The Rise of Enterprise Guardrails with Acrolinx’s Matt Blumberg
Matt Blumberg, CEO of Acrolinx, discusses the rise of enterprise guardrails as businesses seek a more robust approach to content governance.
Transcript
This is Tech Drunk tv. Hi everyone. Welcome back here to Tech Drunk tv.
My next guest is someone I've been trying to get on here for weeks now, but things got in the way, but I'm glad he's finally he who's persistent and his persistent paid off. Uh, Matt Blumberg. Matt is the CEO of Acrolinx, and we're gonna tell you all about them in a second.
But let's first welcome Matt. Hey, Matt, welcome to Text Drunk tv. It's great to have you on here, Alan, it's good to see you.
Thank you for having me. Thank you. So, Matt, we're gonna talk plenty about acro links, but let's spend a few minutes talking about you.
Don't be uncomfortable. I know none of us, like, well, some of us like to talk about ourselves. Um, but tell people a little bit about you.
What's your journey even like? Yeah, I, um, uh, I was, um, lucky enough in 1994 and 1995 to be, you know, sort of, uh, early years of career when the internet was getting started. Uh, so, uh, my career journey, I've, I've been the, the leader of, uh, four different, um, tech companies, uh, since 1995.
And, uh, four really different ones in four really different, um, pockets of the internet. com. Um, and those of you who've been around for a long time, remember seven, seven, seven film and the movie phone, telephone service, uh, and I joined the company.
Absolutely. com era. Um, after we sold, uh, movie phone, I started, uh, a company, uh, called Return Path.
Uh, return Path was, uh, uh, B2B enterprise SaaS business in the email space. Uh, and, um, ran and scaled Brad Feld, Sorry, Brad Feld. Brad Feld, uh, who, who, uh, is a, a close friend of mine and has been on a couple different with me.
Um, so, uh, yeah, Brad was one of my, uh, VCs there. And, uh, we scaled that business for 20 years. Um, it, uh, ended up as a, a pretty large global, uh, enterprise player.
Uh, we got acquired in 2019, uh, by Providence Equity, um, a rollup that they were backing, uh, in MarTech. Um, I started and ran a company called Bolster for four years after that, which was an AI assisted marketplace in the executive recruiting space. Uh, and, um, then I, uh, joined Acrolinx as the CEO in January.
So I'm a couple months in, uh, on a new job, and Acrolinx is, uh, in yet another space, the, uh, sort of AI, generative AI space, and I know we'll talk about that a lot. Um, but, you know, throughout my career, these four, uh, you know, very different businesses, they've all had some really consistent through lines about, uh, you know, business model disruption, uh, you know, harnessing the power of technology and data and data science, uh, to change industries. And, um, and a real through line for me has always been about, uh, people and culture and, uh, and leading businesses, um, in, uh, in a different kind of way.
Um, the other thing about me along the way, I've produced a lot of content myself so frequently, I'm on your side of the equation here. Yep. Um, have a, a great podcast called The Daily Bolster, which has about, uh, 200, 250 episodes of interviewing CEOs.
Um, I've written three books, uh, startup, CEO startup, CXO, and then co-authored with Brad Feld and another vc, um, Hendra Ani, uh, a book called Startup Boards. Um, so, uh, I re I remember the book In the tech scene in Forever, so Matt, yeah. Forever.
I don't know if you remember, but when you said it, the light bulb went off. We were in the Mobius incubator with you. At what, what company were you at at Time?
I was one of the co-founders and still Secure with Raj. Oh, with Ra. I just talked to Raj the other day.
Um, did, Yeah. John Claude. Yeah.
So Raj, absolutely. So I don't want to take up our interview on this, we'll talk after. No, but that is funny 'cause you and I met 25, 23 years ago.
I had Hair. Uh, but anyway, we'll, we'll talk after this, but there's a lot of commonality in here. But man, and, and for people who don't know, return Path was a power.
They, they look spam is a problem, it's still a problem. We all get too much spam in our email boxes, but Return Path probably did more to, to cut down on the spam in your email box. And I don't know if many people out there really realize it than just about any company or statute or regulation, statute or regulation out there.
I mean, it, it, it kind of set the market for how do you do responsible email. So many thanks to you for that, Matt. Um, but we're here to talk about Acro link.
Yes. Let, let's focus in, give us, give us the, the genesis here. What's the story?
Yeah, so Acro links, again, I've only been here a couple of months. Uh, so, uh, some of my language here may not be perfect or polished, but, uh, acro Links is a gem of a company. The company has, um, has been around for a long time.
It's over 20 years old. And, um, uh, the product was built as AI before large language models. Uh, and, uh, you know, we have about a dozen PhDs in computational linguistics at the business, um, and, uh, you know, have done some really interesting work over the years.
Um, we are in a space that I would call content compliance. And content compliance itself is going through a revolution right now, um, because the world of content is going through a revolution. Uh, and, um, I was, uh, brought in in January to, uh, kind of, uh, shepherd the company into, uh, uh, into that revolution.
