3390: How Veritone Is Making AI Work for Media, Government, and Beyond

[00:00:04] What happens when AI stops being just another buzzword and starts transforming some of the most complex, high-stakes industries on the planet right now? Well, my guest today isn't here to speculate. He's here to share how it's already happening.

[00:00:20] His name is Ryan Steelberg. He is the CEO of Veritone, and he spent years steering AI beyond theory into practical revenue-driving solutions for everything from media to law enforcement, defence, and even government. And from processing over 58 million hours of audio and video in a single year to breaking through the procurement barriers of federal contracts,

[00:00:49] Veritone has gone from infrastructure builder to a creator of targeted AI applications, applications that solve real operational problems. And we'll also unpack today why unstructured data, yet those endless hours of audio-video, is becoming the new gold rush for enterprises. And on top of that, how Veritone is positioning itself in the $17 billion training data market,

[00:01:19] and why Ryan believes that AI's growth trajectory mirrors the industrial revolution. So whether you're an investor or a tech leader or just curious about how AI will change your sector, hopefully this conversation today will connect a lot of dots. Now, I also want to highlight here that at the end of every episode, I ask listeners to contact me. And you do contact me in your droves, and I try and fit everybody in.

[00:01:46] But I currently have 200 interviews booked in until November. But I do have a small number of fast-track paid options, which allow people to get on the podcast a little earlier, and help pay for my hosting fees of releasing 400 interviews a year. So the guest today did pay us a fee solely to expedite the release of this episode. But rest assured, all opinions expressed are those of the guest.

[00:02:13] Tech Talks Daily will always retain full editorial control over the content. No endorsements are implied, but we have been able to get Ryan on the show a little earlier. But enough from me. Let's get him on the podcast now. So a massive warm welcome to the show, Ryan. Can you tell everyone listening a little about who you are and what you do? Well, my name is Ryan Steelberg. I'm chairman and CEO of Veritone. We are a public AI company.

[00:02:41] We went public a while ago, back in 2017. I started this business in 2014, really to harness and leverage AI to help tackle this big unstructured data problem that we've now been dealing with well over a decade. So it's great to be here, Neil. And hopefully I can educate you and your audience about what we do and what makes us different.

[00:03:07] I'm so glad you're joining me on the podcast today, because I think over the last three years, enterprises have been on somewhat of a journey. Everyone went crazy over LLM's generative AI. This year is all about agentic AI and AI agents. And then we've had the ROI problem, AI project stuck in pilot phase, for example. I think enterprises are waking up to the fact now the importance of data.

[00:03:34] And one of the things that amuses me somewhat is you've been doing this before it was cool. I mean, you were building enterprise AI solutions back in 2014, long before the current wave of hype that we're seeing. So I'm curious from your viewpoint here, what's changed in the landscape over the last decade? And how has your approach evolved maybe as well? Well, I think we're thankful that we were nimble. So, you know, my background, I've had great success, primarily in the ad tech industry.

[00:04:04] I started some of the, you know, the larger ad tech companies going back to the mid nineties. And, you know, why that is relevant to Veriton AI is, you know, ad tech is an AI problem. You know, we, you know, you know, I sold my last business to Google. Obviously you can sort of credit them for a lot of the latest developments for the transformer based models, right? With deep mind and other things. But the point here is, you know, it's a data problem.

[00:04:30] And when we started this business, you know, way back in 2014, candidly, I just said, hey, build it and they will come. And we were a little wrong. It was a little too early, right? Meaning most companies were still trying to figure out what their data assets were, right? There really is no AI, right? In the software ecosystem without data. And, and so at that time we were spending frankly so much time with enterprises, helping them prepare their data sets even before we could apply AI to it.

[00:04:59] So I think for us, it's been a long journey. Thankfully, just with our background and previous successes, we had staying power to sort of get through this migration. Obviously now we don't have to go in and try to sell AI per se as a concept, but now you touched on a great point is how is this AI truly impactful to my business, right? Is it really driving KPIs, right? Can we really evaluate ROI quickly?

[00:05:26] So my point here is, frankly, I didn't want to build specific applications or end solutions when we first started Veritone. I wanted to build AI where, which is our platform, expose our APIs and let everybody, all the businesses come to us. But because of kind of the slow start of the adoption of AI when we started Veritone, it forced us, Neil, thankfully, to expand.

