Matt Blumberg on Acrolinx Rebranding to Markup AI and the Launch of Content Guardian Agents

Hey everyone, it's Alan Hummel. Welcome back here to Tech Drunk tv. I'm really happy to have my next guest back on.

You know, I, I know this gentleman, geez, 15, 18 something years. We, we shared an office together at, early on when previous company Matt had started, and I had co-founded, um, Matt Loberg, CEO markup ai. Matt, did I get everything right there?

You did, absolutely. Very cool. Good to see you.

Um, good to see you, Matt. So, you know, we were just talking off camera and I couldn't believe it was that long ago that we talk. I felt like it was maybe two or three months.

You're saying it's six or seven, but it's all good. Gives us more chance, more to catch up on Matt. I, I mentioned, uh, you know, we know each other a long time.

You were a CEO of a company, another company that you had, uh, founded. And, and you know, actually, why don't, I'm not gonna tell them your story. You tell them your story.

Uh, well, yeah, I think you, so you and I met, uh, when I was running Return Path, uh, which Yes, uh, uh, which was an enterprise, uh, global enterprise SaaS company in, uh, email data and analytics. Um, which I started in 1999, uh, last century feels like last century. And, uh, it was, we sold, uh, sold that company in 2019, and I've done a couple other startups before that and after that, and now, uh, very excited to be at the helm of, uh, uh, of, uh, the company formerly known as Acrolinx.

And, uh, as of September 17th is now Markup ai. Very cool. I love it.

So, so September 17th, that's fresh. Uh, It's coming up to you right? As we, as We, as we record, it's next week as people hear this, it'll be probably right after.

Yep. Um, so Matt, give us the scoop, why the name change, what's going on? Yeah, we're very excited about this launch.

We, we feel like we're launching a new category as well as a new platform and, uh, a new brand, uh, to go with those two things. Um, the category that we are, uh, so proud to, uh, to spearhead is what we're calling Content Guardian agents. Uh, so the origin of that name is, um, is actually Gartner, um, you know, big, uh, analyst group, um, that have increasingly been talking about this concept of guardian agents.

And, uh, you know, it's obviously a play on guardian angels and, uh, one, one for the AI ages here. Uh, but the concept of a guardian agent is very simple. It's an AI system that is designed to oversee another AI system.

Uh, so if you think about, if you think about it, AI systems do things so fast and at such volume that it's just not possible, uh, to hire enough humans to oversee every action of the system. So you actually need a different kind of AI system, uh, to oversee, um, an AI system and to kick things out. So it's sort of this human in the loop, um, motion, um, for things that are aberrations or don't look right.

So that's the concept of a guardian agent. And what we are looking, um, is, um, what we're calling a content guardian agent. There will be guardian agents that do other things, for example, things around security, uh, but content guardian agents are really, uh, going to be, uh, critical for sort of the, the whole architecture or scaffolding of generative ai.

Um, so if you think about it, companies are adopting generative ai. Most companies are using multiple models, multiple departments are using multiple models, um, to produce more and more content every day. Um, and our new platform at markup ai, um, we think is gonna be the industry standard for overseeing the quality, uh, and accuracy of content coming out of generative AI systems.

So content Guardian, I love it, is the category markup is the brand because what do you do when you wanna fix a document? You mark it up, lawyers mark things up, editors mark things up, developers use markup languages. So, uh, we're excited, uh, to bring our, our new brand and our new platform to market.

Matt, that's a, uh, you know, this is this, you know, sometimes you hear things Matt, and you say, well, how come no one thought of that before? You know, that's a good idea. Why didn't I think of that?

But, but it, it is, it's, it's just so common sense that, you know, of course we would want this, you know, we were talking off camera and I was telling you I was out in Napa last week or, or this week rather, and I was there for this jfr Swamp up, uh, event. And, and they, they released something called AI Catalog, which all it does, and all, it's, it's still a big thing, is it goes through and finds all the ais, all the gen AI models and, you know, running in your infrastructure Okay. Sort of going after the shadow AI problem.

Yeah, yeah, exactly. And Then documents and catalogs, and then theoretically you can decide, I want this one, I don't want this one, Don't want that one. Yeah.

And, and that's a value. I'm not saying that's not valuable, that's a valuable piece of functionality, but what you are doing here is you're going one better, if you will, in saying, Hey, we're going to, you know, people are going to use ai. There, there, there's no, you know, you could lock the door, but they're coming in through the windows.

But when you're using ai, we're going to watch the content you're developing with it. We're gonna make sure that that content is okay. It's right.

It's not patently false, it's not somehow illegal. It, it doesn't violate policy. Yeah, exactly.

I mean, that, again, the, the sort of, the concept of the Guardian agent is that Guardian agents do three things, review, monitor, and protect, uh, and kick things out for human in the loop review. So that's what, uh, the markup, uh, platform, uh, is, uh, is going to do for, uh, for content, um, monitor, review, protect, uh, and help companies really confidently and, and rapidly scale their use of generative ai. Got it.

Um, now it works across all of kind of the frontier models, the usual suspects. It it does. Yeah.

I mean, it's, um, our system is very easy to use. It's a series of APIs, uh, and, uh, the, the system itself is a good blend of, um, of using large language models ourselves, using the output of whatever the client has, and then blending that with deterministic rule sets. So, so companies have deterministic rule sets that can be quite complex.

