How to Use Artificial Intelligence WIthout Losing Trust

AI has become the new business buffet. Everybody's walking past the table with a plate in their hand saying, I'll take one chatbot. I'll take 2 automations, a scoop of predictive analytics, and whatever that thing is that promises to replace Linda in accounting.

Sorry, Linda.

And I get it. The pressure is real. Your competitors are talking about AI. Your vendors are talking about AI. Your cousin who just discovered ChatGPT is now calling himself a prompt architect, which apparently is a thing now because LinkedIn looked at humility and said, no, no, no thank you. But here's the problem. Many businesses are not adopting AI because they have a clear problem to solve. They are adopting AI because they're scared. They're scared of falling behind, scared of missing out, scared of looking outdated, and scared that if they don't jump on the AI train, they'll be left standing at the station holding a fax machine and a Blockbuster card. And that's what we're talking about today on Marc Haine Live. Welcome to this episode of Marc Haine Live. Of course, this is where small business owners and entrepreneurs pick up some core skill sets to help them create the show-stopping, jaw-dropping experience that your customers and your employees deserve. I am your host, Customer Employee Experience Strategist Marc Haine. And I'm so glad that you're here with me today because my guest is Dr. Alex Leithardt, the Human First AI Advisor. And today, we are going to be exploring how leaders can adopt AI safely, responsibly, and with purpose. purpose without letting fear, hype, or a very confident software salesperson drive the bus for you. The price of admission today is please make sure that you subscribe to the podcast. And please make sure that you share this episode with your circle of friends. I'd love to see the business community really supporting the work that I've been doing all these years. Gosh, since March of 2020. Can you believe it? So, here's the thing. AI is not quietly knocking on the door of business anymore. It has kicked the door wide open, walked into the boardroom, plugged itself into the projector, and started using phrases like disruptive transformation before anybody has had their coffee. And for business owners and managers, that creates a very real problem because you are being told 2 things at the same time. One group is saying, adopt AI now or get left behind. And the other group is saying, be very, very careful. AI is so dangerous. Helpful, right? That's like telling somebody, run quickly through the dark room but also watch out for the snakes. So, leaders are trying their very best to make big decisions with incomplete information. Should we buy the tool? Should we build the tool? Should we ban the tool? Should we train the team? Wait for regulations? Maybe we can start with marketing or operations. Or maybe we could start with the break room toaster. And that brings us to our Question of the Day. So, where in your business are you currently using, testing, or considering AI? And do you know what business problem it is supposed to solve? I'd love to know what your experience is. Go ahead and share a link to this episode on your favorite social media platform. Hashtag it #ExperienceLeadership. Put your comments down. I'd love for you to be part of this conversation. As I mentioned, today's guest is Dr. Alex Lethard, also known as the Human First AI Advisor. And I love that title because in a world where everyone seems determined to turn business into sci-fi movie with quarterly targets, Alex brings us back to the one thing AI is supposed to serve and that's people. Alex has had what he calls a wiggly career, which is one of my favorite professional descriptions I've ever heard because it sounds honest, British, and slightly like a garden hose under pressure. He started in civil engineering, moved through academia, earned a PhD in engineering focused on climate change impacts, and built deep experience in data science. machine learning, and AI. From there, he went into consulting with big 4 firms, boutique firms, industry, and now independent consulting. Over the past 15+ years, Alex has worked on data and AI problems with governments, banks, asset managers, prop tech companies, market intelligence firms, financial services, and climate tech businesses across 5 continents. He's very young looking for 78 years old. And when he's not helping organizations make smarter AI decisions, Alex can be found playing the trombone exploring Sussex countryside, or spending time with his family, which frankly sounds much healthier than arguing with a chatbot at midnight. Alex, it's so great to have you on the show. Welcome!

Thank you, Marc. What a wonderful intro as well. Yeah, I'm delighted to be here. Let's get into some deep conversations about AI and business.

Yeah, so just before we get kickstarted, tell me a little bit about who your clients are and how you're serving them right now.