[00:05:53] Expand our model a little bit from a pure tech infrastructure platform, managing all these different AI models, etc. To building a host of very specific AI applications that the end users and end companies are using. Think of it analogous to Microsoft Office and Windows, right? We, in effect, built like Windows and an AI operating system, which we call AI where.

[00:06:18] But the industry wasn't really ready to buy and adopt the infrastructure of AI where, but they were ready to start buying the applications. And so we built like Microsoft Office, if you will, a suite of applications for our big meeting entertainment customers, a suite of applications for our law enforcement customers. And that's what they're using.

[00:06:38] So, Neil, ultimately is AI is the engine and tool, but ultimately it's the end application workflows that's turning this horsepower into value. Are we saving these companies money? Are we increasing their revenue base, right? Are we bringing operational efficiencies? So I think that's what that's what we're kind of dealing with now. There's there's a gap between, you know, state of the art, large, big language models to distilling that down into enterprise organizations. Is it really moving the needle? Right.

[00:07:08] Am I really seeing tangible results in an ROI quickly that I can substantiate? So, Veritone has gone through a journey, but we're going to talk today about, you know, a bunch of different specific applications of why this is not novelty, why these are not POCs. And there's a reason why we have 3000 customers and over a 90% retention rate. So happy to be here.

[00:07:29] And I mentioned a few moments ago the problem with data because of the importance of data in AI and yet many enterprises still struggling with data silos and unstructured data. And the reason I bring this up, I was reading before you came on today that you've described unstructured data, audio, video and text as ultimately the new frontier for enterprise AI. So why is this such an important battleground? And what is it that makes it so challenging to operationalize?

[00:07:58] Because that's what we seem to be seeing more and more of at the moment. Well, you can make the analogy what we're dealing with today is kind of how Google first came on the on this on this in this into the space way back when when they were indexing the open web for search.

[00:08:10] Right. The first thing the first thing these model companies did, you know, and there's a there's a parallel to what open AI and Claude and other groups did is the first thing they did to acquire training data is to use scrapers and other technologies to go out there and suck down any available content, mostly text from the open web. Right. But instead of building it into kind of a search index, right, then then putting like the Google search engine on top of it, they put obviously they have to transfer.

[00:08:39] They put it embedded within the transformer model and now they have a natural language interface that you and I all know is chat GPT or some other derivative. But at the end of the day, it starts with the knowledge graph. Can you get enough different data points which you touched on different data silos to create that knowledge graph? The first version knowledge graph for these large language models was primarily text based as they got more sophisticated as the GPU and CPU horsepower continue to increase.

[00:09:08] We and enterprises, we don't just want a corpus knowledge graph of text. There's other forms of data out there that we want to include in this in this knowledge graph. Audio and video is a huge category. The fastest growing segment of unstructured data is audio and video. This this podcast is a form of unstructured data. Everything that you watch on OTT and television and YouTube, that's unstructured, messy data.

[00:09:35] When I'm when I say unstructured unstructured, it's not like it comes with a data schema that specifically defines right what's in this frame of video. You have to somebody has to act upon it. You can either use manual data labeling companies like scale AI or you can use ironically AI like from Veritone to create the metadata to turn it into structure.

[00:09:55] So the so what we've gone here from from these the model training is primarily text based training to image based training and now full blown audio and video at scale. And that's and that's kind of the state of the art of where we're seeing a lot of I mean tons of investment from the from the big hyperscalers and the large AI model builders is huge, huge hundreds and hundreds of millions of dollars of investment into training data. And we're here and Veritone's here to help support that transformation.

[00:10:25] Incredibly cool. And I was also reading before you came on, I think it was in in 2024 alone. Veritone has processed over 58 million hours of media using hundreds of AI models, which is just mind blowing. But what I'm curious, what kind of insights are your customers extracting at that scale and how is it changing their day to day decisions? Because the numbers and the stats are incredibly impressive. But for the business leaders, what does that mean?

[00:10:53] So let's let's use two examples. You know, every company, it's their data. Right. And likewise, it's not it's not homogenous in terms of the use case. Everybody can take advantage of this knowledge differently. But let's take ESPN. We all know ESPN, right? The big sports network and portfolio under Disney. They've been a client for us of ours for years. We help them ingest all of their different data sources. What you see on SportsCenter on TV, what goes over the air on AM and FM radio.

[00:11:21] Right. Or all their podcast. We help ingest it. We use our AI to index it. And what do they use it for is really, I'll say for them, like four primary pillars. Number one, advertising optimization. You know, Neil, you and I are doing this podcast. And for example, organically, let's say I do a LifeLock ad right in context in this podcast. Right. You need either humans to label that, that I even talked about LifeLock and how long I talked about it and make a correlation to the audience size.

