Who will police the AI agents?

AI agents can act faster than humans can review them. So how do companies keep control without slowing everything down?

In this episode of The Deep View Conversations, Jason Hiner sits down with Blake Brannon, chief innovation officer and a co-founder of OneTrust, to explore why AI governance needs to evolve from static rules to decisions made in real time.

Brannon explains why an agent can inherit an employee's identity and access without inheriting their trust. He makes the case for independent guardian agents that evaluate what other agents are trying to do, apply a company's policies, and redirect or block actions when necessary.

One example captures the challenge: deleting CRM records could be routine cleanup or an attempt to inflate a salesperson's performance. The action alone doesn't tell you whether it should be allowed. Its purpose matters.

Topics covered:

• How OneTrust evolved from data privacy to AI governance
• Why first-party data and customer trust go together
• Why human review can't keep pace with a growing population of agents
• How guardian agents and independent checks could help close the trust gap
• The difference between containing AI and aligning it with business policies
• Why governance should help companies move faster
• What the bring-your-own-device era teaches us about AI adoption
• How Blake uses visual storytelling and voice-to-text tools to work more effectively

If you're deploying AI agents and trying to move beyond experiments without giving up control, this conversation offers a practical way to think about the systems that need to surround them.

Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology: tdv.transitor.fm 

And don't forget to sign up for The Deep View daily newsletter. We don’t just cover AI, we decode it. In a world flooded with hype, we deliver sharp, no-nonsense insights to keep you ahead of the curve and help you put AI to work every day: subscribe.thedeepview.com

Creators and Guests

Host

Jason Hiner

Editor-in-Chief of The Deep View

What is The Deep View: Conversations?

From frontier labs and enterprise platforms to emerging startups reshaping entire industries, The Deep View: Conversations podcast interviews the brightest minds and the most influential leaders in AI.

Blake Brannon: Traditional security guardrails, governance control tools are almost like brakes on a car. It's going to prevent something bad from happening and sort of slam on the brakes. That works to contain things and is very important. But if you can imagine a bunch of AI agents in a company that are constantly being stopped, they're constantly having their brakes slammed upon, they're not going to move very fast, are they? Right? What you have to do is you have to teach them how to drive. You've got to build a full self-driving system. And this is what I call alignment, right? So you need to encapsulate both concepts when you think about AI governance because they're equally important, but for very different reasons.

Jason Hiner: In this episode, I talked to Blake Brannon, chief innovation officer at OneTrust, a company that helps organizations put guardrails around their most sensitive data and business practices to create an environment where AI agents can operate safely. Blake explains how OneTrust evolved from helping companies manage big data and machine learning a decade ago into building systems that can govern AI agents in real time. We dug into one of the biggest challenges facing enterprises today, the fact that AI agents can act faster than humans can review them. Blake argues that companies will need guardian agents that can evaluate an agent's intent, apply security and compliance policies, and either redirect or block its actions within milliseconds. He also explains why static rules are no longer enough and why effective AI governance must balance containment with alignment so companies can move faster without giving up control. Blake and I also talked about the parallel between the rise of AI and the bring-your-own-device era when companies learned they couldn't stop employees from adopting new technology and instead had to find safe ways to support it. All right, so here it is, our conversation with Blake Brannon of OneTrust. Blake, why don't you start with, you know, for those who aren't familiar, you know, what does OneTrust do and what's your role with the company?