So my clients can vary quite widely. The typical point of entry that people have for working with someone like me is they've been using AI and data platforms for a while, or they're curious and they don't know where to start, but they're not seeing the value that they expected materialize. And so they will reach out to someone like me who's independent and they'll say, can you come in? Can you lift up the bonnet? Can you kick the tires? Can you help us understand why things aren't performing the way that we expected them to perform. And their assumptions can be it's the tech or it's the team. But I'm much more holistic and think, well, look, let's look broadly across the organization. Let's unpack your governance. Let's unpack why you're doing this in the first place. And let's see if everything lines up and makes sense.

Yeah, it's interesting because, you know, I do a lot of presentations to businesses and to groups. I do like an AI 101. program to introduce people to help demystify AI a little bit because organizations know that their people are kind of toying with it.

Yes.

But they don't have any kind of internal policies. They don't have any real understanding of what it can do for them. And so, people are basically using ChatGPT as their email engine and, hey, make a name, and they send it off. And they're kind of underusing the tool because they don't really understand it. But at the same time, there seems to be a big fear around how much adoption should we actually be doing in an organization?

Yeah, it's such an interesting phenomenon, isn't it, that there's all this consumption and use that almost goes under the radar. So what people call shadow AR use, where you're using it on your phone and you're a bit afraid to tell your colleagues, this is how I'm using it. There's concern about personal information or private information, confidential information going in. So we're a bit cagey about that. And then you get this dynamic pull and push between parts of the business. The commercial side of the business is always going, let's go faster, let's sell more, let's engage with this technology and see what advantage we can get from it. And then you've got the IT who's a bit more curmudgeonly, maybe the head of data, a bit more sort of slow, a bit more risk-averse. And they're saying, well, hold on, I understand a bit about how this technology works, I'm worried that we're going to be breaking some rules here by deploying this. Let's just go a bit more slowly and take a back seat. And what I often tend to find is when you get deep into the weeds of the people who want to go really fast, and you start to take the toys out of the box and go, well, why do we want to go fast? What problem do you think we're solving by doing that? What advantage do you think you're going to get? We don't know. We just thought it was really good and we can go faster with it.

Yeah. And it's interesting. It's like, it's kind of like that Alice in Wonderland thing about, you know, where do I have to go? Where do you want to go? I have no idea. Then it doesn't matter which route you take, right? So, when you look at how organizations are adopting AI right now, what worries you the most?

Oh gosh, so many things. So, I get quite concerned about the amount of risk that they're exposing themselves to. in terms of things like which tools they're adopting, whether or not they're hedging their bets across multiple vendors, or they're really going deep on, say, Anthropic. I worry about geopolitical risk. So we've seen some really interesting behaviors in the last 12 months, but certainly in the last month, where access to tools has been restricted by government. And so if you're building your entire business stack around Anthropic Claude and the government turns off access to that product for you, what's your backup plan? So I see businesses pick up these tools and run with them, but they almost don't put in place, or they very often don't put in place at all, the real business scaffolding that goes underneath any process, you know, standard operating procedures, training your staff. And so I ask people, well, if this tool went away tomorrow, what would you do? How would you continue operating? And very often there is just no answer to that, especially for sort of micro businesses and startups where they're saying, well, our entire business runs on AI. We don't employ any staff who aren't delivery. And that to me is a huge risk, I think.

You know, it's interesting because through kind of social platforms and stuff, there's been talk for years now. Like I would say, a good decade about people like myself who kind of house my videos and stuff, let's say on YouTube.

Mm-hmm.

And I don't own the platform. And if YouTube decides to take down my stuff or do whatever, I have no control over it. And we're kind of, you're kind of saying the same thing about all these different kinds of AI engines that at some point, the government could step in now and change regulations about what can be done with it, what can't be done to it. Anthropic just recently got their got one of their models yanked from the marketplace because the U.S. government got really nervous about it. And so, to that point is this idea of putting all our eggs into this one basket. When we think about it, is this something that will stop businesses, you think, from adopting AI and going, well, since we don't control the platform, we better not start playing with this?

