Hitting “publish” used to feel routine. Now, when a prompt writes the first draft, that button can feel like a gamble. We sit down with Matt Blumberg, CEO of Markup AI, to talk about what’s really happening inside enterprise content teams as large language models crank out documentation, support articles, and marketing copy faster than humans can reliably review it. Matt lays out a simple framing that sticks: Markup AI is “AI to check your AI.” We dig into what content guardrails look like when you take brand voice, compliance, and accuracy seriously, and why “AI slop” is the one thing every client agrees they cannot ship. We also explore the overlooked frontier of AI search visibility, where a new rulebook is emerging for how content gets found and cited by tools like ChatGPT. The conversation goes deeper than product. Matt shares the behind-the-scenes reality of reinventing a 25-year-old NLP company into an AI-native, multi-agent platform, including failed first attempts, the need to integrate into real workflows (Docs, Word, Figma, APIs), and why enterprise software still matters even when you can build a clever prompt. We close on lessons from turnarounds, the new pace of shipping, and how to avoid product sprawl when code becomes cheap and fast. If you care about enterprise content governance, LLM guardrails, compliance risk, and the future of AI-driven content workflows, this one will sharpen your thinking. Subscribe, share with a teammate who owns “publish,” and leave a review with the guardrail your team needs most.
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
SPEAKER_02Someone said to me once, I can't remember who, or if I read it somewhere, the future is already here. It's just not evenly distributed yet.
SPEAKER_00Okay.
SPEAKER_02Yeah. Yeah. Yeah. Yeah. We get a really different answer. If we're we have clients we talk to who are early adopters who are like, oh yeah, yeah, I've already built that. And then it turns out like they've built like 10% of it. And we have clients we talk to who say, like, oh yeah, we're not allowed to use LLMs. So it's really, it's just a very wide spectrum of these things. What I would say all clients have in common is they all know that there's something called AI slop and they don't want it.
SPEAKER_00Today on Customer Land, I'm with Matt Blumberg, who is CEO of Markup AI, which we'll learn more about in a moment. A couple of things you should know about Matt early on. Matt's got a giant history in this space with companies you absolutely will have heard of, a handful that you may not have heard of, but his his experience as a CEO in this space, I think is going to lend something to this conversation that goes beyond just an operator. Because our space, however you define it, is expanding and morphing and reshaping itself kind of at the speed of sound right now.
SPEAKER_00And Matt's got an awful lot to talk about here. So with all that, Matt, thanks for joining me. Absolutely, Mike. Good to see you. So perhaps you can do a better job than I did of describing what markups about. And I want to talk a little bit about your history too, because I think that's important.
SPEAKER_02Yeah, sure. So this is my uh fourth tech company going back to the mid-90s, uh, which uh I realize makes me sound like a dinosaur. But uh, you know, all all internet going back to 1995, you know, one of the very, very first uh uh websites. So that uh that one was a company called Movie Phone. Uh it was what it was a top 50 web property in 1995 that was the internet's first movie, you know, movie website, and and some real groundbreaking stuff in in 1995. It was real-time aggregation of of uh movie showtime data across the whole country.
SPEAKER_02Uh, it was the first one of those. It was uh the first time you could buy a movie ticket online, you know, first time you could, you know, do collaborative filtering and a bunch of a bunch of terms that don't even really exist anymore. One of the very early email commercial email applications we built called Movie Mail. Uh so a bunch of things from 95 to 99 uh in that context. In 1999, I started a company called Return Path, uh, which grew over the course of 20 years to be a fairly sizable player in the email marketing space. We had about 4,500 clients at the end.
SPEAKER_02It was a big global footprint, and we had a really unique angle on the email business, which was deliverability and certification. So we were not a delivery business, which of there are you know hundreds of platforms out there from constant contact on the low end to Salesforce.com, which used to be exact target on the high end and everything in between. But we were um a specialized analytics business. And we had this really powerful program called Certification, which was basically the only whitelisting program that helped filters and security companies understand what was legitimate mail and what was spam.