[00:11:51] Or you can use AI to index it and saying, Ryan Stielberg said LifeLock and it was in this context for this long. OK, so ESPN and a lot of our media and entertainment customers who are who have an advertising model use us to help increase the research or intelligence on their ad objects. OK, how valuable is their media from an advertising perspective? The number two one, which seems kind of obvious, is just search and discovery.

[00:12:18] If you're ESPN, you have literally millions of hours of content that you're now trying to ingest to ultimately create SportsCenter. Right. And now they have to push out content fast. So their business is ingest, index, organize and push out to you and I, the consumers. The speed by which they can do that, the level of personalization is critical.

[00:12:39] So now Veritone, through our technology and indexing, can help them search through content and compile, right, personalized relevant content packages fast. Right. So I'll say there's an advertising benefit, there's a programming benefit and there's a research benefit. So this is one that you as a host potentially would not like, but they can analyze the ratings.

[00:13:02] Right. So if they so if you're a big organization like ESPN and you have made huge investments into the talent, your hosts, all right, of all these things, they need to perform. And one proxy by how you evaluate performance is ratings. And so what we've done is because now I can monitor exactly when Ryan Steelberg's face is on screen and how long, pretend I'm a host, I can start to make correlations.

[00:13:29] Right. Or assumptions that we have an issue with Ryan. And he's losing his appeal with a demographic. Right. Because I can now correlate time on screen for Ryan and ratings. So there's a research component. Right. And then ultimately production. Right. How quickly can I actually compile and build programming? So that's just ESPN. All that starts, everything I just described starts with having a good foundation.

[00:13:54] Right. Of adjusted content corpus and great indexing using AI and the AI workflows. So everybody's a little different. CNBC is a big client. They, you know, they they they have the ingestion that AI, but their use cases may be a little different. Right. And some of them, Neil, want to keep them proprietary. They don't want everybody to know how they're using the their their respective AI from baritone to differentiate their business. So I would say that kind of fits in two categories.

[00:14:21] One is based upon the use cases that I just described. It provides huge operational efficiency, obviously, but also increased revenue opportunity. Right. We can't afford not to have a productive host. Right. We can't afford not to take care of our advertising clients, which is obviously a proxy to revenue. And what I was also reading up on you guys, I was also learning you're gaining traction in some traditionally hard to reach industries.

[00:14:48] And by that, I mean, federal law enforcement, media defense. And I could go on and on. I'm curious, what's driving adoption in these sectors? And what are the biggest barriers that you've had to overcome? Because as you said, each use case is different. Each industry is different, which must represent a challenge on its own. But what are the biggest barriers that you've had to overcome? Well, for us, the learning curve was steep, not from a technology perspective, but I'm a commercial guy.

[00:15:15] My entire career has been selling directly to commercial enterprises. It's taken us five years plus to prepare baritone through, again, some changes in our technology stack to make sure we can put it into secure environments. Right. And we have the right security clearances, etc. So I'll say there's some obvious security and compliance things that we have to do. But then it's understanding the contract vehicles. How do you get on a contract? Right. Or an award for a big government contract from the Air Force.

[00:15:44] So ironically, the easiest part of the entire equation, Neil, was our software. The hardest part was just being prepared to sell into the space. So thankfully, we've sort of forced us and we got up the learning curve. And so now we've broken into state and local law enforcement. We have several hundred here in the United States. We have several hundred police and sheriff departments that use our software. They're not using it for tracking advertising. Obviously, they're looking at for tracking bad guys on video footage, for example.

[00:16:12] And so the use cases are different. But we now are successfully a mission critical vendor now for law enforcement sheriff departments across the United States. And a lot of those same use cases now are being adopted through recent awards with the DOD and the Department of Justice. The big one that we just announced was with the Air Force and their entire law enforcement policing body around the world. It's actually big.

[00:16:41] All of our big branches in the U.S. of agencies have large, if you will, law enforcement agencies. So we're helping modernize a lot of these very old legacy systems. So we're excited about it. And again, just showing you the scale of what started as helping a sports programmer like ESPN has now expanded to now servicing multiple branches of our federal government and state and local law enforcement.

