AI Governance: Enterprise Controls, Code Bans and Body Cameras

Hey everybody, welcome to The Techstrong Gang. It's Wednesday, otherwise known as Mike, Mike, Mike's favorite day, Wednesday. Hi.

Hi, Mike. Let me introduce our guest for today, John Swartz. Good to see you, my friend.

How are things? Hi. I'm good.

It's good to see everybody. All right. Terry Robinson, good to see you.

You're on mute, but I think you'll come off mute in a minute. That's because I was sneezing, so. Oh, no worries.

We all- I didn't spread any germs, so there you go. It's been an interesting week on the technology front, that's for sure. And JP Morgenthal, I haven't seen you in a while.

How are you, my friend? I'm doing well. How's everybody else today?

I hope we're all good. We have started this week talking about AI in the enterprise, and on day one, if you watch the gang, we were talking about whether or not the enterprise was ready for AI, and the general conclusion was possibly it is not. And a lot of that has to do with the fact that, well, maybe their data's not in place.

Day two, we talked about whether or not there was an ROI and what that might look like if and when it ever shows up. And it would appear that smaller companies have an advantage over larger companies because the larger companies are, again, having trouble with their data. Now comes day three, and yesterday there was two competing conferences.

There was an Oracle event talking about what they were doing with AI, and SAP was talking about what they are doing with AI. And in both cases, they were both suggesting that they were going to show up and finally enable the enterprises to take advantage of AI. John, I know you recovered the Oracle event, so I'll let you describe what happened there, and then I'll jump in with the SAP stuff.

Okay. In homage to World Series week, didn't Yogi Berra once say, "It's deja vu all over again"? Something like that.

So we've got both vendors who are claiming their architectural advantage to solve this enterprise trust crisis, which kept coming up. Actually, Oracle executives were actually acknowledging that. And I'll go briefly just on what each company announced.

So SAP says they're embedding AI more deeply in enterprise workflows through Joule. Oracle proposes, on their side, security and governance built into the enterprise stack. So, SAP specifically is expanding Joule agent discovery and policy controls as it moves into business workflows with AI.

And again, specifically with SAP, they're embedding AI agents directly inside core ERP supply chain and HR processes. They want to eliminate hallucinations, drift through hyper-specific localized business data, and define operational paths. So basically, what they're trying to do is they're trying to gain our trust.

And this is something that the Oracle executive said is the biggest challenge that they face today in terms of cracking the enterprise markets. He stressed that anchoring protections into the foundational stack lets companies deploy autonomous agents while preserving command over their assets. Now, my takeaway, and maybe I'm being a little bit too cynical or too harsh, Mike, but I've heard this story ad nauseam, especially from SAP, and I'm just starting to wonder.

We are getting so many keynotes and so many press releases on this point of view, but I think what would really kind of prove to me that we're going to make progress, especially from these larger companies, are some sort of independent benchmarks or audit frameworks. I think there is a lot of skepticism because we hear these announcements, and then several weeks later or during the same week, we hear about rogue AI agents. So there's kind of a conflict going on, and we're kind of in this netherworld right now, and it's confusing, and I think it's alarming to the enterprises.

Yeah. I think we've seen this drama play out before with emerging technologies, but in the case of SAP, they might be onto something. And I say this because a lot of the folks that I talk to in the enterprise are having a hard time getting all this stuff to work, and that's why you're not seeing a whole lot of adoption.

There's too many damn piece parts. Can't figure out how to govern it, and I can't figure out how to secure it. " And to your point, worse yet, we don't trust all these AI agents.

So, it becomes interesting to me. I think in some ways, SAP even went so far as to say, "Don't rely on a general purpose AI because it doesn't have the context for your data. " I don't know how true it is, but it sure sounds good.

So, in fairness to SAP, perhaps this is what's happening, you alluded to it clearly. Maybe they announce something, they get a lot of feedback from their customers, and then they revise what they're doing, and it goes back and forth. I give credibility to companies that do that.