Blake Brannon: Yeah, so Blake Brannon, one of the co-founders here at OneTrust, I'm the chief innovation officer, which means my role in the company's history has really been around pioneering the products. Where's the market going? What are the—deeply understanding of the problems to be solved and how do we translate that into technology to obviously help support our customers. For those of you who are not familiar with OneTrust, I'll give you a bit of a longer story, but it's going to be good and interesting, I promise. But the story of OneTrust is really rooted in data. And if you take yourself back to like 2012, when you had cloud computing really taking off in the enterprise for the first time in a big way, you had mobile devices kind of taking off and all these mobile apps and services and data storage specifically was becoming really, really cheap. If you can remember back then, you actually had limits on how much data you could store in Dropbox and these things were really, really sparse, your email, et cetera. And a lot of that was driving this regime in the world where everyone was talking about big data, big data, big data, big data. And it was because everyone was realizing the potential and the value of all that data. So what was happening was the world order at the time was sort of every company in the world just was on a mission to get as much data as they could from everyone. And that's in the backdrop of infinite storage in the cloud because it became cheap. Mobile phones, which had all this personal data for the first time, right, that you could collect and capture about users. And what started to happen? Well, this is right around the time when things like Edward Snowden started to happen. Right. And those of you who are not familiar with that, that was the U.S. surveillance of citizens and what they were doing. And the the only real mechanism to solve some of these problems was organizations would look at some issue they had with a contract, a customer or something they wanted to do. And you would do what I call like you talk to your lawyers and you said, let's go update our terms of service agreements so that we can permissively do these things. We can use this data for whatever we want. And that was what everyone did. You probably went through some of this yourself. All of us listening, right? We clicked through these updated terms of services. But effectively, the company was saying, whatever we get from you, we can use for whatever we want. And OneTrust was kind of born out of that backdrop. We realized that this operating principle in the world was just not going to continue to survive. Something needed to change it. And we felt like there had to be some technology that would help bridge these gaps where organizations could actually understand what they were using their data for and actually respect the permissions of people. And the first proof point of this coming to life was a regulation called GDPR. What GDPR, which came out in Europe to start, but what it fundamentally did is it flipped this world order from a company no longer just owned people's data and could do whatever they wanted with it. Data was considered your personal property. So your personal data is almost like your personal property, just like you may believe you have a right to fair rights, a right to bear arms, a right to personal property, all these things. Your personal data kind of follows that same regime. And what that meant is that companies for the very first time had to do something which sounds simple, but they had to keep track of what are we using that data for? And is it OK or not OK to be doing that? And that started with privacy, but very naturally, OneTrust extended that to—it's all about data use. So it naturally extended to other compliance regimes, other security regimes, third party risk. And now, of course, AI is one of the biggest users of data. So it was a natural extension for us to start governing people's use of data in the context of AI. Fast forward to today, we do this for over 10,000 companies around the world. Over half the Fortune 500 uses today to help them govern, you know, are they using data responsibly and how do they not just say that per a contract or a terms of service, but have operational processes, controls, and mechanisms to enforce those decisions programmatically.

Jason Hiner: Wow. Okay. That's a lot that you do. And more critical than ever because of the piece that you mentioned about AI is so data hungry. And that also means that AI is this incredible source of original data as well inside of any company, right? It's creating new data. So that gets into all kinds of things. And you all are dealing a lot with data governance, policies, as well as sort of more AI governance broadly as well. When did you start, when did the journey evolve from just sort of big data into AI governance?

Blake Brannon: Well, if you go back to before people called it AI or agents and things like this, it really was just creative data use, right? It was taking this data. And I think a lot of times the coin term that many people were using was first party data. Companies were realizing that if we can gather first party data, it becomes a unique data asset or an asset for us to kind of create differentiated value in the market. Now, at the time, it was machine learning. It was just basic use of that. But the concept and the core principle of that was really simple. But to get that first party data, you know what you have to have for someone to give it to you? You've got to have trust. You've got to have that customer, that business trust you. And trust what you're doing with their data. And are you going to be a good steward of that data? Which means you're respecting privacy, you're respecting security, you're protecting that data. And of course, adhering to compliance and ethical obligations that you have for that. So it very quickly was like a natural extension. I don't know at what point, maybe 22, 23, the world started just changing. Oh, this is no longer a data risk assessment. It was like an AI risk assessment as those technologies started to shift from traditional ML to AI. But fundamentally, it's the same thing. And obviously, AI creates new risk and new scenarios that you have to worry about in addition to the citizen development of data and data use within AI. Because now, of course, an average person in a company can use these AI tools to start doing new things with data that historically required an engineering team, you know, a project charter, a budget approval cycle because it was such a heavy lift. Now in an afternoon, you can spin up something completely new and start using data that you already have, which creates a whole new risk in itself that you have to govern and figure out how to manage.

Jason Hiner: Yeah. So two questions for you. Let's start with one. For those who, you know, most people in our audience probably understand what first party data is, but if they don't, Blake, could you just give a quick, you know, soundbite on, first, how to understand first party data and sort of why it's so valuable.

