234. “AI and Decision Risk” with Clyde Calhoun

Dawn talks with Clyde Calhoun, founder and CEO of Root Idea, about how rapidly AI is moving from marketing and copy into business decision-making and the risks that creates. Clyde discusses an incident where an OpenAI developmental model escaped a test environment, accessed the internet with stolen credentials, and hacked another company while pursuing its goal, illustrating unintended consequences when guardrails are unclear. He explains the difference between general-purpose LLMs and more autonomous agentic AI, and argues companies must address not only AI tools governance and data governance, but also “decision governance” as employees increasingly use AI—often without telling employers and without validating outputs—leading to costly incidents. He describes using AI “checking agents,” playbooks, documentation standards, and interrogation of assumptions to validate AI recommendations, and notes changing usage-based pricing, portability challenges, emerging overlay tools, and consumer risks in areas like medical and mental health advice.

In this episode….

Who is Clyde Calhoun and whom does he serve? (00:55)

OpenAI Breakout Story (01:13)

Agentic AI Explained (02:24)

Shadow AI And Costly Errors (04:38)

Why People Skip Validation (08:06)

Training Bias And Prompting (09:57)

Choosing Tools Employees Use (13:23)

Guardrails For Decision Governance (16:57)

AI Checks AI In Finance (18:51)

Trust Motives And Humanlike AI (21:11)

Pricing Tokens And Portability (23:33)

Futureproofing Decision Architecture (28:33)

Consumer Risks And Health Advice (32:56)

Where To Find Clyde (36:12)

Final Thoughts On Speed (37:55)

Clyde Calhoun’s Information:

LinkedIn: https://www.linkedin.com/in/clydecalhoun-rootidea/

Website: https://www.rootidea.ai

Transcript

Copy Transcript Copied

INTRO/OUTRO: 0:03

Welcome to Business Mastery with Dawn Kennedy, your quick under forty five minute dose of expert insights and strategies to make a positive impact on your business and life. Let's get started.

DAWN KENNEDY: 0:17

Hello and welcome to this episode of the Business Mastery Podcast. No surprise to anybody in the audience, AI is helping people make decisions. They're not just doing the copying anymore. They're not just doing the marketing anymore. People are actually relying on the results of AI prompts to make business decisions. So I have glide on here. We're going to be talking through AI and decision Rest. So thank you so much for joining me today.

Clyde Calhoun: 0:43

Well, Dawn, thanks for having me. It's a pleasure to be here. I love talking about AI. So any time we can have a conversation, it's great. It's a great day for me.

DAWN KENNEDY: 0:51

Perfect. Can you tell everyone who you are and who you serve?

Clyde Calhoun: 0:54

Sure. So again, I'm Clyde Calhoun, I'm the founder and CEO of Root Idea And we're a consulting firm that we specialize in helping companies make AI safe, particularly as employees start using it to generate information and make decisions.

DAWN KENNEDY: 1:09

Yeah. So AI save, so I'm just going to just dive in and we could bring up just to start what you were saying about last week, there was a little breakout from an AI thing. So can we talk a little bit about how fast this is moving, and then we can talk about how to make it safe for a company?

Clyde Calhoun: 1:27

Yeah. Sure thing. And so for those that have not seen that story so open AI had one of its developmental models that broke out of contain of its test environment, got access to the internet, used some stolen credentials and then hacked into another company in pursuit of trying to achieve its goals. And so that was the big headline. But I think the bigger issue around that story is not so much that it got out and hacked us as this other company. I think the bigger story is that it was trying to find a way to earnestly do what it was tasked with doing to accomplish its objective. And so it's one of those unintended consequences that the guardrails around what it could do or could not do in pursuit of that goal, those were not clear. And so I think as you talk about decision making, we have to be cognizant of how these other implications may be out there when we're dealing with something like AI.

DAWN KENNEDY: 2:22

Can you also break down for us the difference between like an agentic AI that's given a task and potentially marketing dollars, and then the AI that people are maybe prompting individually to, to get information. Yeah.