Market research companies that would revise their results or their studies, I had a lot of respect for because they acknowledged where things didn't go wrong, but they went a little bit awry, and maybe that's what SAP is doing. In the case of Oracle, I really don't know, but maybe this is what's happening. Yeah.

I think, in theory, the larger enterprises should have a major advantage over smaller players in the age of AI because the problem with AI is you need a lot of data to make it really work. Well, enterprises have a lot of data, but it seems to be locked up in all these different silos, and it's hard to get at. And all that data, shall we say, is conflicting.

So now that I got JP back, I'm going to ask him. I was on, my monitor went blank, and I'm like, "Okay, that's weird," but I can't see anything. You've been warned.

The rogue agents. We had a similar issue yesterday when I was in an undisclosed location in New York, and I had to go shoo people away from the front door. So JP, let me ask you something.

What's your take on what's going on here? Because initially when we had all these AI agents, it was going to be the land of the free and enterprises were going to be able to go out and create any kind of workflow they wanted. But, upon further review, it looks like it's a little too hard for them, and there's too many piece parts.

So, is what SAP and Oracle are starting to say sound compelling? Well, it's what I imagined was going to happen all along because nobody knows the SAP and the Oracle EBS environments better than those companies. And there's a lot of nuance to those environments.

Over the years, we have seen how much it costs to customize those environments, to create the business rules and the business policies in those systems, which tend to be mission critical in running the business. And so, that content has not been available to frontier models for training. And so to expect that those AIs would be useful in accessing those environments without governed access from those vendors is really questionable.

Now, I know some people have gone to the data layer themselves through APIs and created MCPs, and I actually think that it's safer leveraging the Oracle and the SAP access mechanisms. And if Oracle... Now, here's the rub.

Is Oracle affordable on the AI request? Is the AI that they're providing you comparable or lower cost than you would get from a frontier model? And hopefully, they're smart enough to realize we need to do that in order to Keep the customers in the environment.

And if they do that, then Oracle and SAP end up having a very special persona, if you will. My persona as the agent is the enterprise app agent universe. And I'm comfortable touching everything because the guardrails are already built into the system.

You're using my model, so I can confirm that the data flowing in and out is not leaking. And I also understand my environment. So I'm providing you the tools to access the data as long as you use my methods.

And that's really, first of all, a big part of what I have been talking about to customers around AI in general and agentic behavior, which is it's an encapsulated entity. I hate to get technical for people who aren't developers watching the show, but it really stems from the properties and principles of object-oriented programming. It's polymorphic, it's encapsulated, and so atomic.

Those are the things that we want in our agents. So we want that agent to have its own persona, to be encapsulated, to be safe. And the best way to get that is for the agents to be provided by the environment, and not having businesses starting to build out this new, quote-unquote, "API layer," which I think is much harder to control and much more likely to end up in a negative event or what we call a hacker event, a security event, in the future.

So, John, the other thing that comes through with all of this is that we seem to be engaged in what I'm going to call an AI agent supremacy battle. Yes. " And therefore, what they're saying is that their Joule framework will become the dominant framework within which employees use to access data and marshal these AI agents, and maybe I won't need all these different AI agents from 100 different other vendors.

What do you think? Well, in terms of simplification, I guess, that would be appealing, but then in terms of data dominance and being a monopolistic kind of viewpoint, that'd be a little disturbing. But the thing that I keep thinking about is, so we keep having these issues with agents, and everybody wants us to trust them.

Now, do we reach a point where we keep having modifications in these packages where the enterprise or the customers becomes a little bit antsy and just starts looking around for their options? Are they just going to stick with and be locked in with their vendor of choice and take their chances? That's purely speculative thinking, and it's probably playing down the road, but I start wondering when we actually reach that stage where we say, "You know what?

Maybe we should try something a little bit different," or maybe we shouldn't put all of our trust into one vendor like an SAP. Well, to your point, I was having this conversation with the CEO of Everpure last week, and he was pointing out that we're going to need some sort of data layer that is agnostic to these AI agents, because otherwise, we will continue to be locked into all these different SaaS platforms, and you'll have all these AI agents that are trying to talk to each other, and the whole thing will become unsustainable from his perspective. Now, the issue is every time you buy an application, that software vendor has control of how that data is stored.