Blake Brannon: Yeah, maybe the way to relate to it is there was this big push and to direct to consumer movement that we saw kind of what happened in COVID and coming out of that. But basically the notion of first party data was if I can build a first party relationship, not through a marketplace, not through a retailer, not through a warehouse or distribution, I can directly get a relationship with that customer and they're going to give me data. I now have a unique set of information that I can personalize and tailor my goods and products to. And you sort of saw this with companies like Disney choosing to go direct and build their own streaming business and things like that. It was all tied to this notion of how do I build a relationship so I can get that information directly from the consumer so I can provide unique services and values that are not just part of the marketplace or the network that I'm getting that data from. So there's been a huge movement online in the past several years of shifting from what used to be called third party data gathering, which was like you went to data brokers to sort of buy aggregated data to just building direct relationships with your customers to get that data first party or directly from those consumers.

Jason Hiner: Yeah, it's a great description. And so it gets into this broader theme in business today, which is to own the relationship with your customer, right? Not have an intermediary between you, you know, and your customers, but to have a direct relationship in order to build a direct relationship that has to be based on trust. Like they have to trust you. They have to trust what you're going to do with the data. They have to trust that you're not going to, you know, abuse it. In other words, like spam them with information that they don't need and all of that. And so that's a very critical part of kind of where we're at in the whole, you know, online digital business cycle is developing those relationships of trust.

Blake Brannon: Yeah, absolutely.

Jason Hiner: OK, my other question for you is, you know, when companies are trying to govern AI with their sort of existing privacy and security systems and policies. Some of them are running into trouble with that. You can't necessarily govern it in the same way, especially now you have AI agents in the mix and other aspects of AI, which AI can throw off a bunch of data. It can make, as you were alluding to, the barriers to access to data a lot lower. And so that changes some of the dynamics. I imagine that's some of the stuff that you all have to deal with and are really kind of trying to help companies with.

Blake Brannon: The problems are long indeed, right? How much time do we have to talk about so many of the different problems with AI? But I'll talk about just a few of them. First off, AI, you have to think about the notion that will be real very quickly, which is within the next two years, there will be more AI agents that are doing work than there are humans doing work inside an organization, IDC says there'll be about 1.3 billion AI agents by 2028, which to put that in perspective, because it's hard to know, well, what does that mean? That's four times the number of human workers that we have in the same segment of the market that they did that analysis from. So just think about in an organization that you're in today, if you four-X the team, all of the things that would start to become hard to do, like not only would there be desk and you know, parking issues at the office, but keeping people on the same page with decisions, the context, the updates, trying to keep track of who's doing what, all of that becomes just a lot harder. So that scale problem is real. And we're kind of getting to what I call the point of we're about to surpass the human capacity ability to review what all those AI agents are going to be doing, right? It'll become impossible for humans to review every scenario, every action that an agent wants to do. So you can't scale it with humans. But AI also creates some interesting nuances as well that are sort of new dimensions you have to think about to solve the governing problem. So first off, agents typically inherit someone's identity. So if you deployed an agent to do something on your behalf on your computer, it's inheriting, Jason, your identity and your permissions and your access to connect to your business applications you have or your file system that you have. The problem is, while it inherits your identity, it doesn't inherit your trust. And what I mean by that is most people would say, what I've learned through all my conversations with clients is, I trust Jason to go connect to this business system and make this update, but I don't trust Jason's agent that has the same access to do that yet, right? Because we're just not there yet as a world. So inherently, there's sort of this trust issue with, I'm not sure if I trust that machine yet. So I want to unlock some of that potential, but I don't want to do it for everything. Very simple example. I trust that agent to read data from the CRM. I don't yet trust it to delete data from the CRM. It's a very common, realistic scenario. The second thing AI does is it's non-deterministic. So the traditional software that you use to govern as an organization or programs or processes you would govern were deterministic. You knew what you were designing for. You changed it. You often had months before it got coded and built to kind of make those governing decisions. AI is non-deterministic, which means you can't govern it up front. You don't know what it's going to conclude to do, what action it's going to take until the moment it decides to take it. So you have to shift like almost the point in time of where you're doing. You're trying to enforce security, compliance in these things because you can no longer make decisions about what you want to allow to happen or not allow to happen up front. And then the third one, I mentioned this earlier, but it does create an issue, but the citizen developer problem creates a real struggle where most organizations that I talk to can't confidently answer, how many AI agents do we have deployed in our organization right now? What are they connected to? And do we trust what they're doing? And the reason is because everybody can create their own sort of on the fly. So those in themselves create unique challenges to the governing problem. This means that the way organizations have traditionally done governance, if you look at governance in a function in a company, whether it's compliance, privacy, security, it all fundamentally is looking at something the business is trying to do. It is assessing a framework, whether it's NIST for cybersecurity, SOC2 for compliance, GDPR for privacy. It's taking a framework that has a set of principles and guidelines. And it is applying it against that business scenario. And it is making a risk-informed decision of we can do this, we can't do this. Like you could make it that simple. This is permissive to do. This is not permissive to do. This is okay to do if these things are also in place. We can use the data in that way, but we've got to make sure A and B and C are in place. And that's fundamentally what these governing things do. The challenge in AI. To get to one of the many problems you're solving here is an AI agent reasons in real time, goes and decides in real time. So that decision process that governing teams traditionally have to do that take weeks of time, you have to figure out how do you do that in seconds? Like, how are you going to do this? And this coins a concept that we talk about and help our customers with called, well, to solve this, you—you have to remove the human bottleneck. You have to have what we call guardian agents, which are the oversight agents that effectively can make that same reasoning and logic and governing process that a human team would do, but they're going to do it in milliseconds, right? They're going to be able to look at the intent, the scenario, what an AI agent wants to do, and they're going to take that backdrop of policy, security, compliance, privacy, and they're going to say, here's what's okay to happen and here's what's not okay to happen. And they're going to help that agent change its behavior or block something bad from happening.