SPEAKER_02So big email marketing business for 20 years. Then after we sold that company, a group of us started an executive recruiting marketplace called Bolster, which was designed to reinvent executive search for tech startups because tech startups are frequently incompatible with the speed and cost and uh sluggishness of executive search. We did that for about four years. And then uh Markup AI is vintage 2025 for me, so it's only a year and a half in.
SPEAKER_02And it's an interesting story and a different story in some respects. Um, I didn't start the business, but I was hired into a legacy software business to reinvent it as a native AI company. So I brought in a new team and we started a new platform from scratch, which we just launched about 90 days ago, that's doing some really interesting work in AI and building on the legacy of a business that's been around for 25 years doing natural language processing and some other related businesses, uh related products.
SPEAKER_02So that's the arc of the story.
SPEAKER_00Uh so you've just generated 30 or so questions that probably really need to be answered before we go much further. But maybe what we'll do is this give us a high level of what Markup AI is, and then we want to leave that there, jump back a couple of steps and talk about what the company was and how and why it got to this point. So I think that'll set us up for the rest of the conversation.
SPEAKER_02Yeah, that sounds good. So Markup AI, we call ourselves content guardian agents. So we are guardrails for enterprise content. We are AI to check your AI. Okay. And, you know, at this particular moment, so we're recording this on September 30th, 2026. You cannot pick up a newspaper or listen to a podcast where people aren't freaked out about AI running amok. And what people are talking about is AI destroying humanity. And to be clear, Markup AI does not solve that problem.
SPEAKER_02And we could talk about that problem if that's interesting. But um we do know that AI is uh is imperfect and AI kind of goes off and does its own thing sometimes. Companies have become more and more dependent on the large language models to generate their content, whether that content is like product documentation or customer support, or more marketing and customer care and and um outbound communication to customers. And you know, AI does a great job of getting your content 80% right fast. And it does a really terrible job of getting it 100% right every time.
SPEAKER_02And what we've found is that more and more companies have cut back on the number of people writing content because they said, oh, AI can do it. Right. But now they got a big quality problem on their hands. So Markup AI is a multi-agent platform that does four things for your content. Um, it makes sure your content is on brand, it makes sure your content uh is compliant with uh the relevant laws and regulations, it makes sure your content is accurate, and it makes sure your content is discoverable by AI search. And it does that in a multi-surface way.
SPEAKER_02So you can use our MCP server to integrate with your models at the point of content creation. We integrate with applications like Word and Docs and Figma and PowerPoint and Adobe so that you can use it as kind of an interactive sidebar while you're editing or approving content. Um, and there's a really robust API that you can bake into any workflow in your organization. So that is what uh Markup AI does.
SPEAKER_00Okay. All right. Now let's talk about what it what this company was. And I think what what I'm most interested in, what I think our listeners will be most interested in, is the thought process that went from okay, this is what the canvas looks like right now. These are our the components we have to play with. This is where we want to take it. What was that thought process like? What did you have to find out in those interim steps to get you to where you are?
SPEAKER_02Yeah, it's uh so how many hours do we have here? Yeah, it's it now it's an interesting topic. So this company is 25 years old. It had a different name. So, among other things, um, you know, reposition, new products, et cetera. It's markup AI is a new name. The company used to be called AcroLinks. And here's the interesting thing. AcroLynx was a carve out of something called the Berlin Artificial Intelligence Institute in 2002. Wow. So what we all think of AI as like a shiny new object, uh-uh, large language models are a shiny new object, has lived for a long time in different incarnations.
SPEAKER_02So this was a business that was created by some real pioneers with language models. These, you know, these are people who have computational linguistics PhDs. One of them, one of the founders, and founders are well removed from the business now. I didn't take it over from them. One of them is one of the engineering leaders at Gemini. One of them was involved in the in the white paper that produced the Transformer. Um, so like these are people that have been around AI for a really long time. And in 2002, the only technology was available based on compute was natural language processing.
SPEAKER_02So everything I described about Markup AI today, one quarter of it is what the legacy system did using NLP, which is it kept your content on brand. Interesting. So it was a system that got trained with a very painful, long, human-oriented process on a company's brand. And uh a lot of the legacy clients that our company served have 700-page brand books. Like these are really large companies with complex brands and multiple brands, and took a long time to onboard a customer, create almost a custom NLP model.