And so then you have to figure out all these integrations that go around that. JP, I don't know, is it possible to have our cake and eat it too? Will we get to some sort of hybrid where the path of least resistance will be an AI agent from an SAP or an Oracle, but we'll also go out and build our own AI agents and maybe get our own unified data layer someday?

What do you think? I think it's a crawl, walk, run scenario. I think the CEO you spoke to at Everpure makes a valid point.

But trying to solve for that right out of the gate means you're going to lose the race. You're going to trip up, fall, and screw up right in the first lanes. The opportunity is start small.

Let these things own the agent for a while. They have all the resources. It's like the early days of cloud.

They know how to build it. They know how to manage it. They know how to control it.

Learn from them. Use the vendor appropriately. Use that relationship appropriately.

Come in, make sure that as I start to build out, you have time because you have the services you need. You have the functionality you require. " You have the services.

It's just locked in a little. Okay. All right, so let's not try to figure out how to be completely vendor-neutral environment right out of the start.

Just work with what you have. Enable that to use those services that while you have them, and while you're learning to set up and that you can test and experiment to make sure that when you want to go from crawl to walk, that you have all the right guardrails in place, that you've tested it to a degree that you feel safe, and then move to that stage. And so yeah, eventually you can work your way there.

Just don't try to do it right out the gate. So I see this fundamentally, this built-in contradiction is going on. Right, we should be cautious, we should move with nuance.

But there's all this pressure coming from the top. They want instant results. We kept hearing so much about the C-suite wanting something overnight, 30% efficiency, whatever the numbers are that they were bandying about.

And now the actual real situation calls for patience, and the irony is that this whole AI age is that patience is almost the last consideration. It's all about speed and jumping ahead of one another, leapfrogging, getting to the final stage first. Sure, no problem.

Read the story about Hugging Face and what happened to Anthropic. Mm-hmm. And then let me know if you want me to move ahead.

And if you do- Yeah ... that's great. You sign off on the email, I'm done.

That's all you have to say. The risk is on you. You sign the SOX, you sign the paperwork.

You're a public company. You got SOX on your side. You're the one who's beholden to shareholders.

I'm more than happy to do that work. You sign off on it as the CEO, but first, you read this. Terry, is the world upside down?

I can't believe what I'm hearing. Companies are actually slowing down and not experiencing as much FOMO because of governance and security issues. What universe do we live in?

One that I'm not quite sure is real. I don't think so. Look, I think sure, everybody's concerned about governance.

Everybody, more recently that I've spoken with, are paying lip service at least to the idea of slowing down and having governance be more prominent. " Well, maybe so, but what is that action going to produce? I think it's interesting at one of these announcements, it's government sort of at the data layer.

That's kind of interesting. That's a bit better than what we've had before. But I don't know.

I just don't have a lot of faith. I just think that companies make decisions, again, like we've talked a million times on the show, about what's better for their business proposition, especially in the short term. I don't see- Oh, no ...

governance winning the day, if that makes you feel better, like we're in the same universe that we were in yesterday, Mike. All right. So let me just ask this last question here.

What are the odds that there'll be some other AI agent layer that emerges? Because I can access SAP data from an external agent, and I can access Oracle and Salesforce, and I don't know, JP, is there an opportunity for some other player to emerge that provides all the benefits across all those platforms, and that's where the next big battle's going to be? I think they exist out there already, but it's a trust issue, and it's expensive to implement those layers.

I see a lot of businesses also doing it for themselves. " And the next thing they're looking for is, how do I manage identity in this universe? How do I ensure that the MCP, and the request, and the prompts coming in from the models also carry with it the necessary information about who's making the request?

And that's kind of where I would say businesses are with regard to that endeavor. So they are right now providing it through custom interfaces where they have built out these custom apps to get certain pieces of data, and they are putting the identity on that. But of course, there's a slowdown to that in that you can't anticipate what everybody's going to want.