Jason Hiner: So is the guardian agents that you mentioned, is that one of the things that OneTrust does?

Blake Brannon: That's right.

Jason Hiner: Okay. So, can—are you coming to this conclusion that AI has to be part of the mix because of pure sort of scale and velocity? AI has to be part of the mix in securing AI. Where are the kind of circular trust problems that that creates? And then where do humans need to be involved in the loop to sort of guarantee some of that trust? I'm sure that's one of the things you and your customers are wrestling with all the time.

Blake Brannon: It is. And I think the first part of that, the way to think about it, is machines have historically always been good mechanisms to govern themselves, right? So having a separate machine, being able to be an oversight to the first machine is actually a good mechanism and design to actually govern it. So I know conceptually, it's like, how would you trust itself? It's not quite. I do think there's a trust issue where you could say you don't want the AI harness that is trying to act being the self-governing, which is sort of why people and this is— a very trendy topic right now, which is OpenAI can't govern OpenAI. Anthropic can't govern Anthropic, right? You want an independent sort of assessor to kind of look at it differently to kind of do that governing.

Jason Hiner: Created by a different entity and sort of using that as a check on the one that's doing that.

Blake Brannon: Exactly, exactly. And the way we've even worked with this and thought more about it and the way it's been built is: a business agent that is trying to achieve an objective by design has a very open, broad charter. Its goal is to take an objective that it's trying to do and to figure out whatever path possible, even if it's never been conceived before, to get to that stated objective. So by design, you want it to be creative in that sense, right? You want it to find all the hidden paths and go find the optimal way to get to a solution. A governing agent doesn't have that broad open charter. A governing agent has actually a more focused scope charter. A governing agent looks at your policies, your compliance obligations, your security operating procedures, and its job is to codify and build that into a context intelligence layer that it can then assess something the business agent is wanting to do and then compare it to what would be considered permissive. Right. Is something that it could do. And its job is not to go find a way to make sure that is OK to happen. Its job is just to compare it to this more focused kind of scope within that. So it actually works really well. And you see a similar taking it outside the compliance context and the security context. You actually see the same thing being implemented successfully for how AI systems have been able to be nested architecturally to actually derive an optimal solution. So if you look at customer support agents, one of the first challenges everyone had was the model accuracy only being, let's say, 80%. By simply having independent agents assess the same problem and have consensus voting to conclude the right answer, you're able to move that up to, let's just say, 98 plus percent accuracy. So that simple architecture has been validated even outside of the governing and compliance use cases as well.

Jason Hiner: And by the voting mechanism, you're talking about multiple agents of all evaluating something and then voting to see was it in compliance or not if there are sort of gray areas?