So if you, if you think about the world of content creation in, in, um, large corporations, and that's typically who our clients are, sort of mid-size and large corporations, um, before large language models, you had armies of people producing, you know, pages and pages and pages of content. Um, and, uh, you know, some of our clients have over 10,000 content creators and over 10 million pages of online content. And sitting in between those two are the need to check the content, uh, for, um, a couple of things historically.
One is, um, uh, is the content on brand. Uh, another would be is the content compliant with the laws and regulations? And then there are a bunch of other things.
Um, so Acrolinx is a software company that does that. We sit in between content creators and publish content. Uh, we integrate with, um, large companies, content systems, uh, and, uh, we make sure that their content is on brand.
That is principally our job today, although we do edge a little bit into, um, legal and regulatory as well. But what's happened in the last couple of years with large language models is that the problem has gotten bigger and scarier for large enterprises. So now for our big clients, it's not just that they have 10,000 people writing content, it's that all of those people are using large language models to help them.
So the amount of content is proliferating. Yeah. And the human judgment element on content as a quotient of all the content is decreasing.
And companies can say all they want, they can policy all they want with their employees, like, don't use large language models or only use this, or only use that, but they know that people are going off to the side and doing stuff on chat GPT. And that's fine. They should, large language models are incredibly powerful and they're gonna change the way all of us do everything.
Um, but what that means is for, uh, midsize and large companies, um, the need to make sure that their content is checked and checked in automated ways has gone up by 10 x or a hundred x in the last couple of years. So Acrolinx, um, is feverishly expanding our set of capabilities to make sure that we're staying at pace with that, and that we're helping our clients make sure that their content is on brand compliant with laws and regulations. And over time, we'll be checking it for other things too.
You know, do they have some weird thing that showed up from an LLM that's a hallucination that might pose some ethical or reputational risks to the company? Um, do or IP or, uh, and that was my next one. I was gonna say, do, are they using someone else's copyrighted material and they don't realize it because it came out of a model, um, uh, et cetera.
So, uh, that's the way to think about us. Instant content compliance, uh, helping companies make sure that the content they produce is on brand and safe. You know, man, I'm listening to you and shaking my head.
So you are talking about one large enterprise that has, you know, a lot of content, put yourself in my shoes. We operate, I don't know, a dozen different websites plus the video. com or, or what have you.
And, um, this has become, this has become a menace for us, to tell you the truth because on one hand, yeah, you have, you're drowning in content, right? Because AI could pump out content really, really fast. And you have a lot of people you know who are pitching that content, please publish this, publish this, publish it, and we're a content house.
We wanna publish content. But you, what you've highlighted is really, really the issue. I'll go one better.
I just got done, uh, taping text on gang, and one of the segments was this same issue on video. Sure. 41% of the video being produced now has AI involved in it, either in pre-production or post-production, Right?
It's writing the scripts, it's writing the storyboards, it's, it's actually drawing the backgrounds and the captions. And then in post it's taking, you know, you and I are gonna talk for 20 minutes, but on LinkedIn, I'm gonna run a 32nd, I'm gonna do a YouTube short for a minute and, and et cetera. AI is cutting that up and doing it.
Did they do it right? Is it, is it on brand? Is it capturing what you and I are talking about?
This is, this is a problem for our types, right? Uh, yes it is. No, and my point is it's been a problem forever, but it's been a smaller problem, right?
Uh, and the problem just got a hundred times worse, Thousandths times worse because I, so I, you know, I was kidding around. I said, I left the venture back, came in about 10 years ago. I started, it was Brad actually who, Brad and Raj who helped me start this.
But, um, you know, I did, I never, I never fat in the deepest recesses of my mind that this would be the issue that I gotta deal with today, which is I've got too much content, but I gotta make sure it's just, because the other thing is quite frankly, and, and again we discussed it on the gang this morning, is, it does, is AI creating the compelling content that humans can create, right? I mean, probably look, 50% of content, Matt is mediocre at best, right? Whether an AI wrote it or a person wrote it, it, it's not grabbing you.
But can AI scale the heights to really create compelling content that is on brand that is going to get people to click or register or do whatever you want 'em to do with that content? Yeah. And the answer to that is without question.
Um, and part of that is that the large language models keep getting better and better and better. And part of that is that the people using them start understanding how to use them more Better propped. Uh, and you know, it's a little, it's, it's not quite garbage in garbage out.
'cause the models are actually pretty good at refining garbage for you. Uh mm-hmm. But, uh, the better people get at, um, at using them or at at writing automation to use them, um, the, the better the content is gonna be.
But that still doesn't mean that it's perfect, and it still doesn't mean that it's compliance or on brand at all times. And, you know, I was talking to a client the other day that's a, a large global financial services company. She's got 20 lawyers That Are constantly looking at all the content that's going through, and she said she needs to hire more lawyers because there's more content.