[00:17:08] And you mentioned there how you've recently secured a sole source contract with the U.S. Air Force of Special Investigations, adding to your existing work with the Defense Logistics Agency. And I've got to ask, what is it or what does it take to scale AI inside government environments that are known for legacy technology being slow to adapt to change? They're risk adverse and highly regulated. Where do you begin on something like that?

[00:17:36] And how have you found working with these government agencies as well? Is there a real desire for change? Well, this is not an 80-part podcast series, so I'll try to summarize it a little bit. But first, from a technology stack perspective, we needed to do some work to both the AIware full stack and the applications to make sure that we could deploy those applications into a secure environment.

[00:18:03] In some areas, Neil, network isolated, meaning there's no comms going in and out. So meaning everything, all this AI, everything we have to do has to literally run in a private cloud or literally even on-prem. Okay? So that took time. So technology took time. The second is, in the past, you know, we did try to go and build relationships through partners,

[00:18:26] which at times, once a technology and an industry has matured, often going through, and what I mean by that is, at times, we may not be the prime contractor. We don't have the direct contract with such agency. We tried that approach in the past with not as good of results. Okay? It's like introducing a new cutting-edge technology like ours, but I have to go through a middleman before I can actually deliver the solution to the Air Force.

[00:18:53] So I think what we tried really hard was to establish Veritone as a prime contractor. And that's what these contracts with the DLA and with the Air Force are. It's our paper directly with them, meaning Veritone's people are interfacing directly with our constituent partners at the Air Force. That has greatly accelerated our ability to get this software deployed, right, and get them trained so they can immediately start seeing value out of it.

[00:19:18] So I think, you know, just, again, bureaucracy is not just the time it takes to get a contract done historically, but just the classic design of it where there's so many tiers before, like if I have a piece of technology, by the time it's used by the end user, it may go through three or four hands, and that creates, and that created a lot of inefficiencies. When you're talking about cutting-edge software and AI, the closer you can get to the mission, the closer you can get to the end user, the better and the better performance that we're seeing.

[00:19:47] So I think it's a combination of us getting prepared technically, right, for having the right securities and deployment, and then it's having the right business go-to-market strategy where we can interface and sell directly with the end mission controllers, in this instance, the DLA and the Air Force. And I'm so glad you've repeatedly said the word value today because we have seen a lot of hype over the last three years. It feels like we're coming out of that hype cycle now.

[00:20:14] So how are you steering a public AI company to be both innovative and financially resilient? Because there is so much, as I said, we're moving away from the hype. There's a lot of exciting things happening here. So where do you begin to steer this? Is there anything you can share around your approach? You know, I think it's just classic enterprise and or SaaS business models. I mean, at the end of the day, you know, the hype, you know, first of all, the AI space is, you know, we may be in a micro bubble. Let's be very clear.

[00:20:43] The AI space is going to be huge. It's going to, I think it's going to be, you know, an order or two bigger than cloud. And it's not, we're not putting the genie back in the bottle, right? But again, it is, there is a lot of noise out there. So for us, ultimately is first and foremost, are we landing customers? Are we keeping customers, right? Are they re-signing their deals? And I like to say is it goes from a novelty to a POC, right? When does it become mission critical, right?

[00:21:10] When everybody's under budget constraints and they're trying to say, hmm, are we going to renew Salesforce.com next year? And how many seat licenses do we need, right? Are we going to renew NetSuite next year, right? This question, every CEO, every CIO, right, has in CTO has to go through. We're no different. Is Veritone software, right, good enough and valuable enough for that customer, for that enterprise or that sheriff department, right? To want to renew that contract or expand that contract.

[00:21:39] So again, let's, you know, obviously what we think is, you know, what we're building is state-of-the-art and we love at times, you know, the attention and the hype that we get because it's new and novel and it's AI. But from a business perspective, let's not, let's be clear here. There's a reason why a Microsoft is doing so well, right? They had the applications already, right? For example, they didn't make a new Microsoft Word. They made Word better.

[00:22:07] If you remember the old commercial, what is it? As a kid, BASF. We don't make all the things that you use. We make all the things you use better. I think that was the slogan years ago, right? Yeah. So I think that's for a lot of the things that you're seeing successful is that AI can be applied as a force multiplier to previous existing great applications or workflows. So, for example, in our business, if we're law enforcement, redaction, right?

[00:22:34] Having a human sit behind a desk and using old software to go frame by frame to blur your face, Neil, for example, just a trivial takes hours and hours and hours. Right. But if I can automate that with AI, right, with great precision and accuracy, right? Boom. It's I haven't had to come up. You don't have to come up with something radically new and novel. AI can be a force multiplier, right? By. But again, what's really sold in the first place is this is the solution of redaction. Right.