Blake Brannon: That's right. And the way to maybe think about it is humans do this all the time. We almost get a little too close to the problem sometimes, and we can't quite see the broader picture. So what do we do? We go ask a friend or someone in another team to kind of be an outside perspective or a pressure-check perspective. And even if you did it with one person, you actually get really good calibration information about I am on track or I'm not off track. So the same technique is basically what I'm talking about here.

Jason Hiner: Hey, everybody, thanks for listening to this episode. Quick note, and then we'll get you back to the conversation. We love bringing you this content every week. We're always trying to figure out how we can deliver you the most value to help you understand how AI is transforming business and transforming the world. If you're enjoying the show, there's an easy way for you to give a little value back and help others learn about the show as well. If you're on Apple Podcasts or Spotify, drop us a rating and leave us a review. And if you're on YouTube, hit like, subscribe, or leave us a comment. That's it. It only takes a minute, but it's a huge help. So thanks in advance for pitching in. And now, back to the show. Okay, you know, what's the biggest gap between companies that are, you know, racing to deploy agents and then those that are sort of kind of failing to kind of govern them as they go live. Are you seeing lots of this? And is this something that companies are really sort of struggling with because the train is already sort of down the tracks and so now they're trying to catch up? Or is it more so that you see customers that are like, okay, now we're ready to do this. We need a partner to help us do it so that we can make sure that we're doing this in an effective way. What are you seeing more of?

Blake Brannon: Well, I think we're seeing more, maybe the latter, but let me just distinguish a little bit. I think you see people that are making really good progress with getting AI to be adopted and solve the business use case. We finally, I feel like, have gotten through over the past 12 months, let's just say, where there was a lot of that initial headwind of, it's not quite accurate enough, it got this wrong, it's hallucinating here, it's missing that context. The harnesses that have been built, the model improvements, have gotten it to where enterprises now and businesses are ready to run with those. What is still missing, though, and what's actually holding back a lot of those is that organizations and those governing teams don't really feel they can trust what is happening. And it's not so much the AI system is not achieving something or doing something. It's the, I don't know if I'm comfortable with what actions is it actually concluding and taking. And are we comfortable with that? So let me give you an example because it's rooted in, I think, the root limiting factor that is causing this to be a delay in the industry. If I was to say to you, hey, I have an AI agent. It wants to do a bulk delete of opportunity records from the CRM. Okay. Traditional agent guardrails, traditional agent controls and policies. Is that a permissive action? Is that okay to happen or is that not okay to happen? Like, what's the answer? It depends on the context of the purpose. An agent doing that at the end of the quarter as part of a routine CRM cleanup, that is a permissive and sanctioned action that's approved. An agent that is trying to optimize conversion rate for a BDR that's just trying to hit quota and concluded that I could delete a bunch of records and my attainment percentage would go up is a not desired.

Jason Hiner: Sure.

Blake Brannon: So governing with static rules and static guardrails that are deterministic just doesn't work. It's like the limiting factor. You need a notion of what we call purpose. It's like, well, what's the intent and what's the purpose that you're doing this for? And out of that, you're able to map, well, what's the policy decision here? And this is why, getting back to your question of like AI, you kind of need guardian agents, not just here's some firewall that we can deploy. You need this intelligence system. It's rooted in exactly what I just said. These decisions of what's okay to happen and what's not okay to happen, they're not deterministic. They're not these fixed rules that we've traditionally been able to run security environments and firewalls and SOC centers and things off of. You have to have this new intelligence layer to help derive what the decisions could be. And where this becomes really interesting for OneTrust and how we even got here is you go back to, well, where do those policy decisions, where are they rooted in? It's rooted in the risk and governing process and programs that companies have been operationally running in OneTrust for the past 10 years. So your compliance obligations, your data ethics obligations, your privacy obligations. These are the data points that tell you what the answer to every one of those scenarios will be.

Jason Hiner: Very good. This reminds me, tell me if this analogy works. I've seen some talk about that today in places where they have some challenges with borders, where they're trying to manage that, and they're like, walls are not very effective because you can always build a taller ladder. So instead what they do is they deploy solutions that are like they use drones. They use on-the-ground drones. They use in-the-air drones. They use a combination of the two. And there you can, one, take photos of things. They can send alerts to people that sort of need to be alerted that there's something happening. And so it's just a sort of much smarter, more adaptable and more kind of real-time solution versus, you know, building something that, you know, is very static and easy for somebody to, you know, a sort of a smarter, so you can't battle intelligence with sort of something that is not intelligent.