SPEAKER_02And then the NLP model was used to keep human authors in check. Because a human can't keep a 700-page brand book in in her head or in his head. So the the journey um started with uh you know, with a business that had been around for a very long time with a lot of big customers that had been customers for you know 10, 15 years. And and essentially that the the product was was brand consistency. Yeah, exactly. Okay. Yeah, it was content governance. It was called the same, not the same thing we call it now, but yeah, brand consistency. Okay. Um along come large language models and start to disrupt the business.
SPEAKER_02And they start to disrupt the product, they start to disrupt the buyers, the customers. You know, just to hear one, you know, kind of cheeky example, but the the uh legacy product was sold on a per seat basis, and we had a whole bunch of clients that fired 90% of the people that wrote content. So we woke up one day and we're still doing the same amount of work for them, but they're paying us 10% as much because it's per seat. Right. So the business was just being disrupted and had a had a real claim to deep expertise in enterprise content workflows, you know, had figured out how to build integrations into like a hundred different really quirky systems.
SPEAKER_02Um there was a lot of of kind of richness to the business, but it really needed to be updated to put it mildly, into something that fit the contemporary landscape and stack. So um, you know, the board brought me in, I brought a new team in, combined new team with the old team. So we preserved a lot of that rich history and legacy. And it took us a year to figure out exactly what we needed to do. So you're right that there was a journey there, and not necessarily an obvious journey either. You know, we started off by saying, which I think a lot of people did two, three years ago, by saying, oh, well, let's you know, just look at the old system and we'll add some large language models to it.
SPEAKER_02And we sort of took a look at that and we're like, well, that doesn't really make sense. Like it's a really old system, it's very expensive to operate. Um, you know, sprinkling in a little AI doesn't, you know, doesn't make it a contemporary stack.
SPEAKER_00Did you know where you were headed at that point, just trying to figure out how to get there? Or was it still a function of what can we do with this thing we've been given?
SPEAKER_02No, it it it started with what can we do with the thing we've been given. Then it was really clear that the priority was to build something that that served the existing customer set so that we could retain that business and move it forward into the into the future with us, you know, while creating something that was compelling for early adopters. So it was a journey and it took us a while. We took, you know, one shot. Uh, you know, I'm a believer you got to take lots of shots on goal and they don't all go in. And you know, the first shot on goal didn't quite work. You know, we said, oh, well, we're just gonna redo everything from scratch, and uh, you know, that'll be you know, that'll be fast and that'll be interesting, and we'll just make it an API and everyone's a builder, and you know, so we got something out pretty quickly, and the uh it was rejected by both the new uh customer set and the old customer set for different reasons.
SPEAKER_02So the new customer set looked at it and said, great, but I don't only want an API. Like I need something that integrates with my existing uh authoring systems and I need an MCP server. And it was kind of rejected by the old customers who said, Hey, I actually like the output that I get from you today. Don't give me new output, just give it to me in a new framework. Um, and then yeah, if you want to add some new stuff to it, add some new stuff to it, but like don't mess with the thing that works. So um, you know, we learned a lot from that first shot on goal, and then you know, started again.
SPEAKER_02Uh and you know, the good news with building now using AI tools is you can do that, right? It does the cost of development is pretty low. So it took us a bunch of months to get to that point. Starting this year, uh, hit the ground running, built a new stack in six months, launched it in June, um, and it does what needed to be done. So for the old customers, it preserved their functionality, but moved the brain into a new body for new customers. It started scratching some new itches as well.
SPEAKER_00So I I'm I'm assuming, tell me if I'm wrong here, but the the company that you took over was still cash flowing fairly positively. You weren't you weren't running after a a runway per se, so much as just trying to reconceive what you had. Is that a fair assessment?
SPEAKER_02It started that way, but then the disruption gathered steam. And that took the form of things like I said a minute ago, like we had a perfectly good customer that all of a sudden dropped by 90% in revenue while retaining the business. So the you know, the the business really started to lose steam and it and uh we've had to you know do do a lot of work that you would classify more as turnaround than anything else. Right. Um so uh it's been an interesting mix of things and and uh really interesting couple of years.