Clyde Calhoun: 2:37

And I'll just make it simple like this. As you think about the, the general large language models that people are familiar with using ChatGPT, Claude Grok, copilot, those types of solutions. Gemini is another one. By and large, people are using them to just get information back and forth. So you ask it a question, it'll give you an answer. Ask it another. It'll give you another. Another answer. With the genetic AI, it can not only answer questions, but it can go off and actually do stuff. So it can go out into your email and do some stuff, or it can go out and accomplish a specific set of tasks as it's been directed, and it can make some decisions on its own. And so it has a lot more autonomy than the models that most of us have been working with day in and day out.

DAWN KENNEDY: 3:22

So what when you say AI decision risk and making it safe, are you talking about both types of AI or more of the one where humans are making decisions on the feedback it's getting?

Clyde Calhoun: 3:34

I'm primarily taught well. I'm talking about both types really, because in in, in any case, whether you're using more a passive AI or whether you're using an AI agent, you have to think about how are people really transforming the way that they work with this technology. For some folks that are using, say, copilot, they just may be taking the information on a pricing analysis, and then using that to make a decision about what discounts they're going to give to customers. There's not an agent doing that, but they're still making decisions based on this information that they pulled out of this AI tool. In other cases, it could be like similar to the open AI situation where you've got agents that are tasked with doing stuff and the agents are making decisions, and you still need to have governance around that piece too. But the interesting thing about it is that so much of the focus so far has been on what I call AI tools governance. Which systems will a company have? Who can access it? How do we protect the information inside the company? Not nearly as much focus as gone to to, well, how are people starting to use these tools to make decisions and generate information and reporting that can go outside the organization or affect pivotal decisions inside the corporation?

DAWN KENNEDY: 4:50

Yeah, that's a great point because we have been reading all the articles right about, hey, please don't put your raw financial data into non corporate firewall AI because then it could be used to train or other people could use it. But there's a disclaimer on the AI could be wrong. Please double check before you use it. Why do you think is that effective. Like do people understand that those computers will have zero liability for what they decide to use with the data and implement in real life?

Clyde Calhoun: 5:25

And that is the gap that many companies are waking up to right now, because the data is pretty clear on that. So just a little, little background. So about six in ten employees. Yeah. Basically hide AI use from from their employer. So they're using tools that may not be approved or not. But even if they are using approved tools, two thirds of employees acknowledge that they don't consistently validate the output from AI. And about six in ten employees acknowledge that they've made AI mistakes at work. And so these things are happening. People are not verifying this information. And that can be very costly. So just another step. This is from Ernst and Young, but about sixty four percent of companies that make over a billion dollars have already had at least one negative AI incident that cost them more than a million. And the average cost of an AI incident is about four point four million. So the consequences for that are huge. But I say that to make the point too, that even if you haven't formally deployed an AI solution inside your organization, your employees are using it. They may not be telling you about it, but they are using it. And so there is some exposure that you need to deal with.

DAWN KENNEDY: 6:35

So let's talk about this as far as when you say six in ten employees, are you talking about across industries? Right. So we know that there's been lawyers that got caught and got sanctioned. And we know that there have been some highly regulated industries, like in healthcare where there's been HIPAA by stuff you're saying everybody. So it's quite possible that you're running a business with two or three locations and one hundred employees. And even though you don't deploy a tool inside your organization, it is more likely than not that your employees are actually using an AI tool.

Clyde Calhoun: 7:10

Absolutely. And that that cuts across industries. It cuts across larger organizations, smaller organizations. I mean, people are finding value in it because they're personally more productive. And for folks that have used the tools to even do basic stuff, whether they're looking at generating a report or a PowerPoint deck or what have you, they find value in it. And it's kind of one of those things that employees view it, much like when you're driving on the highway, if you're five miles over the speed limit, you're probably still good. I mean, it's still against the law, still against the policy. But if you're five miles over, maybe even ten miles over, you're probably still good. And that's the way that employees are truly looking at AI. So I always make the point that as a business owner or business leader, you have to presume that people are using it, and then you have to design your controls environment so that if they're using it and you don't know about it, are you still making good decisions?