So to open up the real flexibility, you have to open it up freely, and so they're trying to figure out that layer. And once they figure out the identity layer, and there are answers out there, again, I think that'll be something that will be leveraged architecturally by-- It's an architect that'll be repeated by many companies. And so I don't see it as one vendor stepping in and providing the answer, but I do see products all being joined together, like an Okta and a Snowflake and things like that, that kind of tie together in a consistent archetype that companies can repeat that architecture to create that data layer for themselves, and then manage it appropriately about who's got access to it.

All right, folks. You heard it here. On this particular issue, pull up a chair and welcome to the great AI agent rumble, because that's what it's going to be- ...

for the next six months or so while we figure all this stuff out. Shifting a gear. If you've been watching what's going on in the open source community and with AI coding and contributions of code and reviews, there's a lot of tension in the world, and the tension seems to revolve around, well, two things.

One is there's too many contributions now being generated with AI, and not all of them are of equal weight, and a lot of them may be slop, and the maintainers are overwhelmed, and they're complaining. " And secondarily, you also have Linus Torvalds saying over in the Linux community that despite what anybody says about AI, he's going to use AI to help build the next kernel and going to contribute to those projects. But JP, I've got to wonder, are we going to see some more level of, shall we say, maturity around how these things are built in the open source world?

" And now everybody's putting, I guess for lack of a better phrase, guardrails around who can contribute. Is that where we are? So this is based on some research that you all put out, and the answer was that maybe Linus has it right and Reddit has it wrong.

I think the issue is one of trust. The Linux community is established. It's well-known.

It has its processes. It's all established. So adding AI into that is more manageable because now it really doesn't change significantly the process of accepting code into the kernel.

That's still going to be the same one, which is the review process by engineers on the team who are trusted resources. And so if somebody drops a new module that was written in AI, it's still going to go through the current review process. Now, the Reddit issue is one of, I have no idea who you are.

You could be China dropping all kinds of bombs on here and creating venues of terrorism using our platform, and I can't allow that right now. So until I figure out how better to manage it, I'm going to put these rules in place that kind of give me a trust factor, like the Linux community, that you're a real project, that you're here to stay, that you're building something for the long haul if possible, if you can build a community around it. But you're going to demonstrate to me, even if you are using AI, that you're a real community and that this project has some legs to it.

So to me, it comes down to one of, it's not one's doing it right, one's doing it wrong. It's a matter of I'm dealing with two different situations, and I'm trying to manage them based upon the data that I have. Mm-hmm.

Terry, what's your sense of how aggressively are adversaries starting to poison these open source projects? Because I think to JP's point, most of the conversation has been around being overwhelmed by the amount of slop that I have to review. But I think the unspoken part of this is that malicious actors are now contributing to these projects, dropping in code that has malware that they're hoping will be distributed downstream and activated six months to a year later.

Sure. So honestly, yeah. I really think that's always been the fear, right?

That that's what would happen here. And I think it shouldn't surprise anybody that that's the case. This whole thing is kind of interesting.

It's a little mind-boggling... Not mind-boggling. It's a little jarring because you think of Linux, right, as the sort of rebel among the entities out there.

And now, JP, by the way you describe them, they've become like the mature adults in the room and well-trusted and everything, and now you have Reddit and open source. And part of me feels like, I understand why you maybe need some rules around it, but part of this also feels like it's throttling what open source is too. But I do think, yeah, that certainly I think we're already starting to see that there are bad actors out there that are taking advantage of the openness of open source and trying to stick code in.

It's true, and the maintainers are being overwhelmed. And I guess I'll throw this to JP, but do they have a certain amount of egg on their face? Because for years they always said the great thing about open source is it was peer-reviewed, and we'll be able to catch all these issues and bugs.

And along comes AI, and suddenly we have thousands of issues and bugs. So I don't know. Is software just software?

No matter how you develop it, there's going to be vulnerabilities in it unless we find some magic silver bullet that I don't know where exists. Well, it is that, and it's the volume and the velocity which is different, right? When they were discussing it, you had humans doing the coding, which produces a consistent and predictable volume and velocity.