SPEAKER_00So here we are, uh as you said, uh last day of September 26. Uh Markup AI has been in market with the new platform. I'm putting your quotes out there. 90 days for 90 days. It's not a whole lot of time to assess market receptivity, but early world moves fast these days.
SPEAKER_02So I've heard uh yeah, we uh we now have half a dozen of our large enterprise clients that have migrated over uh from the old platform to the new platform. And I think we have 400 active users. We have a, you know, it's like any AI startup, it's uh you can self-serve, self-provision, don't need to talk to a sales rep, you know, click, click, click. And I think our fastest time to value now for a walk-up customer is two minutes. The system's pretty good at getting you in and getting you going fast. Uh so early returns are good, but it's early returns, and we know we got a lot to do. And uh my friend Orrin Hoffman, who's you know, real thought leader in entrepreneurship and and now in the AI space, he's a VC, former operator now, VC, and uh he was the founder of LiveRamp, among other things.
SPEAKER_02Um he put out this uh this great Substack article the other day that said the only competitive mode that exists anymore is how fast you ship new product. And he's right. Yeah, I don't think it's the only one, but I think it's a pretty darn important one.
SPEAKER_00Yeah, I think so too. And it and it's a massive wake-up call to people who have been building startups, you know, a certain way over the past, say, four or five years, maybe even two or three years. The pace is just just let alone the last 20 years or 30 years, right? Yeah, yeah. I I don't even know how you operate, how you survive with those kinds of paradigms floating around your head right now. Yeah, you can't. It's it's it's new, it's all new. Yeah, it's it's yeah, everything about it's new. Content is the next code.
SPEAKER_00I'm quoting, I'm gonna use that, but tell me what you mean by that in the context of this discussion and what it means for for Markup AI.
SPEAKER_02Yeah, it is one of my one of my mantras at the moment. And it's it's it's I would describe it more as like a thought leadership mantra than a customer marketing mantra. But um what I mean by that is a couple of things. One is um the number one enterprise use case for large language models is code. The number two use case is content. And you know, that tracks spend if you look at anything Gardner puts out or Forrester puts out, right? But what I mean by content as the next code is that because code is such a uh a richly developed ecosystem, there are lots and lots of uh of products out there and vendors and processes that are like pretty well worn at this point and well known and battle tested.
SPEAKER_02And if you look at sort of the life cycle of content of code production, there are tools that people use at every step in the in the process. Um, you know, there are tools you use for generation, they're tools you use for quality, there are tools you use for testing, they're tools you use for review and approval, security and compliance, publishing, uh, observability, rollback, et cetera, et cetera. And you know, you can, I won't go through them, but there are multiple vendors for each of those things. And some of them overlap across those different kind of stages of workflow. Um and if you look at the content ecosystem, one or two of those seven or eight stages are densely populated, but most of them are are not.
SPEAKER_02So obviously there are tons of content repositories out there, right? The the CMS world, plenty of plenty of uh uh vendors that are that are in that space. Content generation is obviously where large language models excel, but it's a lot of that stuff in the middle that's kind of missing. You know, there's sort of no well-worn path or series of vendors that do automated testing and verification and review and approval and compliance and you know, orchestration and observability and rollback and audit trail. And that's really the the area that Markup AI is trying to kind of stake out territory in, is really helping enterprises put together the scaffolding they need to zoom forward and use AI for content production safely, smartly, securely.
SPEAKER_02And you know, when you go to hit that green publish button, you don't have that moment of panic of like shit, right? You know, is it all okay? Because I didn't write it. You know, two years ago I wrote it, so I knew it was okay. Now I didn't write it. I gave it a prompt.
SPEAKER_00Yeah, right now everybody's a proofreader or an editor, if that. If that.
SPEAKER_02Yeah. So anyway, so that's what we mean when we say content's the next code.
SPEAKER_00Just to go a little bit deeper into the weeds there, explain what a content guardian angel is and does without without giving me a pitch for it. Yeah. What does it instantly? What does it need to know about the company? How does it flag problems? How does it moderate or qc all this stuff? It takes an awful lot of outside context to make those kinds of QC decisions.