[00:23:04] Can you make redaction better and more cost effective? So I think that there is a lot of it. So that's why I think from if you always, always say is where's where's the money going? Right. The big dollars are going to the application layer right now. Right. Sequoia Capital, who, you know, I've known and respect for years in Andreessen, you know, they put out a lot of recent reports and studies of it. Yes, the big bets have been made in infrastructure. Right. We all know who the leaders of the chips are. We know about the data center.

[00:23:29] And we obviously all of us know about the few huge public multibillion dollar large language models. Right. But what you're going to see now is it is the much bigger ecosystem in aggregate is the workflow application layer. Right. And it's always been that case. So, again, make sure your your end solution don't stop trying to sell it as the magic of the AI inside. What is ultimately doing and how is it serving the end customer?

[00:23:55] Now, the Veritone Data Refinery or VDR is positioning itself to tap into the global training data market that is projected at 17 billion dollars, which is a huge number. So how does VDR fit into your long term strategy now? And what kind of demand signals are you seeing from both public and private clients to prove that that demand and that adoption is there? You know, we've been around for a while.

[00:24:22] And because we've been working with such large media and entertainment and other data centric companies, you know, we've been ingesting and indexing and cleaning up, as you've touched on earlier, tens of millions of hours of audio and video. And and now what what we were getting paid for to help them organize that huge corpus of data is now gold as a new training supply for these next generation AI models.

[00:24:51] So now we are working with the largest hyperscalers, the largest AI model developers, many of the names we've mentioned previously on this podcast and helping take that tonnage, that messy unstructured data. It's gone through Veritone AI where we've cleaned it. We've indexed it. In effect, we prepared it to turn into training data that now when we're seeing these next generation gen AI models that are just blowing our minds about how well they're creating imageries or movie scenes.

[00:25:21] Veritone is part of that ecosystem as a I'll call premier partner for training data. So it's a purely great organic new line of business. It's probably the fastest line of business in terms of growth. We just came to market in the fourth quarter of last year with this. And we are and we already have kind of an immediately addressable like customer pipeline of like 20 million dollars for this new line of business. So for our size of a business where, you know, again, we're 100 million dollar plus revenue business. That's big for us. That's big for us.

[00:25:50] And we see that this new VDR line continuing to accelerate, selling across both commercial and public sector, which is really exciting. And listening to you today and your story, you've obviously worked at the intersection of AI, media and marketing for years now. So I've got to ask, what's your perspective on where enterprise AI might be heading?

[00:26:15] And as someone that loves to bust myths and misconceptions on this podcast every day, what's still being misunderstood by investors, customers or all the media? There's a lot of scary stories on our news feeds that might be swaying people. So what are you seeing here? I think we take a step back. It's the growth has been spectacular. If you've kind of benchmark this phenomena against anything we've seen in the past, if you want to get romantic and compare it against the industrial revolution in the past.

[00:26:43] Or you want to look at, you know, how is this new explosion of opportunity compared to the mobile explosion with the iPhone versus the World Wide Web onset? So, you know, they're just like it back in there. Back in the day, there was a lot of hype. There was a lot of true meat and there was a lot of vaporware in every one of these things. OK, this one is no different. It's just bigger.

[00:27:06] And the other thing that thinks a little bit different this time is a lot of these technologies were given to us, introduced to us that we can all use immediately. Right. Unlike some of these legacy technologies. Right. We didn't have to wait 15 years before OpenAI decided to let us use ChatGBT. Right. To be clear, Google. Right. Who came to market later with Gemini. They had this technology for many years.

[00:27:32] They could have introduced and released Gemini, as you know, years ago, well before ChatGBT. You know, obviously their team, in my mind, right, were the foundation of the Transformer model team. OK. Now, there's some debate about that. But having been at Google, I was kind of around, right, when a lot of this stuff was was in the works and even precursor. So the point is. There's hype. There's like any industry. Investors just need to find out, you know, is this a real AI company? Do they have staying powered?

[00:28:00] And do they, at the end of the day, have a business model? Right. And at times, it's exciting for us with the hype. But it's also frustrating for us is to say, how is Veritone different from other company, you know, ABC? I'm using them as just an example. But again, it goes back to me as the fundamentals. Do they have history? Do they have education in data and AI? And how long have they had that experience? You know, B, do they have customers? And are the customers resigning? Are they coming back? Right.