Yeah, I think there's So one of the issues that I think I've identified and other people have seen, and it's quite acute in the UK, I think it's much less of a problem in the US, but is this slow adoption, this reticence to pick up this technology and run with it for all of these reasons, mostly because people don't trust it.

Yes.

You know, there was a recent survey in the UK where, you know, the main reason why people aren't adopting the technology at all is lack of trust and transparency.

Right.

And the question then is, well, if there isn't demand, then there's no reason to build the data centers. There's no reason to build the GPUs that power the models. There's no reason to build the models at all. And so the grand promise of AI can solve any problem disappears very quickly under that type of doubt. And so I think in the UK, if you go and talk to companies that our data center manufacturers, you know, builders, constructors, they're saying, well, we're not putting spades in the ground yet because we don't think there's a strong enough demand signal.

Interesting.

And that's, yeah, which is really, you know, it's also challenging because you sort of go, well, if your data, if someone could guarantee that your data were resident in the UK and that your data would be processed in the UK, would you feel different? And people do feel different under that scenario. But we have another couple of challenges here, which is even if you want to geo-restrict data within an existing data center, we don't have enough compute capacity in the existing data centers to run the latest models.

Right.

So there's this, you know, there's this real kind of challenge. And to that, I often say to people, well, do you even need the latest model? Based on the problem you're trying to solve, you're writing some emails or you're analyzing patterns in data. All of these tools are good at that. Yes, yes, yes.

So, for the people like for business owners and managers tuning in right now who's looking at AI, they don't really— they've heard it. They're hearing it from like vendors and from consultants and even from a LinkedIn prophet with a ring light. What do these business owners need to understand first before they start thinking about this idea of implementation?

So, I like to start with people by saying AI isn't one technology. It's not language models, for example, or image generation. It's a whole family of different solutions to problems. And some of these technologies have existed for a very long time. It's just that now we can do them faster, better, more accurately. But really, it's about figuring out where are the pain points in your business? What are the bottlenecks? What is the value that you'd like to create that you don't have a process or a person for right now? You know, what's holding you back from doing that? It's, I know it probably sounds like the absolute classic consultant's battery of questions, but I think they're still valid in this context. It's just that everyone's getting a bit excited and going, well, I can use AI to solve any problem.

Yeah. And any problem is not a definitive problem. You're not defining anything. So, it's really interesting because to your point, I see some people look at AI kind of like— I don't know if you ever watched the first episode of Downton Abbey when Mary goes to see Granny.

I've never seen Downton Abbey.

So, in the first episode, Mary goes to see Granny and walks into Granny's house. And Granny's got candles everywhere. And she's like, but Granny, we installed electricity. And Granny looks at them and goes, oh, I don't know about this electricity. Maybe I don't want all this fire in my walls. And so, Granny had this perspective or perception of what light means to her. And she's thinking that in order for electricity to work, there's going to be fire inside the walls, right? Yeah. And so, I think we fall into the same category when it comes to AI. It's like, it's all of a sudden, it's going to become the magic key that all of a sudden, we're going to become more productive. We're going to become We're going to become less human dependent because let's face it, every business is under-resourced. People are trying to do twice as much as they did 10 years ago. So, what are the warning signs that a business is adopting AI kind of jump with this whole kind of jumping in without a clear plan? What are things that you're seeing?

Gosh! So, I absolutely see people set up innovation teams. And they say, well, we've got a special innovation team of 5, 6 people and they're all very expensive and all working on AI. And you go, well, why? What problem are they solving? Why aren't those people, the other people who are in the business who are actually doing the work that you're then upskilling to use these tools, who can then apply them to the real problems that they're facing every day? So why are we trying to kind of make this special team out of nothing?