SPEAKER_02And that's the great thing about AI. So when I say our business is AI to check your other AI, that's what it is. It has that context. And um, you know, I'll give you a bunch of examples. So agents that optimize for AI search visibility, there is a rule set out there. And this is the common theme is rule set, right? It's it's determinism outside context, rule set, however you want to refer to it. There is an evolving, an ever-evolving body of rules around AI search visibility.
SPEAKER_02What does it take for your brand to show up in Chat GPT? And just like there is a whole bunch of rules of the road for SEO, right? And for SEO, it's some geeky stuff like the meta tags and you know stuff you don't necessarily see. Um a lot more of it for uh AI search visibility is in how content is structured, uh freshness of the content, snippet readiness. So there's this whole list of rules there. So our agents that do AI search visibility will apply that rule set to a piece of content and essentially mark it up, which is the name of the company, right?
SPEAKER_02It's marked up for that content. Um so that's one example, but I'll give you some other examples that that I think will give you sort of the the some of the nuances of the business. So something that our old product did really, really effectively was um terminology. So small companies don't necessarily worry about this very much. Large companies can have terminology dictionaries with a hundred thousand entries. And let me give you one more that's uh that's probably uh more nuanced and will I think will will make sense in the context of of what we all know about AI and its strengths and weaknesses.
SPEAKER_02So uh one of the things we do is accuracy. Accuracy is a really tough thing, and it's not one thing, like it's a million things. No, yeah, it gets there. Well, so so one form of so so uh the the simple form of accuracy is link checking. Okay, fine. A lot of people can do link checking. We have an agent that does link checking, great. You're not gonna use that in a vacuum. Um, but it's a useful thing to have in a tool belt. But what about accuracy that helps marketers, marketing copywriters at a pharmaceutical company make sure that the claims that they're making in a piece of collateral for a phase three drug trial make sure that the claims that they're making match the drug specifications?
SPEAKER_02That's a context issue, right? That's a rule set and context issue. And and it's one of those things that uh that AI is great at doing, right? It can say, uh I'm gonna look at this thing against this repository. So that's that's what we mean when we say content guardian agents.
SPEAKER_00Okay.
SPEAKER_02And it's and then it's all of those together, and then it's understanding how those things work and interplay with each other.
SPEAKER_00So it seems to me that that Markup AI is going to have certain areas where your domain expertise or access to domain expertise is probably greater than others. If you're doing work in the pharmaceutical area, as you just mentioned as an example, like you kind of know how to navigate where the where the quality indicators and proof points need to come from, right?
SPEAKER_02Yeah, it it is interesting. People ask me all the time, are you a horizontal or are you a vertical? And the answer is we're horizontal or we are horizontal that is made up of verticals. So there are things we do that apply to everybody and things we do that apply to a vertical, or where vertical knowledge or ontology or knowledge of workflow is helpful. But there are also, quite frankly, most of the things we do are company specific. So, you know, and everything's configurable, and our agents will configure the things for you if you point it at the right, right direction.
SPEAKER_00So you mentioned an earlier part of this conversation. Conversation that onboarding or or the phrase you use was like, you know, time to value, I think. Fastest time to value of two minutes. Of two minutes. Yeah. That's not so bad. That that's pretty impressive.
SPEAKER_02Um, it's pretty good. So here's how that works. You come in, you are sign up for an account with your domain. While you're signing up, we go scrape your website, we come back to you and say, here's your brand footprint or your brand voice print. Is this right? Great. Check a document. Click here to check your first document.
SPEAKER_00Okay.
SPEAKER_02Right. Or if you have a brand book, not every company does, you know, click here to upload your brand book. Click here to upload the Claude skill that you built that's trying to do this on your own. So time to value on some of these things can be really, really fast.
SPEAKER_00I've got lots of questions around how this value proposition, set of value propositions, are pitched to and received by the enterprise world of users, of buyers of this. What what you've produced and what you're selling is a comprehensive suite of QC services for content. How is that received?
SPEAKER_00Is it is it kind of like, oh yeah, we we're doing this, we got to do this, uh, in some verticals, or is it kind of like, well, wait a minute, you know, can I just do this with a better claude prompt?