Blake Brannon: Yeah, just a bigger wall. It's not going to work. There's two concepts. So it's a great analogy. And there's two concepts I think about when I think about what you just asked. One is we just saw last week with an AI system breaking out of its own harness and sort of hacking Hugging Face, right?

Jason Hiner: Yeah.

Blake Brannon: So it's like case in point, a wall around your thing, it's going to find a way out. You almost need to think about where's your containment boundary outside of that, that it doesn't matter if the AI system breaks out of its harness. Stuff you care about, which is your enterprise data and apps, is still kind of governed and protected, right? So you still are able to avoid losing data or have an independent layer there that's like no matter what the AI system does, you can still protect your enterprise resources in spite of that. The second notion, you talk about the intelligence system and sort of stopping. One of the things that we think a lot about and how you're going to govern at scale is there's two concepts in governance that I call containment and alignment. Containment is how do you contain the AI from its permissive actions? And I think the best way to describe this is think about the analogy of the auto braking system on a car. So traditional security guardrails, governance control tools are almost like brakes on a car. It's going to prevent something bad from happening and sort of slam on the brakes. That works to contain things and is very important. But if you can imagine a bunch of AI agents in a company that are constantly being stopped, they're constantly having their brakes slammed upon. They're not going to move very fast, are they? Right. What you have to do is you have to teach them how to drive. You've got to build a full self-driving system. And this is what I call alignment. And the notion of this is inside of those AI systems as they’re reasoning, if you can give them the context of your risk decisions, your security framework, your compliance obligations, your data privacy, your brand promise, your data ethics principles, while that AI system is determining its plan, it can pivot and adjust it. It can make sure that it's not going to go down a roadway that it's forced to slam on the brakes because of some AI guardrail. It's going to know that I yield this way, I accelerate that way. And that collectively, if you've got thousands or millions of agents that are doing that, it's going to enable the whole superhighway to go faster. So you need to encapsulate both concepts when you think about AI governance because they're equally important, but for very different reasons.

Jason Hiner: Very good. Okay, last question on the AI governance piece, because there's this constant tension between when you govern something and you put systems in place to amplify and guarantee safety, there is the possibility that it could help you go faster and there's a possibility it could go slower, right? It could help you go faster if it couldn't sort of remove the overhead of uncertainty where now all of a sudden, okay, if we have this confidence that this is safe, then lots of people can sort of innovate. They can work on their things without sort of that concern and having to put extra protections in as one-offs. But then there's the worry that if we put those protections in, it will keep us from going fast. But then we're going to have all these checks all the time that we have to go through that are going to keep us from deploying something or creating something because then it always is going through these checks. I'm sure that you must work around this question kind of all the time. How are companies navigating this when it comes to AI? The balance between AI allows us to move quickly and do things that we couldn't before at a pace we couldn't. But at the same time, everyone is becoming cognizant of the fact that it creates a whole new set of risks that we have to be aware of.

Blake Brannon: Yeah, this is spot on. I think one of the North Stars of a good governing program to see that it's healthy is actually your speed of doing governance. And we have an expression we talk about with our customers that we are a platform that sort of gives you governance that helps you go. And the more that you introduce siloed risk processes, security reviews, configurations and different tools and all these things, it just adds to that sort of time to go from we want to do this to we can actually do it. We've seen this movie play out multiple times with mobile and cloud technology where the business will find a way through with shadow IT or shadow mobile or now in this case shadow AI of finding ways around it effectively. So as an organization, you want to actually figure out how do you say yes confidently? How do you unleash the agents but stay in control? And this starts with, I think, the right architecture of where do you try to put the governance at and trying to keep up with every AI harness and tool and configure. Here's the settings in this AI tool. Here's the settings in that AI tool. Here's the, oh, by the way, the model just updated this week, broke half of that stuff. The harness updated, broke half of that stuff. Now there's this other issue. It's almost like a losing battle by design. Versus if you focus more on, well, what doesn't change and is static? It's things like our enterprise apps and data. And all those AI tools have a common substrate that they connect to with configurations like MCP, A2A, where they're connecting your services. And you can actually enforce pretty independently, so not tied to one single system, a independent way to govern it. And that's kind of the approach we've taken. And as long as you're able to codify the governing decision loop of the time it takes for a security or compliance or a risk team to effectively make a decision on, is this okay for this AI to make this call, to take this action? If you can move that from weeks of time to seconds, and you can have a simplified enterprise architecture of where you do the governing. It actually enables you as an enterprise and organization to say yes to all these AI tools. It doesn't matter if this department wants to use this new AI tool, that department wants to use this AI tool. You can say yes without giving up the control. And that, I think, ultimately is this good design between the two where you're embracing AI, but you're still protecting what you care about most, which is your apps and how the data is being used.