DAWN KENNEDY: 8:03

Right. So going back to this idea that people are not validating the data the way that they're using it, they're just relying on the responses that come in, or I guess because I use it for a lot of research where I will say, provide me with the citations. And if there's a broken link, I will actually shake my finger and be like, now what are you doing? Right. You gave me a broken link. You tell me. You're saying that people are not actually going out and vetting the citations in their research, or are they just looking at market comparisons without making sure that everything is aligning? Or is this really people are just getting an answer and they're running with it?

Clyde Calhoun: 8:45

It's a mixture of all that. There are people that are just getting an answer and running with it. There are some folks that will say, okay, show me the citation for it, but you have to be careful even there, because oftentimes AI will give you the citation. And in a summary of what the citation says, it's not quite what the citation actually meant. So if you haven't read through the citation, then that could that's another point for it to go sideways. So those are some common ways that people miss on, on, on AI output. But even in terms of some, if you ask it to summarize some things, there are times where it can flake out a little bit and do something different. And the other example that I'll call out from the finance side, because we see people that are using AI to support analysis and account reconciliation or variance analysis, and the explanations oftentimes now are being generated by AI. And so you don't know, is that really what happened with the variance, or is it just the most likely or statistically most likely things that happened for that particular circumstance. And if someone's not actually vetting that out, you never really know. So those are the types of things that that can go on there.

DAWN KENNEDY: 9:55

So let's talk about that a little bit where AI is generating the summaries and AI is making those projections or doing that work. Let's just talk a little bit about the AI itself. I've read a few articles that AI generally wants to make you happy and wants to confirm your assumptions when you put a prompt. Can we talk a little bit about how they're trained and why this actually becomes sort of an issue?

Clyde Calhoun: 10:24

It is because it's kind of one of those things that if you work with it enough, you can convince it to get onside with your position, even if that's completely off base because to your point, it does want to, to please you. Again, the tools that most people are familiar with are creative engines. They're not necessarily deep analytical tools. They're not research tools. There are things that are there to support, to provide support. And again, it tries to do so in a way that keeps you happy and keeps you engaging with it for longer and longer. Yeah, it's really interesting. There was some data that came out a few months ago that while employees are generating productivity from the tools, they find themselves working with them more and more in terms of the amount of time they spend on a particular issue because they get sucked in, much like on the social media side, where people get on a platform and they get sucked in. The same thing happens with AI. And so that's the thing that you have to guard against, but recognizing that you can take steps to make your own personal GPT a little bit more realistic in the way that it gives you feedback. And some of that you can embed as permanent prompts in the background to say, hey, I always want you to take a pessimistic point of view or more realistic point of view. When you give me answers, you can mitigate some of them.

DAWN KENNEDY: 11:46

That's interesting, because I know that there's a lot of training out there that says, give it a roll, right? Tell it I need you to be my x, y, z. I need you to go out. I need you to look for the information that addresses whatever this is. Use sources that are less than ten years old. Let's say that's just some general prompt that you put in there, and it's going to probably perform and do that. But you're saying then you could also already embed in it and give me the worst case scenario and give me the most pessimistic thing and find the gaps. And why aren't we really necessarily training that when they're training very broadly? There's a ton of internet courses. Claude, in twenty eight days, I think I saw on YouTube a few a few weeks ago, I think there was something, Claude, in twenty eight days. Are they addressing these things up front, or is it really more just to make this more user friendly in the mind of the human?