SPEAKER_02The answer is it's all over the map. And someone said to me once, I can't remember who, or if I read it somewhere, the future is already here. It's just not evenly distributed yet.
SPEAKER_00Okay.
SPEAKER_02And you know, you know, we get a really different answer. If we're we have clients we talk to who are early adopters who are like, oh yeah, yeah, I've already built that. And then it turns out like they've built like 10% of it. And we have clients we talk to who say, like, oh yeah, we're not allowed to use LLMs. So it's really, it's just a very wide spectrum of these things. What I would say all clients have in common is they all know that there's something called AI slop and they don't want it. No. Right? They all know that AI makes mistakes and they can't afford to have that.
SPEAKER_02So, you know, at the end of the day, we are enterprise grade software, and enterprise software exists for a reason. And yes, you can build a cloud school skill that does some percentage of what we do. But let's think about that for a second. How do you version control that? How do you deploy it across your enterprise? How do you make changes to it? How do you make sure that it's token optimized? How do you deal with the next model that comes out and backtesting it? How do you make sure it's properly integrated into your DAM and your marketing cloud and your, you know, your GitHub system and everything else all at the same time?
SPEAKER_02Like that's what enterprise software does. And, you know, it is true that you can build stuff on your own now that you couldn't build in the past. But, you know, the question is, should you? And uh, you know, it was interesting. I was in, I sit on a bunch of outside boards and uh I was in a board meeting the other day for a company that's not a technology company at all. They're services business, a financial services business. And uh and they do a ton of stuff with AI and they do a really good job at it, and they've transformed their business and hats off to them. Um and uh someone in the room said, Hey, we're a software company now.
SPEAKER_02And I said, Oh, that's interesting. Like, so who's in your hiring plan next year to do all the maintenance and to do the brake fix and to answer the support tickets and to do this and to do that and to do the other thing? It's like, well, you know, day is gonna handle all that. It's like, all right.
SPEAKER_00Not so much.
SPEAKER_02Yeah, maybe in some ways, but not all.
SPEAKER_00Yeah, yeah. So from your your description a moment ago of of how this is being pitched and sold, you're presenting these enterprises with a de-risking solution for their content. This is this is about risk mitigation, period. Giving them the confidence to hit the publish button. So in conversations with other technology providers, if you're selling technology, if you can attribute it to uh directly to an ROI or to some bottom line effect, good or bad, it makes your sale a lot more concrete to the people who have to approve it and ultimately sign the check.
SPEAKER_00De-risking uh a large content operation seems to me to be a little difficult to boil down to a a direct ROI or or de-risked amount. I know you have to have that. You you you know, you've probably got the calculation, the formula, you know, written on your foreign arm worse, where I'm mine would be. But how does that work in your world? Because I'm sure it's it's a question that gets asked.
unknownYeah.
SPEAKER_02Look, in my experience, enterprise selling even for the same product, uh, falls along something that one of my former mentors refers to as the fear-greed continuum. And you can find a buyer for the for the same product at very different places on the fear-greed continuum in terms of what motivates them. And it's your job as a salesperson to discover that and to use the right pitch. Um, and you're right, there are there are times where it's hard to sell insurance, right?
SPEAKER_02It's hard to sell risk mitigation. Now, you go to a client that that's already been fined by you know the by some consumer protection authority because they had some you know bad content, or you know, the regulation says that you know this drug has to be marketed at the sixth grade reading level, and it wasn't, like that's a little easier. But you're right, frequently you have to sell on ROI, and you can sell on ROI, like, hey, how many, how many people do you have reviewing content? How many people do you have writing content? Or you know, what's what's the value to you of making that person five times as efficient, efficient, 10 times as efficient? So you can sell on ROI.
SPEAKER_02The other thing that I've found with with um AI products, and I think this is this is true of me as a user and a buyer, as much as it's true of us as a vendor, is something one of my board members said, which is that uh sometimes you can measure the impact of AI in terms of ROI, and sometimes you measure it in the previously unimaginable. And that's a little bit different. It's you know, like if you look at the fear-greed continuum, it's not exactly on that. I think it may it adds like a third dimension to it. Yeah. Right. But you know, but there are AI products out there that you, I mean, you know, as a user, you like you use something, you're like, oh my God, that's magic.