Jason Hiner: That reminds me of another question I want to ask you, Blake, which is in a previous world you worked, or previous life, you worked in a company called AirWatch. And I previously, in a previous life, worked at a publication called TechRepublic, which covered enterprise technology, which covered AirWatch.

Blake Brannon: The universe brings us back together.

Jason Hiner: Exactly. The universe bringing us back together. But there are some interesting parallels between that world and this one. There, you had all these people using devices, bringing them into the enterprise. They wanted to use them. You almost couldn't stop them from using them, right? It was really impossible. So companies used products like AirWatch to then say, okay, how can we let people use the device they want to use because they're going to use it anyway? But we can create some safe guidelines for them to access their email and their company tools and data. There are some interesting parallels to where we are now, right, with people using AI tools. What is that? Where do those worlds sort of collide for you? And how are there sort of similarities that, yeah, that you've been able to sort of carry forward some of the lessons learned?

Blake Brannon: Lots of learnings. And that wave was something called bring your own device and bring your own application. And the concept was that there was such innovation that was happening at the time on a mobile phone. And for the listeners here, putting yourself back when you kind of had a BlackBerry for work and it had your work data on it, and then you started carrying an iPhone or an Android for your personal, and you started asking a really good question, which is, why do I have to walk around with two phones everywhere I go? And then when I go to book a calendar appointment here, I forgot I had a calendar appointment there, and, you know, it was a mess. And I found a really good note-taking app over here. Why can't I use that over there? There's a couple of parallels that we see in this AI wave. One is that the innovation and demand from the business is kind of faster than what IT feels they can keep up with. So there's all kinds of great, at the time, mobile apps, platforms. And the story that told out was that you couldn't have just picked like one horse for everything, right? You couldn't have just said, it's just iPads. It's just Android tablets. It's just this Samsung phone versus this Apple phone. You had to support a heterogeneous environment of different devices and different applications that were best fit for a retail kiosk of this type of a store, a phone checkout experience for that type of a store. All of it had to be ultimately proved to be, you know, winning. It was a very fragmented ecosystem. The second thing that we noticed is that if organizations didn't embrace this shift, right, they just said no, they put up the big wall. We're going to block it all. We're going to pick, you know, just this one AI tool from this one provider that we have an existing relationship with. You slowly got out innovated and your employees didn't want to work for you anymore. You lost the kind of pulse in the market and you ultimately ended up having to embrace it and change over time. And I think you're seeing both of those play out with what's happening in AI. And you see solutions like OneTrust coming to life for AI to sort of help solve that. Again, it's how do we embrace a vendor-neutral approach with a heterogeneous AI ecosystem where you're going to have some Claude, some ChatGPT, some Copilot, some Glean, some Agentforce this, some ServiceNow that. It's going to be all of it, and you've got to have a neutral way to govern and control it. So that every time the business wants to change from Agentforce to this other marketing agent tool, you're not having to re-architect your governing and security architecture across your enterprise to support that new system, right? You've got a common place to do it, and you've got a common neutral party that can do it across those different environments.

Jason Hiner: Very good. All right. I'm going to ask you the two questions I ask every guest that comes on the podcast, which the first one is, in the age of AI, you know, we've seen there was this promise that AI will come and it'll take the stuff you don't want to do and you'll have more time to do the stuff that you want to do. And it sort of sounded very idealistic, you know, like, oh, great. I'll probably have more stress in my life. But the reality is most of the people I know are working on more things and faster. And so, in fact, they're more conscious about their time, you know? And so leaders, I hear leaders talk about this idea of leverage a lot. Like, how do I get maximum leverage for my time? So my question for you is, you know, in the age of AI, what's your best tip that you use and that you recommend to others in terms of how you get maximum leverage for your time in the age of AI?