It's interesting because is it a fear that, you know, oh well, you know, we're doing some, we got this special team working with AI to help us with our accountants. And because of that, we're going to end up losing accountants on our team. We better not let the accountants do the work because they're going to be in it to keep their jobs. Could it be something like— could it be rooted in fear like that?

I have heard people talk about that. I think the issue is more your point around businesses are massively under pressure. Everybody is stacked out with work. We are under-resourced left, right, and center. Nobody wants additional work on their plate. Nobody wants to be accountable for anything else. So if you start saying to people, now you've got to have a successful pilot and research project in order to get funding, it's just another problem for people to deal with, which I do think is the wrong approach anyway. If you start saying to people, you know, let's help you develop these skills, let's see how you figure out the right way to use them. What we'll do as the sort of central business overlord is we'll help build an environment that is safe and constructed for you to do that in.

Mm-hmm.

And I see real friction there happening where, I'm going to sound like an awful colleague, but where the IT team just sort of goes, you've got Copilot now, off you go. Please don't put any personal data in it or confidential data. But they don't put any of the systems in place that could exist to prevent that information from egressing from the business.

Right.

Whereas what I like to see and what I like to kind of instruct people around is let's put data loss prevention at the heart of this program because A, it's good practice anyway in your email chains and your Slack channels and all of that, but it's great for your AI models because then your colleagues will feel comfortable. They're not going to make a huge mistake. They're not going to accidentally sell out the whole business's competitive advantage in an email. Because we're going to help them plug that hole in the dike. And that just, it gives people the freedom and the confidence to try things out safely.

It really is about putting systems in place and creating a culture of intentionality.

Yes, absolutely.

No different than, you know, I was talking to a cybersecurity expert and we were talking about this whole thing about changing passwords and all this sort of stuff. And that the IT person will send out a mass email to everybody and go, okay, it's 1st of July. Time to change your passwords.

Right?

And any Any hacker who figures out that that's their routine all of a sudden goes in there and have people redirect to a different website. And all of a sudden now, they're stealing copious amounts of corporate information. And so, we have this belief system now. We're training people to be more intentional with their emails and with what they're clicking and how they navigate past our kind of internal systems. And so, I mean really, that intentionality is just as valid when we're looking at developing AI and creating policies around AI and playing with AI.

And I don't think that's necessarily changed. But what I think we're exposing now because of these tools, because of the way they work, is we're exposing the sort of lack of development of EQ in the workforce. You know, we haven't developed people's coaching abilities or their communication abilities. And when they go and they try to use one of these tools, they're unable to give it the kind of explicit direction that it needs in order to create a quality response. So to your point around prompt engineering before, that's really all that is.

Mm-hmm.

You know, you've gotta coach it because it's gonna come back to you with a result. Maybe that result is what you want, maybe it's not. But just like a person, you need to be able to communicate very clearly I like this section. I need more work on this. And here's why. Here's some other evidence you can bring in. This is the kind of language I would like to use. This is the impact I would like to have on the customer.

And it's interesting because to that point, I think a lot of people start tapping into it. And like, I show my 91-year-old mom how like we— I'll do things. She's like, oh my God, it's like magic. It's like, where does— how does it know? Like, where does this information come from? And it seems like it's magical and it's creating all this stuff. for us. But to your point, if we don't have a system in there to test against it, to be able to push back against it, to look at assumptions. Because I think all too many, like we've seen tons and tons of stories of lawyers doing court briefs and all that and handing it off to the judge. And the judge is like, what are you talking about? There's no such, right? And But people get so— seem to— the people who are using it seem to get really, really comfortable with it.

Yes.

And yet, there's no check and balance on making sure that we can push back because, you know, ChatGPT loves you.

Oh, they all love me. They're all very positive about what I say to them, which is lovely. Like, but I mean, there is an interesting core truth in that around validation. So from a human communication perspective, we love to be validated. We love to feel that someone else is interested in what we're saying.

Yeah.