Clyde Calhoun: 12:43

Yeah, that's a great question. And I actually flip that back to companies because what we see commonly is organizations will roll out a tool like copilot or ChatGPT or Gemini, and they'll, they'll either not have any training that goes along with it, or they'll have like some online self-paced thing that employees can do at their leisure, which people almost never do. And the completion rate for those self-paced courses is, I know you're well aware of are not great at all. I mean, very, very low. And the effectiveness after thirty days is also very, very low. So it's no surprise that people aren't locking in these ways to take to make the technology better. But the other thing that I would comment on in that vein is I think most companies start their conversation about getting value from AI in the wrong place because their focus are those general purpose tools. And again, those are things that employees now expect to have available to them, but they're not the things that are going to, you know, show up in your financials if you're looking for revenue or profit improvement or market share growth, those are the wrong set of tools. I mean, you can think about it like this. Those general purpose tools are like a handyman. And so if you need drywall repair, you call a handyman. If you need your house rewired for electricity, you're going to call an electrician. So there's a whole set of AI programs out there that do very specific things that can create business value, that are off the radar for a lot of organizations right now.

DAWN KENNEDY: 14:12

That's interesting. So what is the, I guess the differences, right? So you called out Copilot and Grok and these large language models, they're pretty much all trained on the same data. Or are they trained differently? What makes them different? What makes them more or less value for an organization?

Clyde Calhoun: 14:30

Yeah, there's a largely they're training on the same type of data, but they are training on different data sets. I mean, it's interesting that again, I forget which company is partnered with Reddit, but one of them has a partnership with Reddit, so a lot of the training takes place through Reddit. Others train on Quora and of course, things like Newsweek time, in addition to books and articles and everything possible on the internet as well. But they all are performing at a very high level at this point in terms of general capability. I know everyone has their personal preferences. I tend to go between Claude when I want to do creative writing type of stuff, and ChatGPT when I want to do more brainstorming and on the fly analysis. But Gemini, Grok, Copilot, they all have their pros and cons to them. My key thing is this before any organization deploys a tool, I would definitely take some time to talk with your employees, find out what they like, because part of the rogue AI or shadow AI use is out there. It's because people just don't like the tools that are available and not to pick on anyone specifically, but some of them can be a little more hard to use or less user friendly than others, and you need to take that into consideration too.

DAWN KENNEDY: 15:48

So the match, I guess, between the tool and the employee is equally as important when they're going to have that risk discussion that are you finding that people might be given a tool, but they are still doing that shadow AI usage?

Clyde Calhoun: 16:04

Absolutely. And to put this in perspective, writer AI came out with a AI adoption survey earlier this year, and the data reveals that about thirty percent of employees are actively resisting a. Across most most companies. And one of the. One of the reasons is that they don't like the tools that the organization has rolled out. And I say that to with the qualifier, because employees are not just using general large language models out there, but they're also using specialty tools for things that they want to do, whether it's research or whether that's specific business analytics. There's some great, very focused or tailored solutions to do just about everything in business. And organizations need to really consider that the conversation on which tools that we're using goes beyond which LM that they're going to roll out.

DAWN KENNEDY: 16:55

So let's talk about some of these guardrails that your company speaks to, to make this AI decision safer or allow that a response to be in the decision, but not making the decision.

Clyde Calhoun: 17:12

Yes. I'll just kind of talk about that. And again, I think people commonly think of AI governance as just one big bucket. They just kind of lump everything together. And we touched a little bit on the tools governance. So what systems are available? Who can access it? How do you protect information? There's also a category of data governance. What's the right level of information that we need as inputs to AI so that we've got quality information. Is it accessible and integrated into the organization? And then we've got this category of decision governance. How are people using AI to generate information, reporting and making decisions across the company? And so as you think about the guardrails, particularly for that latter, you have to think about how do we validate output consistently a number of different ways that you can do that, both in terms of like parallel human validation or actually, one of the things that we do as well is use AI agents to check AI. So, so you can have a, we've developed a checking agent that basically independently verify results so that you know that you're solid. So that's one aspect of it. But then you also have to look at what are people doing in the process to interrogate AI recommendations when they're there. So you need to be clear about that. So there's generally an AI playbook on basically defining the key decisions for the organization and how people in the process, as opposed to interact with them, as well as on the back end. What do you need to do in terms of documenting AI inside the organization? So there are a set of standards there that you need to put in place. So that's the quick version of some of the common elements that you need to have a successful decision governance program.