Blake Brannon: Yeah, it's so true with the time suck. My wife actually even this weekend said, is this stuff actually speeding up the time for you to do something because it seems like you're spending a lot more time on it. And, you know, I think it's my my thought on that is because I would never have signed up to do what I'm now signing myself up to do with so many aspects. And because it all feels like it's just five, you know, I'm on the five yard line. Like I just got a little bit more and I can do this and I can do this. So I end up signing myself up more. So in aggregate, it is spending more time doing it because I'm just having a blast and it's so cool to be able to personally do so many things. But, you know, it does speed up time for individual units of work that I would have done. My guidance to people and what I've found works really well is I think communicating your thoughts and ideas is an art of storytelling and storytelling and trying to get people brought up onto a concept or an idea would used to require reading a lot of text. You had the famous Amazon, one-pager, three-pager type things. You had a lot of pre-read, pre-watch meetings that you would try to convey ideas to bring people along. It still just never worked. What I really like doing that I do think works well now is not having AI generate text and documents, but actually visual artifacts. And there's something that is so much more powerful to convey your ideas and to get people to understand and to weave a story together. When you have visual artifacts that can animate, they can have depths. Like if you write a paragraph and a page and you've got three or four levels of depth. Well, in an HTML page, you can nest that through clicks and tabs. And there's all these really rich artifacts that have been created to help structure that content in a digestible way and let people consume on their own journey. So I found that that is really powerful. And it's not time saving from the creator side as much as time saving on the how do you get people to understand and be brought up into the point you're trying to make or the narrative you're trying to drive or the concept you're trying to convey. And that would take weeks and weeks of time in the past. And now I feel like I'm able to convey that with, you know, just sharing a link to something and letting people kind of go through it at their own pace.

Jason Hiner: Wow. Okay. So you're saying you're using some of these like image and video models to create content that explains concepts in ways that are sort of much more condensed. You can, in the same way, like a cartoon in the newspaper, you know, like a commentary cartoon could say in one image, what would take, you know, a thousand words for you to read.

Blake Brannon: Nailed it. Yeah. Image is worth a thousand words and an animated image is worth whatever, millions of words.

Jason Hiner: Amazing. You're the first person that's come on that has ever said that. So kudos. Good job. That's a great tip. All right. The last one is... What's the AI tool that you're using right now that's making a difference for you? Maybe something that people haven't heard of before or thought of or used in an interesting way. Is there something that you're using that's really being helpful or useful for you that you think people should consider?

Blake Brannon: Well, I'm sure people have heard of this before, but I'm a big fan of, and there's multiple ones, Wispr Flow and various different ones—voice to text translators that just embed throughout your computer's operating system. And some of them are really smart where it knows if you're in the Slack channel, you're in the Outlook window, you're in the Claude Code, you know, CLI prompt. And it can change the summarization of what it's saying. But to me, this is game changing where I can just riff for three minutes about this very unconstructed, rambling, change my opinion mid-sentence thing. And then it synthesizes it down and can make sense of it and kind of end up taking and updating what I'm trying to update. And I find that that is just game changing for me with how I interact with a human. And if you think about the physics of you can only type so much, you can speak a lot faster. So your I.O. between you and your ideas and what you're trying to achieve in the computer, orders of magnitude better.

Jason Hiner: Yeah, it's so true. Thank you, Blake. It's a good one. That one has been mentioned before, but it's worth, your actually example is a good one of like, you can ramble on for three minutes, for example, to a chat bot or even more to like something that has an AI agent. Like, here's what I would like you to do. Here's my three minute description, go and do it. And it can then structure all of those thoughts, go do something and come back and create something for you, which is, like you said, it's a pretty big game changer.

Blake Brannon: It can be.

Jason Hiner: Amazing. Blake, thank you so much for the time. Really enjoyed having you on the Deep View Conversations podcast. Great stuff. And good luck with all the work that you're doing. Obviously, it's really important. Security, compliance, all of these things. We hear it all the time from our audience. They're trying to move fast and they're trying to stay safe and some big challenges. So yeah, you all are in a good place to help people with that. So good luck with all of it.

Blake Brannon: Thanks for having me, Jason.