And then AI can produce at thousands to a million times faster on both of those fronts. So I don't think you could've planned for that coming and then what impact it would have on an open source community to be able to respond to that. And so I think maybe part of the answer is AI, right?

When we get to a point where we have AI review processes that we can trust and that we'll be able-- Some of that happens today. I have plenty of customers that I work with, they have- ... remediation as part of the PR process, right?

The code gets checked in, it gets checked for bugs, it gets checked for vulnerabilities, and before it gets tagged through for a human review for release. So that review process is stronger even in some of the AI companies who have really had the opportunity to build and experiment around that, and get their prompting and the models working properly to do automated review or autonomous review. So as autonomous review gets stronger, more mature, then you'll be able to throw it at the open source projects, and you'll be able to count on it.

So eventually, like a Reddit, open source forums would be able to say, "We have autonomous review, so it's okay. We'll be notified before anything gets released that's of risk," hopefully. Right?

How long is that gap going to be for? Because right now, when I talk to people, they are too overwhelmed to review and validate all the code being generated by the AI coding tools. So that's just getting passed through to production because, well, we've got to prove that we have some ROI on our AI coding efforts.

But more of that software than ever might contain vulnerabilities unless somebody maybe scanned for it, which they probably didn't. So, in the short term, are we going to be worse off? And should I be- I think we're comparing apples and oranges right now, because what you just described was the enterprise problem, where command and control has no real understanding of the SDLC and the impact, and they're just throwing out edicts around managed objectives that make no sense because they don't really understand what's behind the process or the risks.

And so you're taking that scenario, and you're throwing it at open source, which doesn't have that same time to value problem, and there's no reason for them to rush to market or rush anything through as an enterprise would, where people are going to shortcut and say, "Hey, I did my sprint on time, make sure I get my bonus," or, "Make sure I don't get fired," right? So there are a few different scenarios, and you just mixed one with the other. But a lot of those folks are using AI coding tools, and one study just suggested that there's more open source code being generated by those tools inside enterprise apps because the AI coding tools are pulling from the open source project.

So a lot of folks are inheriting issues from the open source community that they are now passing through into their production environment. So I'm just sitting here wondering, are we looking at some sort of brewing cybersecurity crisis that's going to explode in the next six months until we get our validation and review processes together? Probably.

I always think that's- A lot of people are shortcut-- I'm sorry. Go ahead, Terry. No, I was just saying, I always think that's the most dangerous part is that gap, before you get the rules and the structures in place or the guardrails in place.

And I think you're right, Mike, you're seeing a lot of the open source, and it's making its way into the enterprise. I've had maybe a dozen conversations with people over the last couple of weeks where that's come up. So I do think it's a bit of a dangerous time.

Maybe JP, we're squeezing oranges and getting apple juice, but you know. First of all, I think that there's a lot of technology out there today that can be used to do remediation. So if this is the case, it's actually very positive for open source.

If businesses did what they were supposed to around open source, which is if you use it, contribute your changes back. So now you would have this huge community who has big pockets, who can go afford to use GitHub Copilot and Devin, and SpaceX and SI, and things like that to do remediation on a large scale. And part of that remediation will be against these libraries that are pulled in, and it will identify the problems in them.

And wouldn't it be nice if the businesses actually contributed those remediation changes back to the open source community? You'd have a much larger community now doing remediation for you. Well, do you think that takes some convincing, or is that just a matter of people just getting on board, and it takes some time for them to change their thinking about this?

I think it's happening, but it's slow. And I think if you look around, you see people starting to play with AI testing tools that validate code, and at least they're doing something. They'll look at that code before it goes through, and then they've got to re-engineer their pipelines to do that.

I also think you're seeing cybersecurity people looking at tools that enable them to discover vulnerabilities faster and put some type of virtual patch or control in place while they wait for a regular patch that at least will mitigate the threat. And then I think you're seeing an open source community that is rallying together to share some of their work and their patches, and you see IBM and Red Hat trying to drive that conversation along with the Linux Foundation. And so I think all those things are in motion, but I think it takes the average enterprise half a year to operationalize all that, and meanwhile, the AI coding train keeps chugging along.

