and it can’t do it yet. And It’s got a long way to go. And for a number of reasons. But I I take that back to say we were, Derek, you and I, and I think David, you were there as well at Mertek. And all I could think at this entire conference was everyone here just wants to know about the SAS popp SAS pocalypse, right? Almost every seminar

David Corts (48:40): Yeah.

Matt Wampler (48:41): is, should have been, is this part of the tech stack something that will not be replaced? I shouldn’t invest in because is gonna get replaced with AI in the next six or twelve months, or I can replace today with AI. How do you view the tech stack today? And what do you think are your core things? Like, you know, Squares made it unbelievably easy to get data access. It’s powered your thing. You’re not, you’re not gonna build a POS anytime soon. You know, how do you think about the tech stack itself and what is durable?

Deric Rosenbaum (49:17): I I think getting getting that the data layer the transactional layer, the data layer and and the brand layer dialed in now gives me optionality in that in that that operations layer where where we have specific partners. You know, we’re we’re a a nimble and lean team. I I I’m not looking to have a specialist in everything that we do. Like I I don’t need somebody trying to build clear cogs on my team. Like I I’m not looking to reinvent the wheel. And but I think so many of the tools that we have and so many of the dashboards we have are prescriptive, right? They tell us what’s already happened. What’s happening now is we’re going to get more into predictive tools. I mean, yes, prescriptive is inherent to predictive. You have to have that to understand what’s coming. And

Matt Wampler (50:11): You got the you got the descriptive

Deric Rosenbaum (50:11): I think in

Matt Wampler (50:12): in the background, the things that had happened in the past. We’re moving to predictive. And then the goal is for us to get to prescriptive, where like we know there’s a problem coming. This is what you need to do in advance.

Deric Rosenbaum (50:25): And I I think inherently certain tools will get folded into other tools. I mean, I I don’t know that it’s a SAS pocalypse. I think it’s a consolidation that anybody that’s been in tech stack has seen coming since like two thousand twenty-two. Like you can’t have twelve loyalty providers. You can’t have, you know, go to any trade show this year and count how many voice AI booths there are. Like probably twenty six. Like, you know

David Corts (50:54): Yeah.

Deric Rosenbaum (50:56): Many are gonna how many how many are gonna make it to the top? Maybe two, maybe three, you know. It’s like anything else, like when third party first started. It was a circular firing squad. There was Doordash and Uber and Waiter and you know, your local place and another regional place. You know, now we’re down to basically two. And I I I think that’s just a natural evolution of how technology works. And it’s not that it replaced jobs, it it’s that. those things evolved and it allows you to go those people to become specialists in something else. Or i i it’s not about replacing jobs and It does change the way I think about it, but it really just narrows my thinking into more hyper specialists. And then let the synthesis layer maybe fill in some of those secondary and tertiary vendors that we use now that aren’t it’s not a dashboard that I I may I might go into once a month, right? It’s doing a job, it’s running in the background. But is it pride and real value? Probably maybe for half the chain, right? So these are the things you can start looking at, I think, once you get the other two pieces completed.

Matt Wampler (52:14): Get the infrastructure in place and all of a sudden you’ve got the optionality to swap some of these out.

Deric Rosenbaum (52:19): That’s the word. You have optionality.

Matt Wampler (52:22): David, what do you think about the whole idea of a SAS pocalypse?

David Corts (52:26): Yeah, I I I kinda didn’t buy it when I when that was the rage whoa four months ago or whenever that was that we were talking about that. I totally got the idea that you can code anybody can code things now. I do it all the time. what I don’t buy is that restaurant operators or fill in the blank want to also be software developers. I know it’s easier, but maintaining software is not Derek’s job, right? And so I think that and everybody realizes, like it’s easy to vibe code, it’s really hard to maintain, iterate on, make secure all this software. Maybe they’ll get better and better at that, but I still think anybody who traditionally buys software is gonna do calculus around how much time and investment do I want to make in building software in order to save X, right? Because that’s time that I’m not spending actually building my own business. So I I thought it was a little overblown then, and the big change that’s happened since then is the models seem to be on a trajectory to be somewhat commoditized. I mean, we just talked about you can with the right infrastructure and the right context, a lot of these models can perform the basic business function and like incredibly well that you need them to. It’s like how much money do you really want to pay for that? the whole open source movement. So I feel like a lot of the value is going to migrate back to that application layer. When we can talk about everything down to the to the metal and to the power behind it. But I think so much of that value is gonna start to reaccumulate, but in a different version of a SaaS business. I think the new SAS businesses are going look a lot more like what Derek is describing with