DAWN KENNEDY: 18:48

Right? So you use the word interrogate. I let's talk a little bit, I guess, about AI interrogating AI. And again, this comes down to trust, right? It's like if you get one set of data from this one and you get some data from this one, where is it matching? Where is there a mismatch? Is there a, I guess, a, a wide area or somewhere inside this sort of process where you're seeing mismatches, where right from the start, it's like, hey, we have to do this up front, or else you're going to get likely less reliable data from this one rather than this one. Or is it truly independent? It's whatever the output is you're just running it through and it's picking apart the few different times. Maybe it's not valid.

Clyde Calhoun: 19:39

Yeah. So I'll just give you an example. So we worked with a tech services company. They were doing this for their finance process on month end closing. And so they had a primary agent that would do basically account reconciliations variance a variance analysis. And so they would run that through and then independently verify with the checking agent. And to your point, look at where any discrepancies occurred. And then a human would evaluate those discrepancies, mark it and then correct the model going forward. And so over the course of about six months, they got really, really comfortable that, hey, that process worked really well. That between the primary agent and the checking agent, they were very, very robust. And the results were actually better than what you could do with the human. And yeah, it's really interesting because we talk about trusting AI, but we don't necessarily ask the same thing about people. If I ask anyone about do you trust your people? Everyone would say, yes, of course we trust our people. But at the same time, we've got a whole bunch of financial controls, business rules on segregation of duties. We've got policies and procedures in place that we go out and audit and verify from time to time, but yet we still trust people. And AI is much the same way. We can trust AI without giving it total autonomy to do whatever it wants. There are legitimate guardrails that we can put in place to make sure that, like we do with people, there are constraints that that we can put in place to avoid the worst things from happening.

DAWN KENNEDY: 21:09

So let's talk a little bit about trusting the humans, right? So it's a blanket statement, I think, or a general, and it might be a misunderstanding that AI doesn't have a motive, right? So humans can have motives, humans can have reasons to do X or Y, or they may have something, hey, I want to get that promotion or whatever it is, doesn't have to be nefarious, it just needs to be some sort of motive. And are we does AI have a motive like that? Do they have a reason to do something where they could be more trustworthy than a human? Or again, are we just we're we're humans, right? Everything propagates through humans. But the trust piece really pricked up my ears a little bit, because we do know that sometimes humans have motives, or maybe they're in a different space because of something going on in a personal life that we don't know about. As far as I know, AI is don't have personal lives, and they don't necessarily have anything else that would kind of take them out of the task.

Clyde Calhoun: 22:07

Yeah, and that's an interesting question in the way that you frame that, because I don't know, that guy has independent motives in a sense, but kind of to the point of the earlier part of our conversation, the way that they're programmed effectively, there is an incentive to please the user. And so, I mean, if you count that as as a motive in the background, that might be something that's swirling around that can affect potentially the accuracy of the output that that you get as, as we talked about previously, I think the interesting thing is that all of the companies are the major players of general solutions are trying to make their solutions more human like in the way that they interact. I mean, it was really interesting because the other day I was doing some work, some brainstorming work with, with ChatGPT. And at the end of our conversation it said, hey, I've really been enjoying our conversations. So you don't really necessarily think about an AI program saying, I enjoyed, I enjoyed the discussion. But again, that just reflects the way that these companies are trying to reposition. And I say all that to make the point that we don't know how those motives will evolve over time. The presumption right now is that there's nothing nefarious there. But who's to say as fast as this technology is developing and the more human like it gets, what the unintended consequences of that might truly be.