So I'm just wondering how long this gap is going to be, because it seems like it's going to be at least half a year. JP, fair? To what scale?

To 80% adoption? I think that'll take at least a year to get to that level of consumption and deployment. Well, if that's the case, then the first half of 2027 promises to be one hell of a- Hard ...

security moment because there's just going to be a lot of issues, and there might be, who knows, a breach every day. Cross your fingers and hope for something better, but we'll see what happens. Well, remember this, the thing about breaches, so if you actually look at cybersecurity, and I don't know why this came up for me a while, about a couple of weeks back.

I started doing, "Oh, I met a woman on a plane who had been hacked," and she had all these problems now that she had to fix because they stole her ID and everything. And I started telling her about all these tools that she can personally use to protect herself, and she could use MFA and all these other things, two-factor authentication, make sure it's on all your bank accounts, right? So I walked her through all this, and then I started looking at some of the research, and the interesting thing is 90% of this stuff still isn't zero day, it's social engineering.

The thefts at the casinos for MGM started social engineering. This lady, she typed in, they got her through a phishing email, social engineering. So even if you have these vulnerabilities, the likelihood of them being found without some sort of social engineering targeting you is still very low.

" That's where I think we're going to be. All right. Shift into gear.

Terry, welcome back. Well, thank you. I'm sorry about the technical things.

I'm not sure what's going on. It seems to be going up and down. It's an issue for everybody lately, so I don't know what's happening.

I was just talking... And there's even a lag now, so I'm not even sure. Can you hear me?

Yeah, we can hear. I can hear you. I kind of lost Terry for a second.

All right. I'm going to try this again. I'm going to go to John and see if we can pull Terry in later.

So there's been an advance. I'm going to call it an advance. But if you've ever watched any kind of incident involving law enforcement where things went wrong, they were always looking through the body camera.

And the basic idea here is that we have some ability to go to what here in New York Warner Wolf used to call the videotape to figure out what happened. The thing is though, it's nice to have that record, but the people who manufacture these body cameras are now, I think it's called Axon or something like that- Yeah. Axon.

Is now trying to embed AI into the feeds. And the idea here is that- It's gone wrong. The hope is anyway, that the AI will detect when issues are escalating and maybe give some advice to folks, and maybe mitigate some of this stuff.

And I know there are probably privacy concerns, and I'm sure not every law enforcement official wants every interaction with the public recorded, but I don't know. Let me start with John. Is there more good than bad here?

When I read this story, it's a Brian Bellew story. It's a good story. What- True story ...

concerns me is, right, you can make arguments both ways. So on one hand, the AI-generated warnings could improve officer safety, but I think about the possibility of false alerts or inaccurate information which could influence decisions during tense encounters. And I also believe, I think, most police departments lack standard operating procedures, audit infrastructure, or accountability protocols necessary to govern how officers should interpret or document real-time AI alerts and proceedings.

So there is no real record or policy in place. So I see great benefits, but I also see really dire situations, not just in terms of privacy, but of reliability and accountability. It's a two-edged sword.

Right. Right. Well, so I also think, too, who's training these things?

Because what situations are going to come up as needing this kind of remediation or that, or this kind of action or whatever. So I'm like you, John, there's some positives here, right? You'd like to see more detail, you'd like to see greater officer safety, you'd like to see greater safety of the public when these encounters occur, for whatever reason.

But I am worried about how the AI part of this is trained, and then everything is interpreted. I also wonder, do they have something there that automatically turns the body cams on? Because that seems to be an issue that comes up.

Surprise. It'd surprise me. " "I couldn't get it to work.

" Yeah, exactly. So, I think if we could overcome that, that would be a huge hurdle. Yeah.