Matt Wampler (54:21): Yeah, I also think that AI is kind of the scapegoat on this in some ways. Part of me thinks we’re going through a restaurant technology renaissance. I mean, for the early 2000s, what were the biggest innovations? We we got cloud-based POS and online ordering. I mean, outside of that, there really wasn’t much innovation in the industry. And then all of a sudden, we went through the whole You know, zero interest rates and all these tech startups got funded and solutions just started popping up left and right. I mean, it’s like the golden and those those solutions have now grown up. They’re series A companies, and there’s a ton of them. And everyone’s like, Okay, this was great. We really enjoyed having all this new technology, but we bought everything. And the 80 SaaS vendors we have are. probably unnecessary. We can start to consolidate. And, you know, the scapegoat is, you know, it’s the SAS pocalypse. It’s not that we were a feature the whole time.

Deric Rosenbaum (55:27): I mean, i if you really think about it, it i if you look at it in the layers that I’ve been talking about, you got the transactional, the brand, the operations, and the synthesis. That operations layer is where all these SAS products reside, right? The synthesis layer is new. It’s coming. It’s it’s here-ish, not not widely adopted. The difference is is this the brands that have gotten the one and two right, and the four is coming, everything’s shifting around three now. So it it’s not that it’s not that it’s changed, it’s shifted. And you know, if you’re not shifting with this change, then yeah, you might be a victim of circumstance.

David Corts (56:09): Yeah, and and whether without AI and whether without the pandemic, which I think fueled a ton of that investment in restaurant tech, and I think it fueled a lot of it at absurd valuations. Like whether without AI, a lot of those companies are going away anyway. Like each industry goes through its sort of digitization moment asynchronously, and restaurants had its moment right around then. and I think a lot of companies got funded in that moment and I think a lot of them are overvalued and a lot of them aren’t gonna survive with or without it.

Matt Wampler (56:48): Yeah, you know it

David Corts (56:49): It happens in every industry and every cycle.

Deric Rosenbaum (56:52): And that’s the advantage of like fresh KDS. Like, you know, it sits in that middle layer and it saw the shifting and now I’m not gonna say it’s the brain of the kitchen, but it it is it’s the conductor of the kitchen, right? I mean it’s not just

David Corts (57:07): Yeah.

Deric Rosenbaum (57:07): it’s not just displaying a ticket, right? It can do so much more than that now. It can display my ClearCOGS reports and my in my inventory reports, or you know, you can strike through as you go, or it it builds you future screen so you can say, Hey man, I got sixteen more bacon and turkey clubs coming. I probably need to drop forty pieces of white toast, right? Or whatever the case may be. So I mean

David Corts (57:27): Yeah. And and what we’re where we’re headed, we’re we’re trying to participate in that synthesis layer that you’re talking about, right? Because we’re we collect so much information and so much data in the orchestration, the stuff you just described, and we wanna be able to synthesize that to make your operation run better.

Deric Rosenbaum (57:49): I mean i if you

Matt Wampler (57:49): Well, it’s distribution. I mean, at the end

Deric Rosenbaum (57:52): if you think about it.

Matt Wampler (57:53): of the day, it’s distribution. If you’re gonna have your your centralized intelligence system and you’ve got all this great information, it needs to get into the brain of a human at some point. You know, where else do you distribute it?

David Corts (58:05): That’s right. It’s right to that screen or to Derek’s phone or to another vendor or to a supplier or to a machine. But right now it’s right back to the screen,

David Corts (58:19): right? It’s right the action. Here’s what’s about to happen. Be prepared for it. Reduce cognitive load.

Deric Rosenbaum (58:27): with the explosive growth of Omni channel and you know, all these different options for our guests to order from, right? I mean, where does everything flow through in the kitchen? The KDS. So how do you solve for universal capacity management? The only place you can solve for that, I think, is at the KDS because it’s the only thing that has the real time bump data going through your kitchen, right? So you got orders coming in from Doordash, Grubhub, Uber, your native channel within your four walls. And yeah. You can regulate a little bit of that, but without the bump data to know the actual output of your kitchen, you can’t calculate these things. So that’s part of the synthesis layer that I think is coming. So he takes that data, then shares that back with Square, and then we adjust our capacity or our order, our load balancing or order throttling based on real time situations to provide a better experience to our guest and a more reasonable expectation.

David Corts (59:24): Yeah. That I mean that’s right. We sometimes we say the POS knows what was ordered, right? The KDS knows what happened. It’s the actual execution and it really matters.