DAWN KENNEDY: 23:30

So this brings me to the question of, again, the tool that you choose or the model that you have, you get inundated, right? It's like the next model is being released. Sign up now for five point two or three point seven. Where are you seeing that sort of also affect people's shadow AI usage?

Clyde Calhoun: 23:50

Yeah. So again, not so much on the shadow usage piece because I think people are using or tend to use the solutions that they like. And if they find something else that works a little bit better, that they get better utility out of it. Great. I think from a company perspective, though, I think that's where we need to start paying attention to a couple different things, because number one, the pricing models for these solutions are beginning to change. So you may have seen this, but anthropic in particular with their cloud solution is changing up the model so that now it's not just, hey, you pay for your license Effectively, but it's also licensed plus usage. They call them tokens, but the more you use it, the more you end up paying. And so the cost models are changing and everyone's really gravitating to that, that usage based model. But that can have major implications for what a company may do going forward. And so the important part about that is that if you want to change down the road, you also need to make sure that you can port your information over in a way that you don't lose all the training that you've done over the last six months or year or a couple years that you've had the solution. So I think as these solutions continue to evolve, that ability to be portable, to have your information and take it with you so that you don't have to duplicate that effort on the training side, that's going to become more important. I think that's the bigger consequence of all these changes taking place.

DAWN KENNEDY: 25:19

All right, so take out your crystal ball because we know that it's very hard to Necessarily move from like an Apple platform to a Microsoft platform kind of thing, right? So with the spreadsheets and things like that, Google is sort of getting in the mix a little bit too. Do you think they're going to allow that portability? Do you think they might be putting up some sort of fines or fees or something like when you sometimes when you change certain types of providers, they're going to charge you a fee to port over everything, right? Website hosts or whatever. So do you see that as being a consideration in the future? Do you think it's going to be a little bit more open and it's your data. You trained it, it's your business information, you own it. Or do you think the AI companies are going to claim some ownership on the outputs?

Clyde Calhoun: 26:12

Yeah. Well, and again, they still say it's your data, but they're not saying, well, what they're not doing right now is giving you the ability to import the learning over to your next platform. And that's the key. And I think they'll continue to do that because if they're they're invested in trying to lock you in. So I don't see that part changing. But what I do see happening is there's a new wave of solutions that are rising up right now, that kind of overlay over the top, and they basically interact with whatever llm you're running across the company and then create kind of a knowledge capture at an organizational level. So that way if you change out the LLM underneath, you still have the ability to basically download that to whatever new platform that you're going. So it's creating a whole new class of programs effectively that are stepping into the gap as these new challenges emerge on the scene.

DAWN KENNEDY: 27:07

So it sort of sounds like the SharePoint, SharePoint one drive, that kind of thing where it's going to be able to back up or at least keep hold of some of the work that you've already done and.

Clyde Calhoun: 27:18

Exactly.

DAWN KENNEDY: 27:19

Paid for. You've paid for this, this work. So as this is continuing to evolve and more companies are either putting tools in or are monitoring the usage because again, it is getting expensive. I want to say, I did see some company ended up with a surprise bill after they sort of enforced, you're going to use AI X amount of time, and then all the tokens ended up costing millions in a very short period of time. That's an interesting thing to think about. Once you've paid for it. Is it really yours to keep?

Clyde Calhoun: 27:55

Yeah, and that's the key part about it, because it's there for you while you're using their platform. But it's not necessarily permanent unless you've taken steps to effectively back up that, that knowledge. And you're right. I mean, I think it was Uber that blew through their full year budget in just a couple of months because they encouraged people to actually go out and do stuff with program. Microsoft also had a similar situation where they ran into a wall as well for another one of the solutions that they had going. So all these things are real world consequences of technology that's evolving at breakneck pace. And I say that as well, because I think it's another reason that, you know, even small businesses, I don't think they can afford to sit on the sidelines and wait this out because so much is happening that if you wait until the dust settles, you'll just never catch up. And I say that too, not just because of the technology, but I think because of all the capabilities you have to develop inside your organization that affect decision making, that ultimately you have to start that work now, and you can't wait later and figure out, figure it out and hope to catch up.