They have brand guidelines. Sometimes those run hundreds and hundreds of pages long. Um, they'll have terminology dictionaries that could have tens of thousands or hundreds of thousands of entries.

Um, they'll have policy handbooks, uh, they'll have laws and regulations that they're subject to. And our system is, is very good at instantly ingesting all of those deterministic rule sets and blending that with a large language model so that we can scan and score and then rewrite content instantly. Um, well, if you think of the use case, right?

A, you know, companies produce, um, you know, hundreds of thousands of pages of content a year, they might have millions of pages of content that are active online. What do you do when you come up with a new policy or change a term or have a brand guideline? How do you fix 6 million pages instantly?

You push a button and our system works in batch mode and flags and fixes all the mistakes. Hmm. What do you do when you're creating a new piece of content and you want to check it somewhere, but you don't want to have to go send it to a lawyer and then send it to someone on the brand team and then send it to someone else, and then what happens if they don't agree with each other?

All that is streamlined through one very easy to access API Got it. Now, I'm assuming you're, you are using some sort of AI on the backend to do all this, obviously We Do, yeah. We're, we use large language models in, in conjunction with, uh, deterministic rule sets and a vector database and everything else.

That's how our system operates. But we can obviously work against any other system that sends us content. Excellent.

And now how would, how would you sell this by the amount of agents you deploy, by the amount of content you're monitoring Can buy the token? Like, like all good native ai ai, Right? That's the way, that's the world.

We're moving to the token world. ai. Uh, you can do a trial for free.

You can, you know, our, our entry level package, uh, once you get past a a free trial is $200 a month with a credit card. So if you're a developer and, uh, you wanna play around with the, uh, the two APIs, one for checking and one for rewriting, it couldn't be simpler. I love it.

Sounds great. Um, and, and you know, Matt, the other thing I wanted to emphasize is, look, if you've got millions of pages of content, hundreds of thousands of pages of content, this is great. But you don't have to have No, you don't.

This isn't, you know, just for the, those folks, for those people, no. Isn't, Isn't, yeah. And then that, right.

That's, that's actually why we took the approach of deploying this very simply with APIs. Uh, and there's also low-code, no-code options. You can find us on Zapier, you can find us on N eight N.

Um, so we're, you know, we wanna make this available to everybody. And you know, pricing is on consumption, so you can, you can use it at small volume. Look, the reality is, yeah, big companies need a lot of it, but small companies need a lot of it too, or need a little bit of it, and they should be able to get that.

So it's, you know, it's the beauty of deploying systems these days, uh, that are built on top of large language models and, and, you know, very, very simple deployment is, uh, it can be used at small scale or at large scale. Excellent. I love, I love it.

Love it. Um, begs the question, what about Ag Agent ai? Is this a form of ag ai, you think?

Yeah, for For sure. Yeah. I mean, the way, the way we're talking about the product, we, uh, are launching on September 17th, or launched on September 17th, depending on when you're listening to this, is, uh, that is our, um, our brand guardian agent.

It is agentic ai, it uses generative AI in the background, but it's really an agent, uh, that is, um, the, the brand guardian agent is actually a bundle of individual agents that do very specific tasks, which is kind of the definition of an agent. So there's a terminology agent, a consistency agent, a tone agent, a clarity agent, and a spelling and grammar agent. But all of those operate in one API.

So this is very, very much ag agentic AI with the power of large language models in the background. I love it. I know you mentioned the website, but mention it one more time.

ai. I love it. Matt, you know what, how long are you there?

Seven, eight months. You changed the name of the company, shook up the whole thing, started ingen ai, uh, product offering. I'd expect nothing less from you.

Well, it, you know, we've been able to do, do this as quickly as we have, partly because we have a, a superb team, and partly because, um, the foundational models and the tools around them are very powerful and easy to build. And, uh, but also because, uh, the history of, um, uh, of our legacy company, Acrolinx, uh, we've been doing similar things like this for large enterprises for over 20 years. So there was a tremendous, even though, um, this is new technology, there was a tremendous amount of institutional know-how about how companies interact with content, what their content supply chains look like, what their approval processes are like, what they care about.

Um, so that, uh, you know, sort of the, the, the, um, uh, legacy knowledge combined with new technology lets us operate at very, very quick speed. And it's also gonna let us, uh, produce, um, the next set of guardian agents around content. So this one is brand, but the, the ones that are coming over the next few months are gonna be policy risk and regulatory.

Uh, there'll be, uh, content optimization and accuracy. Uh, there'll be data leakage. So you can just think about anything that could go wrong with content.

Um, you know, we, we will be the one-stop shop to make sure that, again, it's this concept of the guardian agent. It's overseeing your AI systems to monitor, review, and protect. Excellent.

Excellent. Hey, Matt, I promised I'd get you outta here on time. I know you have another meeting.

We're about out. Congratulations. Keep up the great work.

Let's not wait five, six months till you're back on though the way you're rocking and rolling here. You're gonna have news next month or something, so happy to talk. Stay in touch any Anytime, Alan.

Always A pleasure. All right. Matt Bloomberg, CEO of markup, ai, the newly renamed markup AI launching the Industry First Guardian, a content guardian agents.

Stay tuned for that. You're watching. Text Drunk tv.