SPEAKER_02And you know, then you're not in the conversation about am I selling insurance or am I selling ROI?
SPEAKER_00So there's a there's a lot to talk about here. I'm gonna make my own plug here. As somebody who produces content for a living, I deal with people who produce an awful lot of content on enterprise levels. This is a tool you should look at. I wanted to talk a little bit about how your perspective as a startup CEO has grown, more changed over time, and how that's kind of affected your method, if you will, for building markup AI.
SPEAKER_00You know, your experience at Bolster, which was a really interesting take on need in the marketplace and how to resolve it, didn't necessarily go the way you were hoping it would go. That's correct. Yeah. But you know, you've been through some very successful um companies and leadership roles. That one went a little sideways for you know market reasons, I'm sure. But leading these organizations and as CEO, you've got a giant remit.
SPEAKER_00I don't care how big or little your company is, you've got an awful lot that you've got to kind of tally up in real time to keep the company moving in the right direction. How has that changed over time for you? And how has it affected the way you envision growing markup AI?
SPEAKER_02I would say it's changed in in a couple ways. And there's one that's related to this circumstance, and then there's one that's related to the world of AI and how how the world has changed that I can point to. So, you know, this circumstance is a turnaround, as I said. And the conclusion I have even two years in is that every CEO should be required to do a turnaround sometime in his or her career, uh, because it's just a different game and it sharpens your skills in different ways. And, you know, look, if you're a CEO of your own company long enough, you have to fix stuff because not everything is right all the time.
SPEAKER_02But it's really different when you come into someone else's house and have to do cleanup. And you just look at things differently and you don't have that founder angst of like, oh my gosh, the thing I created needs to be torn down and fixed, which, you know, good CEOs and good founders know how to do, but it's it's like a mo it takes an emotional toll on you. Sure. And you have a lot more distance when it's when it's a different organization. I think that the process of growing your career over a long time is the process of like aggregating new lenses with which to look at the world.
SPEAKER_02And the turnaround lens is a fundamentally different one and and I think a really powerful one that I think, you know, if if my next thing is something I start, I will approach it differently, having had this experience of like looking at something from the, you know, from the outside a little bit more objectively. But that has nothing to do with the moment that we're in now with AI. And that's something about this particular circumstance. I think the moment we have with AI is that the way certainly the way software companies function is getting turned on its head.
SPEAKER_02Sure. And that's not circumstantial to this company at this moment. That's just happening everywhere. And there is not a CEO that I talk to who isn't grappling with this same problem. And it's it's the it's the other side of the coin of automating code production. So, you know, heads, good news. Right. Software is cheaper than ever, faster than ever, it's free. You can take an infinite number of shots on goal, you know, no problem.
SPEAKER_02What are you building?
SPEAKER_00Uh and will it survive, you know, two months down the road?
SPEAKER_02Well, and should it? Should you build it? Right. Um, so when the scarce resource in the organization was capacity to write code, the organization, and what I mean by the organization is software companies, right, oriented themselves around that constraint and said, how do we optimize every minute that engineering is spending writing code? We are going to know when they write code that they're writing the right code. Um, right. So everything was kind of oriented around that. But that's not the constraint anymore.
SPEAKER_02Now the constraint is actually the capacity to internalize customer needs and market dynamics properly because the code production itself has gotten so much easier and so much cheaper. So, you know, it's one thing to say, hey, I believe in taking shots on goal. We're going to take 100 shots on goal. That's all good and well until you realize you have this trail of of code everywhere and what we have started to call product sprawl.
SPEAKER_02And product sprawl will kill software companies at the end because it will confuse the hell out of customers and it'll exhaust go-to-market teams. Yep. And you'll end up with Frankenstein.
SPEAKER_00So raising my hand because I've seen it. I've seen it firsthand. Yeah.
SPEAKER_02So I think that's the biggest thing that we're seeing in this moment. Every company that's producing software is having to rethink how they do business.
SPEAKER_00Can I ask you how that gets worked out within within the four walls of markup AI? But you know, you have your strategy people, you have possibly maybe that's you, you have market understanding and insights, you have consumer understanding and insights. Mashing that all up at the speed that is required now to make coherent decisions that are strategically sound seems to me to be the the thing that that is most in need right now.