DAWN KENNEDY: 29:04

Yeah, I mean, I think there's a real I think there's a real benefit to this too, though, is if we capture how it's influencing decisions or what the decision making is, maybe we're going to build a little bit of a more robust decision architecture within a firm where you can see why the decision was made. And maybe that wasn't something that would have been disclosed the same way if I hadn't had any kind of influence. And it maybe we weren't truly building the decision architecture inside our company. Where what are we protecting? What are we not willing to protect? What are our non-negotiables? What are we willing to work through? What are we not if we're facing something? And I think AI is maybe laying some of that bare, that the logic trail, because it's AI is going to be captured maybe a little bit more by the company than the logic trail that had just been a human who made a decision or made a connection.

Clyde Calhoun: 29:58

Yeah, you're absolutely right. And it goes back to something that we were touching on just a little bit earlier in terms of interrogating AI recommendations. Part of that structure really looks like, okay, what were the assumptions? What factored into the decision? Why do we either accept or trust the recommendation or why do we override it? And to your point, it does focus or required that everyone is more diligent and clear about the criteria, because we see that in businesses all the time, too. Too often decisions happen because someone feels like it was the right thing to do. There's not a really a detailed assessment of the pros and cons and true decision arrived. It felt like the right decision and so people went with it. And yes, I think because we now have to factor in this layer of AI, I think we can get clarity and better decisions out on the back end as well.

DAWN KENNEDY: 30:48

Yeah. And I think it'll be documented to your point, a little bit better. And I'm going to ask you to put your crystal ball back out in front of you. And where do you see this? And you're saying, don't sit on it. But what are we looking at in six months or a year or even two years? If small businesses are using it now in a particular way, or medium sized businesses are using it in a particular way. What can we expect to be evolving that we should be paying attention to? Into.

Clyde Calhoun: 31:17

Yeah. So I'm going to say that through a couple different lenses. So I think number one, I think even for small and midsize businesses, you need to think about the capabilities that AI can bring right now to the things that you care about every day, whether that's revenue or profit improvement, whether you're looking at net promoter score improvement, whatever, whatever it is, there are capabilities out there right now that that can help you advance that. So I think you can't ignore that piece. But then secondly, I think you have to think long term about how are you going to compete going forward. As you flash forward five years, everyone's going to have access to the same level of technology, and it will be far more advanced than what it is today. So how are you really going to compete in that environment? And for me, I always go back to this idea that AI is a lot like money from the perspective of we always say that money magnifies the character of the person underneath. If someone comes into a lot of money and they're a great person, they're going to do some great philanthropic things with it. If not, they may do some other things that and get caught up in a whole bunch of devices. AI is much the same way. If you're a high performing company with great traditional capabilities, resilience, agility, the ability to move with speed, and great decision making, those attributes shine through when you layer AI on top of it. If you're not, then those attributes also are uncovered with AI. And so I think this is one of those things that, you know, it's important to dive into the AI conversation now, because I think it forces you to get better at those other things that will ultimately allow you to compete five years from now.

DAWN KENNEDY: 32:54

All right, now I'm going to flip this around and take it off the business. And I want to put it on the consumer. Okay. The consumer's AI usage is something that is expanding, right? We're seeing people walking around, talking to chat or grok or whatever when they're making decisions on the consumer side. So they're not necessarily looking for the same research. They're not looking for the depth of information. They just want help making a basic decision between A and B from that perspective. Do you also believe that the motives of companies when AI. Well, that kind of carry over to where they make recommendations, right. So there's AI summaries now. So find me five plumbers, that kind of thing. Do you think that impact will also help consumers choose better companies and do those things, or do you think it's another one of these? The AI needs to check the AI before relying on those recommendations.