And what are the lawyers doing? They're reading stuff, you know, manually, they're referring back to some spreadsheets. She kept, she kept referring to the poor soul on her team that has to manage the spreadsheet.
And, you know, when you're talking about things like that, those things don't scale. Um, but you can put software and systems in place that do scale. Yeah.
What makes you think those lawyers aren't using AI to monitor that content? They certainly, they certainly could be. Um, and, uh, you know, at the end of the day, Acrolinx solution is, um, you know, what I would say is it's very horizontal, but it's chopped into verticals because, uh, verticals have different needs and different laws and regulations.
Um, but, uh, you know, it, it is all about, um, making the job easier. And that's really what AI does in general, right? It, it's not that it takes over everything, it's just that it makes the job a lot easier for people to do Ab Absolutely.
A absolutely. Lemme ask you another question. So I, again, this is, this is germane to me personally in my business.
One of the issues we have is, okay, so you can interject yourself, here's my content creator, here's my website, or whatever my distribution, you interject yourself in the middle here and you say, okay, this, this looks kosher. We're gonna put it up on the site, we're gonna distribute it, you know, do our thing. It's an ongoing process, right?
We used to call something called S-E-O-S-E-O is being blown away with ai. And, and, and especially, I mean, a, again, as a content producer, 85% of our traffic used to be from organic search. It's not anymore.
Google keeps that traffic to themselves. They put their AI in there and they're not well organic, You know, and people are using things like perplexity and chat GPT to do rag search now. Uh, yeah.
But at least Perplexity Perplexity gives you the, the the They do. Yeah. The, the actual, you know, link, they Give you the source material.
That's right. Yes. Um, so to me, this is an ongoing thing.
It's not a one and done, right? No, it's not. It's not, nothing is one and done anymore, and it's not one and done for, for, uh, you know, reasons that you would expect, which is the world is constantly moving around you, so you know what you're talking about, which is sort of on the, on the, you know, the backend, you have content that's out there, is it still meeting the mark in terms of, uh, you know, search or just effectiveness in general?
That's gonna change based on what the world does. But the other thing that happens is the, the compliance checking in the middle, the rules change all the time. Yeah.
New laws and regulations pop up. You have 10 million pages of published content. Some of them are gonna be out of sync really quickly.
Um, you have, if you're a large company, you might have a 500 page, uh, brand book or style guide, and you've got people in marketing going in there all the time and editing it. They're Sunset product names. They're deciding, they want language to be more inclusive now.
They want language to be at an X grade reading level instead of a y grade reading level. And again, tens of millions of pages of content that are already out there. But one of the things our solution does is it's constantly scanning and scoring content against a changing rule set, uh, and just flagging things as they need to be, uh, uh, fixed.
It almost screams that you have to use the AI to work with the ai. You know, as I always say to the ai, as I always say, it takes a diamond to cut another diamond. The diamond that exactly it.
That's exactly it, my friend. That's great. This is a job.
It's a job. It's a big job to be done. Excellent.
Um, you know, we didn't mention, we mentioned the name of the company, but what's the website? com. Okay.
And look just about every company, I know you don't have to have tens of millions of pages to have this problem. You could have a relatively modest website and still have this problem For sure. Especially if you are a, you know, you're selling, whether it's B2B or B2C, you, you, things change.
Matt, what's the OnRamp here? Is there some sort of free trial? Is it, is it hard to do?
I mean, is it just a question of I'm gonna install a couple agents and let 'em run? Um, yeah, That's a great, that's a great question. Um, historically, uh, you know, in the world before large language models, it was a long on-ramp.
And one of the things we're doing right now is turning the whole solution into an API solution. Um, so, uh, things like, um, you know, free trials, uh, freemium, sort of easy, easy as like a Twilio or a Stripe to integrate into your, into your workflow, um, are, uh, are coming soon over the next couple of quarters. So any company that wants to work with us, um, of any size, uh, will either use our current solution or there'll be a development partner for the new solution right away.
And, um, they're all using the same, uh, you know, sort of, uh, brilliant engine behind them. Uh, but, uh, you know, we're, we're in the process of modernizing access to it Would've been a new old solution for a new old problem. Exactly.
And again, a problem that's has been around for a long time, but it just got a lot. But it's mushroom clouded now. It's, it's, I mean, we live it here, so it, this, to me, this hits really close to home and, uh, it's God's work, Matt.
Keep doing it. I know you'll do a great job with it. Matt.
We, we, we desperately do need this kind of thing though. That's great. It's important, especially for people in the content world, Matt.
Well, thank you, Alan. I appreciate it. I appreciate you having me on.
com And come back and keep us posted Promise. Okay? Absolutely.
Anytime you'll have me. All right. Matt Blumberg, CEO Acrolinx, here on Tech Drunk tv.
We're gonna take a break. We'll be back.