Matt Wampler (59:35): So this was an interesting experience for me, having been an operator prior and coming into the technology side of things. The people that, you know, I know that have been doing it longer than I have are very jaded about working with other vendors. It it was almost like a hostile industry. And you know, you walk in, it’s almost like we walk into a room and everybody was just having a fight and you can like feel the tension. It it feels like this vendor-to-vendor relationship has always had some tension. I actually want to talk about the human element because, you know, we were at whatever conference and just started talking about this. And Derek started talking about how he wanted everything connected and information in front of his staff to make decisions. And we all said, Hey, let’s let’s get together and make this happen. And it’s been without question the best partnership we’ve ever had. I I’m curious from your guys’ standpoint, why why does it feel like it’s so hard to get restaurant technology companies to work together in harmony and why there’s all this friction.

David Corts (1:00:43): You wanna go first or you don’t

Deric Rosenbaum (1:00:43): I I

David Corts (1:00:44): mean to

Deric Rosenbaum (1:00:45): I I I’ve got a lot of opinions on this. Yeah,

David Corts (1:00:48): Yeah.

Matt Wampler (1:00:48): Hit me.

Deric Rosenbaum (1:00:51): I I think for so long it was controlled by so so few. what you know, i if you really think about it, like cloud based POS solutions have been around maybe ten years, like it it it at a broader adoption rate, right?

Matt Wampler (1:01:05): Which is insane, just a fact on its own.

Deric Rosenbaum (1:01:09): Correct. And prior to that you had these legacy, you know, server client models and there was a five to ten major providers with maybe three really coming to the top, right? And everything was built in a wall garden because they wanted to sell you everything. Or connect you with their partners because, you know, they had some kind of rev share whatever thing going on between these two vendors and They basically built this closed loop model, and for years and years and years, brands’ tech stacks were dictated by their vendors. I think we’ve flipped the script. 100% I think we’ve flipped the script. Your tech stack or your operating system should be dictated by your brand’s objectives. Period. And I don’t believe in best of breed, I believe in best of fit. There’s a lot of people that do a lot of things. There’s several KDS providers out there. There’s well, there’s not too many of you out there. At doing what you’re doing, Matt. But you know, there there’s options, right? So, what is the best fit for your brand’s objectives today, tomorrow, and in 10 years from now? And I think that’s the fundamental shift. And that these legacy providers if we as an industry raise up against this and say, you know, we’re not going to tolerate these walled gardens and this this ecosystem that you I mean, the same thing happened in In broadline food distribution. Like, you know, they have no choice but to adopt. And the forward-thinking vendor partners that we’re working with understand this and realize that it’s in their best interest in the long game to open these things up and bring in best of fit. I mean, that’s ultimately why I chose Square. I mean, they have an open API. I don’t I have yet to find anyone that will tell you that Squares API is not one of the best on the market. Like we can build anything to it. I can connect anything to it. And you know, that I don’t have to go through seven layers of hell to connect clear call, I mean fresh KDS on an API level at my store. We do it ourselves. Like it’s self-service. And that is the new, I think, the new modus operandi and how. We have a responsibility as brands to accept to to accept data

David Corts (1:03:34): Yeah.

Deric Rosenbaum (1:03:34): standards and to man best of fit for our brands. Sorry, I’m off my soapbox.

David Corts (1:03:42): I agree with I agree with everything he

Matt Wampler (1:03:44): Appreciate it. It was good.

David Corts (1:03:47): everything he said is right. And I sum it up with something I said earlier that I’m just a big believer in in my 30 year career in tech. Open beats closed in the long run. Every time. And I think vendors that don’t want to collaborate with other vendors are clearly putting their needs and their financial interest above the operators. And that might feel good and get you a few more bucks in the short run, but in the long run, the operator’s the one doling out what is not a lot of money. Like that, you know what I mean? We’re all fighting for the same wallet share that Derek has. And it’s not like you’re selling to anthropic where they can just buy whatever they want. And so there’s competition there. But I’m a big believer if you’re open and you’re collaborative and you put the operator’s needs. above your own short term financial interest or your own moat if you think that’s the way you’re really gonna build a moat, then you’re not gonna make

Matt Wampler (1:04:53): Yeah, I also think there’s this extractive element to a lot of restaurant technologies where they’re fighting over finite dollars that restaurants have for tech spend. I I I think I’ve said this multiple times, but like if you were to look at restaurants as an industry, like as an index, you’d say it’s worth X trillion dollars. And it’s based on the forward earnings of those companies. You know, you can Increase revenues at the expense of another restaurant, it doesn’t really change the value of the index. But if you can increase the profitability of those restaurants, that index would be worth, you know, X trillion more. And so like coming back to efficiency, if we can just get restaurants to be more profitable, man, it it’s not a zero sum game and just a better world.