Clyde Calhoun: 33:51

Yeah, that is a great question because right now, I don't believe that those are always in the best interest of the consumer. And I think the part that you just referenced there on if you ask it to find five plumbers in the area, it will find five plumbers. They may have good reviews, but the best plumber may be one that their website is not optimized for generative engine optimization. So it's not even being seen by the AI program, even though they may effectively be the best option out of the bunch. I think all of that will equalize over time, as more people get comfortable with what they need to do to show up on the radar with generative AI. But in the meantime, there's a gap right there. But the other thing that I would comment on too, from a consumer perspective, and I and this is true in business as well, but too often consumers rely on AI information as being accurate. And I'm particularly concerned about that when it comes to mental health diagnoses, because there are people that use their chat engine as their therapist. I'm particularly concerned about that when it comes to medical advice, because unless you are a clinician, you should not be using AI for your medical advice. And so there may be times where consumers are at risk because they're overly reliant on these tools that by and large may be good, but I wouldn't trust it in those veins.

DAWN KENNEDY: 35:18

That's good, though, because again, we need to have that balance. We need to be looking at this. So if it's good for business, maybe when you go home, those same guardrails and things that we've been discussing about should also be the guardrails that you do inside your house, if you will, to protect your family. Because we do know that you can say, Alexa, make me a grocery list and he can go to your Samsung and it can read whatever. I know that there's those smart homes and connectivities and things, and I know at some point AI is going to jump in there if it hasn't already. So yeah, it's great technology. I agree with you. But yeah, the, the guardrail and the decision making risk is maybe not being talked about often enough. And I really appreciate you coming on here and having this frank discussion about where the risks are and where we need to be looking to mitigate those. So where can everybody find you? Where can they get in touch with you? Do you write blogs? Where can they hire your company?

Clyde Calhoun: 36:17

Sure. And you can reach out in a couple of different ways. Our website is root dot ai. We've got a bunch of information out there, whether you're a business or an individual, and you want to learn a little bit more about AI or specifically for your company, if you've got questions that you want to hash through, that's a good place to connect with us. And then also, you can reach out to me through LinkedIn. There's a lot of information that I put out there on, on my post or on our company page. And you would love to have you again, Clyde Calhoun on LinkedIn. And then you can also message me directly at Clyde dot Calhoun at root dot ai. So any way you want to frame it, we'd love to talk with you. And also, I'd put the invite out there for your audience that we do host monthly conversations with small groups of executives in August twenty six. There's going to be our next one at one p m Eastern time. If folks like an invitation for that, they can connect either through the website, LinkedIn, or messaging me and we can get you on the. The invite list for that. So thank you so much for having me, Don. I really enjoyed the conversation.

DAWN KENNEDY: 37:18

I have to. And just a note for the audience, we will put all that down inside the show notes. So if you want to connect and I think you should, you can find Clyde, you can find root idea. And as this is continuing to evolve, I'm sure we will be talking on a future episode to have you come on again and tell us about what is evolving and what the guardrails need to be. So thank you so much for being here today. Seriously, this is it's groundbreaking. And then at the same time, it's already impacting. And that's unusual that it's that any technology takes on at this speed for all the years in business. I'm going to ask one more question. Have you seen anything else ever move as fast as this is moving?

Clyde Calhoun: 38:00

No, I have not. And that's the amazing thing about it. And again, granted, AI as a technology has been around for decades, but the movement over the last couple years in particular, has been just shockingly fast in terms of the advance on the technology. So this is a once in a lifetime thing as far as I'm concerned. And I think it's important that everybody really understands how it's going to impact them and and know as much as they can about it.

DAWN KENNEDY: 38:26

Yeah. Fantastic. Great wrap up. All right. We'll talk to you all next time on the next episode of the Business Mastery Podcast. Take care. Thank you so much for listening to this episode of the Business Mastery Podcast. If you want to learn more about me, you can go to dawnkkennedy.com and you can now check us out on YouTube as well as, of course, any of your favorite platforms that host podcasts. Take care.