And I think, you know, there's an interesting signal that these tools are designed to do that, first of all, and it's outside your control how and to what extent they do that. But also, are we all so starved of validation in our day-to-day lives that, you know, a word salami machine produces a random string of words that sounds nice to you and you feel good about yourself? I think there are questions in that for society as well. But I think, you know, to your point as well, almost no businesses that I've seen and worked with are putting in place what I would consider to be robust testing of these systems. So when they test, so I've done a lot of projects in data integration and things like that over the years. We have this idea of there being a sort of one golden successful thread of doing things right. and everybody designs for all the perfect things to happen. And so your data engineering takes maybe a week or 2 to implement. But life isn't like that. There are all manner of things that can go wrong. And to take your perfect thread, everything works correctly engineering problem and turn it into a robust piece of software engineering, that 2 weeks becomes a year. Yeah. And so nobody wants to make that investment. And the same thing happens with things like AI agents. You know, there's one way for it to go right. There's an almost infinite number of ways for it to go wrong. And if you want to build a system that is robust to that, you have to be able to articulate not only what you want to happen, but also as much of what you don't want to happen as possible, and then continuously test for it. Because not only will these things happen when you're watching, they will also happen when you're not watching. And if you put into place some sort of automated system that's replying to phone calls or replying to emails, you've got to be watching the whole time and testing the whole time to make sure that the behavior that you want doesn't drift away into something that you don't want.

Yeah, my worst fear, I think, with AI agents doing my emails is at some point it's going to represent me as like an African prince who's locked up in prison who wants to get—

I mean, I am very strict with access to my data with AI agents that I write. So I don't necessarily use a platform. I don't use sort of Microsoft Copilot Studio or anything like that for myself. I do a lot of coding. And so I write them all myself. I give them very specific guardrails, not only in the prompts, but also if I'm using LLMs, the prompts, but also I deliberately do not enable it to do certain things. It cannot leave its box. And when I walk away from it, it doesn't exist anymore. And I think that's a reasonable approach for a small business.

So, when we come back, I'd love to dig into how organizations can adopt AI safely with a human-first intent. And we're going to get to that right after this.

When the spotlight shines on your business, are customers applauding or yawning? In other words, how is your business performing? Make your business a star with the new book Lights! Camera! Action!, Business Operational Excellence Through the Lens of Live Theater by Marc Haine. Marc uses his business and acting experience to help you see your business like a live show so you can create a performance your customers will never forget. Buy Lights! Camera! Action! Get your copy of Camera Action today at your favorite online retailer or directly at marchaine.com.

Welcome back! I am speaking with the Human First AI Advisor. And that's important, this idea of human first. Alex Leithardt, you described AI adoption as a hypothesis to test, not a yes or no decision. What does that mean like in practical terms?

It means thinking, you know, we talked about intentionality before. It means thinking about what actually is the reason why we are doing this. So, if we believe that we can be more efficient with sending our emails because we're going to get a better relationship with our customers, we're going to have better outcomes from customers, then we need to actually put in place a system where we can measure that intentionally. And so when we implement our AI system, we've baselined what we believe the behavior currently is. We make a change and then we see whether or not that actually has an impact on the outcome. And it may sound like, so people always say to me, well, that sounds like being at school. That sounds like science at school. And I say, yes, that's exactly what it is. Because when we follow the scientific method, which we should all be doing, we probably all do in little ways in our lives all the time. We need to think about what is the theory that we're going to test? What is our evidence? What do we think the current state is? What are we going to experiment with and change? And then, we measure the outcome to see if what we thought would happen actually happened.

I love that. I think that's such a good— I think in terms of even use cases when people are looking at and saying, I'm not sure what AI can do for me but here's a bunch of problems that I have. I look at AI and go, gosh, I wish I had it 30 years ago when I was doing staff scheduling for 400 employees, right? How much easier would it have been to put that puzzle together every week? And so, I look at small businesses especially who have this under-resourced. But they also have kind of this nativity because the people, I like to call them incidental entrepreneurs because they have a subject matter expertise that helped them start their business and grow their business. But they don't have the marketing background, the HR background, the technological background to be able to look around and say, hey, this would be really good. And so, I think they become more subject to kind of these tonic salespeople who are coming into their business and go, hey, how would you like to speed up your processes? We can do it, right? And then, they just kind of give their systems over to something. So, what does a really good AI use case look like to you based on your experience?