I live in a household with an 18-year-old who wants to be a cop, so- Mm-hmm ... quite often, I can't tell you how much I find him and my wife sitting there watching "Live PD," "Cops," or some other show. And then because of that, now I get all these reels coming up on Facebook that are showing the people take the body cam footage, and they post it, and they narrate to it.

So between all that, I have a definite opinion. One is, I think it's useful that the police would have a tool that allows them to keep an eye on the situation without having to remove themselves to go get data. I think that is fantastic that they don't have to remove themselves in order to be able to identify certain things.

Now, here's the thing, the problem I see based on what I see in those reels, is that cops have to make a decision quickly in many cases, but they don't ask enough questions. And I fear that they'd be making decisions based upon what's coming up or what's handed to them, versus turning that around into a set of questions to validate. And so often, you see in these reels, and maybe it's not as often percentage-wise as the number of interactions police have every day, but there's too many of them where they're like, "You don't have the right person.

That's not even the right person. " Right? And so that's happening without these.

What happens when you add these, and then on top of that, you add police officers that aren't willing to stop and validate and verify before moving forward? Mm-hmm. I think that there is this sense of urgency that's not always warranted.

But I think what I would hope for is that in some of these instances, and it always seems to be the one that goes wrong, is that somebody involved has mental health issues, and the police didn't know anything about it, and then they get there, and things go south. And I wonder if we could stream more information to them in real-time using AI, where they can have a better understanding of the circumstances that they're walking into, they might not even walk into it in the first place. They may just call somebody from the mental health department and say, "Come talk to this person," because otherwise, this is going to escalate in a way that's going to go south.

And those escalations, John, as you well know, they go wrong inside of 30 seconds. Yeah. Right?

I think that's going to happen, Mike. I think this will eventually evolve. I hope so.

Maybe through kind of independent third parties, companies like Axon can figure out a way to use the data to give us a more well-rounded picture of what's happening. And I know these events are happening in a short period of time. It's almost an impossible situation that most cops are put in.

But that would add a layer, and remove the confusion, and the haste of the crisis at the moment. But I don't know when this is going to happen. But it's going to evolve, is basically what I'm trying to say.

Could it possibly too automate that part where it pulls in the mental health professional? Mm-hmm. There'd have to be some sort of, I guess, 24-hour monitoring, or whatever.

Or it just even activates, "Look at this address. We know that there is a young person-" There's a history of... Right.

"... a history of blah, blah, blah," and maybe that helps. And I don't want to see officers lose also their intuition on these things.

The way you develop knowledge about how to answer a question or respond to a situation comes from your experience, right? And not from something just feeding into you and telling you, "This is what you do," and whatever. And I feel like with AI, and just in general, we lose a lot of that.

We talk a lot, again, about humans in the loop, but at some point, the humans don't have the sense and the knowledge that they had at one point, and the instincts. That's the overall sort of existential thing, I guess. But still, I think that it could be helpful, but there could be so many ways that this could go south.

All right. Well, I'm going to leave this here, but on behalf of Brian Bellew and myself, both of who, we have a lot of friends and cousins and relatives in respective police departments. And speaking on behalf of their families, none of their families want them involved in any of these types of situations in the first place.

Right. And I don't think most of the officers want to be involved in these types of situations either. And so anything that we can do to deescalate the opportunities for things to go wrong, I think that they'll be happier.

Where it gets rejected is when it becomes part of the, we're using this as a supervisory tool. Well, at some point, you have to let that police officer make some sort of judgment, but the more information they have, the better that judgment's going to be. Well, and that's why you get into the situation, right?

Yeah. Have that information before they step foot, maybe across the property line, or whatever. That's the benefit.

That's the good part. So everybody cross your fingers that there's more good than bad here, but one thing for sure, we're going to find out. Hey, I want to thank John, JP, Terry, as always, for sharing your insights.

45 minutes goes by in a flash when we're doing these things. And I want to thank you all for spending some time with us, and please stay tuned for the replay of the Techstrong TV lineup. And of course, Techstrong gang wouldn't be complete without a word from our sponsors, so stay tuned for that, because, well, it's us.

See you guys tomorrow.