SPEAKER_00But but that's just me. I'm not I'm not leading your company. What do you think about that?
SPEAKER_02We have spent a lot of time thinking about it and talking about it. I don't know that we are doing, I don't know that we've gotten it right yet, but we have a theory that we just have to re-instrument some large chunks of the business. If you think about the the world of signal capture, signal capture has never been easier because everything is recorded now, right? Everything is transcribed now. And if you think about the world of information assimilation and pattern matching, et cetera, like that's a great use of an LLM.
SPEAKER_02So thinking about how you sort of re-instrument the workflow of an organization and the business process of an organization and using the tools that are now available to support that is sort of directionally how how we're trying to think about that. So um, I don't know that we've figured it all out yet. I think we have a theory about it, but um, but there are no tools to there are no packaged off-the-shelf tools that do this. So you gotta do it on your own. Someone will come up with package tools at some point.
SPEAKER_00Yeah, yeah, that's really interesting. That I think that topic in itself is worth another episode of this conversation because it it seems to be where the companies who are who are about to get this right are always it's a constant, I wouldn't say re-envisioning or rethinking, but it's it's kind of a real-time context aware of your place in the market and what you'll need to be doing near-term, medium-term, and who even knows about long-term.
SPEAKER_00Yep. Yeah, 100%. And uh so I I want to see, I want to have a conversation about how that's worked out with you because you have you build those contexts along the way, you build those paradigms along the way. And we were talking about that a little earlier. Like maintaining strategic curiosity over time, it's actually really hard to do and unless you're kind of built that way, I think. And you seem to be built that way.
SPEAKER_00So what's in the vitamins? Did you find the team that thinks like you? Or are you having to kind of piece this together domain expertise by domain expertise?
SPEAKER_02Or look, I there is there's certainly value in working with with some of the same people at mul in multiple contexts, uh, because you have some shared learnings. Um, but we have plenty of people here at the company that that um you know haven't worked with with us over the long haul. There's a lot to be said for intellectual curiosity. And uh, you know, one of the things that I always think about uh about entrepreneurs is there are entrepreneurs that are always thinking about the next thing to go build and the next thing to go do and the next problem to go solve outside the company.
SPEAKER_02And there are entrepreneurs that think about that, but also think about well, what's the next thing we do inside the company to make things better inside. And I don't know that one is better or worse than the other, but I'm definitely type two. So uh, you know, I I do spend uh a lot of cycles thinking about uh, you know, how we can do things better, how we can be better as well. I'll leave you with one thought that relates to your business and probably your audience as well, which is here's what hasn't changed. It's still really important to solve big, hairy problems for your customers.
SPEAKER_02How you do it may have changed, the delivery vehicle may have changed, the underlying technology may have changed, the pace of it may have changed. Um, but you can't build a successful business without solving your customers' biggest problems.
SPEAKER_00Biggest hairy problems. There you have it. Matt Blumberg, CEO of Markup AI. I can't thank you enough for this. And and I'm gonna stick my foot in the doorway there and say before it shuts, we really should be talking about a second episode here because there's there's there's richness. Let's do it. It was great talking to you, Mike.
Cashback is having a quiet comeback and it’s not just about saving a few bucks. I sit down with Carrie Paradas, North American GM at ShopBack, to unpack how a cashback and rewards platform can tap into something much...
AI shopping agents are about to rewrite the rules of retail, and the scary part is simple: your brand can lose without ever getting a chance to persuade a human. We dig into that reality with Talbott Roche, CEO and...
Gift cards look simple from the outside, but the machinery behind them is turning into one of the most useful building blocks in modern rewards and fintech. We talk with Alex Preece, CEO of Tillo, about why the “gift...
Shopping is turning into a conversation, and that changes what “good UX” even means. We sit down with Purvi Shah, Senior Vice President, User Experience at Target, to unpack how user experience works when it’s not just...
The sale isn’t the finish line. It’s the moment you hand your relationship to shipping networks, support queues, return policies, and whatever happens when a real human has a real problem with their order. I’m joined by...