[00:28:30] And third is, do they have any consistency? Are they just small business? You know, are they or is there anything consistent in what they're doing? And obviously, you know, a little, you know, again, to our horn is, you know, we have thousands of customers. They're the biggest names in the categories, whether it's the Air Force right in the Fed space. Or it's, you know, ESPN and Amblin and Entertainment and NBC Universal in the media entertainment space. So I think investors, they're not really missing things. I think it's just the space is big.

[00:28:57] It warrants taking risk and making bets on companies. But you've got to do your extra level of homework because, you know, there's just it's, you know, it's such a large pool that, you know, it's the gold rush. It's the, you know, everybody and their mother wants to get into it because that's how big the ecosystem is. So just be wary and do your diligence. Fantastic advice. I think that's a powerful moment to end on. But before I let you go, it was a big thank you for sparing your time with us to share your insights today.

[00:29:26] I'm going to see if there's something we can do for you now. Some of the biggest names in business, VC funding and tech could be the Bean guests or maybe even listen to this podcast or through six degrees of separation. Maybe we can connect you with someone. So who is a person you'd love to have a private breakfast or lunch with? Who would it be and why? And hopefully let's see what we can manifest together and make something happen. But who would it be? Oh, man. On the spot. Great question.

[00:29:53] Well, you know, I think I love talking to people who have have experience but are dealing with yet another major disruption. And so I would you know, I would love to sit down with, you know, a Steven Spielberg or, you know, somebody who, you know, again, is still in the game. They've had an experience. They went through the CGI revolution. Right. Which everybody, let's be clear, thought it was going to destroy Hollywood, et cetera, et cetera.

[00:30:22] Some people think it did. But to be very clear, the industry grew, you know, a hundredfold. Right. But that would be exciting. You know, again, I don't need to, you know, I can read enough and thankfully they're transparent enough. I get enough, I think, communication from an Elon Musk. No, I would love to have lunch with Elon Musk. But I think it's just, you know, a Brock, I mean, a Brockheimer, you know, some of the most classic, you know, forward thinking directors and producers.

[00:30:48] Because they're, I mean, this is another period of time where there's so much upheaval and unknown about that industry that is mission critical to us. We can't imagine a world without quality media and entertainment. And it's probably the area that's probably most acutely being impacted, right, by generative AI. So that would be a great dinner and conversation. And our last names kind of sound the same. So maybe I can sneak in using my last name for dinner and they'll think I'm saying Spielberg, not Steelberg. What a fantastic answer.

[00:31:18] Absolutely love that. We'll put it out into the universe. Let's see what we can make happen there. And for people wanting to find out more information about all things Veritone, see some of the big stats, figures and the value that you guys are offering. Where would you like to point them? Veritone.com, you know, V-E-R-I-T-O-N-E.com and our investors. So you can either get great data on what we do and who we service on our website. And obviously our investor website has great financial reports and updates for everybody.

[00:31:46] Well, I will add links to absolutely everything there. Make it easy for people to find you. But more than anything, just thank you for bringing this topic to life today. You've had an amazing backstory and I love how it's all come together. And you were doing all this before. Many people listening were even thinking about AI. And best of luck moving forward. I'll be keeping in touch with you, see how things are progressing. But thanks for starting this conversation today. Great. Thank you, Neil. Have a great day. So there you have it.

[00:32:14] That was Ryan Steelberg, CEO of Veritone, sharing an insider's view on where enterprise AI is heading and how unstructured data is shaping the next wave of innovation. So from building technology that can run securely inside federal agencies to opening new revenue streams in training data.

[00:32:36] Ryan's perspective, I think, reminds us that the future of AI will be driven not by hype, but by companies that can scale real world results and deliver that value. So if today's conversation got you thinking about how your organization will manage and extract value from your own unstructured data, please, if you want to dig in a little bit deep into Veritone's work at veritone.com. And of course, I want to hear from you too.

[00:33:04] How do you think AI in law enforcement, government and media and everything in between? Are you happy that it's advancing this quickly or do you think it's time we slowed things down? Please share your thoughts with me on LinkedIn X or wherever you have your big picture tech debates. I'm just at Neil C. Hughes on everything. You can also get me on techblogwriter.co.uk and the new site, techtalksnetwork.com, which houses all eight of the podcasts that I run.

[00:33:32] So keep your thoughts coming in and questions. And I'll be back again tomorrow with another guest. I'll speak with you all then. Bye for now.