Deric Rosenbaum (1:05:48): I mean, unfortunately a lot of the major players in our space are publicly traded. So, you know, it’s makes it a little more difficult for them to step up and play ball at this level.

Matt Wampler (1:05:57): Doesn’t change the fact that if they are to increase their earnings, then their stock price goes up, you know? Although I guess their earnings

Deric Rosenbaum (1:06:02): Correct. Facts. Facts.

Matt Wampler (1:06:04): are based off of royalties, so they just care about revenues going up.

Deric Rosenbaum (1:06:09): Depends on the company.

Matt Wampler (1:06:11): All right. So I just to kind of close things up, you know, we we talked a lot about having this semantic layer that provides information to your team, you know, so that, you know, you’re omnipresent. I think the hope is that that means you’re freeing up mental capacity to not be looking at dashboards or metrics or figure out what to do next and are able to spend more time on the human aspect of it. IQs are going down. Are we getting stupider? Are we just eventually

Deric Rosenbaum (1:06:41): Yeah.

Matt Wampler (1:06:42): becoming like slaves to our AI overlords? Or is that like an unintended byproduct of I don’t know why I’m doing what I’m doing, but I was told to, so I’m just gonna do it.

David Corts (1:06:56): I was gonna ask Derek, does AI make you feel stupid or smart?

Deric Rosenbaum (1:07:01): I was built for AI. I mean, period. Like my brain works this way. I I I think it it really distills down to explaining the why. Like what why are we doing this? Why are we building this? What are we gaining? What are we what are we discerning? How are we, you know, it it part of my job is of bringing change to a brand that’s four generations old, right? I mean, we’re celebrating eighty five years this year, is and you know. bringing that into this modern operating structure. I mean, hell Grouchos was around before the freaking Interstate system, man. So like i i i if I can do it and we can do it, I I think anybody can do it. And, you know This is one of the hardest industries I on the operator side. And you know, you you have to commit to excellence. And it and if you don’t, I just don’t think you’re gonna find it. And explaining the why I have found goes a long way. And then also bringing the proof.

Matt Wampler (1:08:03): David, any thoughts?

Deric Rosenbaum (1:08:04): You’re you’re not gonna get buy in out otherwise.

David Corts (1:08:06): I mean I I think it either makes you a lot smarter and a lot more empowered or it lets you cut every corner in life. And I to me it’s the most exciting time I’ve had in a long time, maybe in my whole career. because I I think I’m like Derek, I am so built for this. I’m a liberal arts guy that’s been trapped in tech for his whole career by choice, because I love it. But I’m wired to ask the next question. I’m a big believer that it’s the the the question is a lot more important than the answer and knowing the next question to ask and how to connect it. And now I have this little box in front of me that is the corpus of all human knowledge and I can ask it anything I want to and follow up with it. I mean, it’s incredible. And then when you add its capabilities to do things. It and then you add the ability to execute on those things, it’s incredible.

Matt Wampler (1:09:05): I I’m on the kick of what or why? If you ask

David Corts (1:09:08): Yeah.

Matt Wampler (1:09:10): what, like what do I need to do next? It’s gonna make you dumber. And if you ask why, you get smarter. Yeah. And

David Corts (1:09:18): How does this work? I mean it’s just yeah.

Matt Wampler (1:09:21): I think the funniest one is watching my, you know, nine, ten year old daughter. When you’re a kid, what’s the thing that always annoys parents? Your your kids are always asking why.

David Corts (1:09:33): Or wired desk, yeah.

Matt Wampler (1:09:35): we’re wired to ask. And so I I think it is funny when you when you give AI to an adult that’s just trying to get through work and their day, they’re just trying to get efficiency out of it. And it may make them a little dumber because they’re not having to critically think, but man, if you’re a 10 year old that is looking at the world in wonder and wondering why everything works, it is just a, as Steve Jobs liked to say, a bicycle for the mind.

David Corts (1:10:02): Yeah, no, it’s incredible. And it’s so funny. I talk to people sometimes that say, I don’t know how to use AI, or I need to take a class or I need to talk to somebody. And I’m like, the most ironic answer you could hear. You have a box in front of you that just ask it.