That's a great question.

Oh, it's a little question too. I know it's just a small little question.

I mean, I can tell you. I'll start with one from my own experience and we'll go from there. So, I am dreadful at emails, frankly. I get a lot of emails. I'm very busy. My attention is constantly being dragged all over the place. I think a lot of, you know, solopreneurs feel that. And so a great use case for me personally was to implement agents that look at my inbox on a periodic basis, every few minutes. They are, they have a certain level of access within my email system. They cannot send anything, but they can move things around within a set of directories that have already been created. And so they're trained to recognize different emails of different levels of criticality from different people as they come in. And what's useful about this is, yes, we've been able to do those things with, you know, Outlook rules and such and so forth for the last few years, but those rules are surprisingly rigid. And what's great about things like language models is you can get them to recognize all sorts of patterns that are completely outside something that you could define within a rule set in Outlook. So there's some great behaviors there that I think small businesses can really benefit from. I also find things like data quality can be assisted by these sorts of tools very easily. So if you're working with a very large dataset or a large enough dataset that is full of holes, is full of poor quality points, maybe you need to do some classifications on some of those data. And you can run a model over those very quickly to be able to assign labels to each record within that dataset. And then a human can go and review them. And very quickly, you can accelerate that process of cleaning up some data, right? Pretty straightforward use case. Other ones that I think I've come across recently, let's have a look. I always find people talk about coding a lot. They talk about writing code very quickly with these types of models. I've done a fair bit of that, and I work with lots of people who dabble in it, and they try and use it, and they're sort of experimenting. They're saying, well, I'm not actually sure how much this has really accelerated what we're doing. And to our point around tests and guardrails, right, they were good. They were actually measuring how much effort tasks were actually taking, so you could see some evidence. But what worked really well for them was starting to write the tests for the code themselves, and then setting off the AI agent to work against those tests. And so essentially, you know, it's a little bit like, if you've done things like staff scheduling, you might have come across things like linear programming and optimization and and things like that. But it's really just kind of throwing random solutions at the wall, seeing what fits, seeing how close it gets to optimizing towards the solution that you want. And you sort of generally wiggle towards the outcome that you're looking for.

It really is like I was showing my wife, I'm doing some renovations at home and I used ChatGPT to give me some pictures of what the renovations could look like and all that, right? And some of the things was, it came up with really ridiculous looking things. And my wife was like, oh, I thought AI was so much smarter. And I said, no, AI is actually stupid because we have to tell it what we want. And then, we have to prompt. And we literally have to look at the evolution of the prompts, right? You can't just turn around and just say, okay, design me a nice foyer.

Right.

We have to really kind of start setting context and we have to learn how to be able to do that. And so, what steps should organizations be doing when it comes down to this idea? And by the way, I like the fact that you had this human in the loop component that you have humans coming back to it and pushing. I also like the fact that you said you go to and have the agent test against itself. So, this idea of using AI to be the ultimate contrarian to rip apart any ideas that you've brainstormed with AI. But when it comes down to this idea of I want to— I have a problem I need solved. I want to speed up my thing because I don't have enough people to be to do that. How can I move this forward in a way that's going to be conducive to the business? It'll be safe. I'll have safeguarded my details and it'll become more productive rather than kind of this time suck of going down this rabbit hole.

And I think that feeds back into your question of what makes a really great AI solution.

Yeah.

Because those things that I was talking about, they all solve a real problem. that an actual person has within their organization, whether it's me with my email headache, whether it's a bunch of guys in a team trying to deliver features for a program very quickly, they all make good commercial sense. So when you look at the numbers of what these businesses are actually trying to make a profit from, you know, I'm not getting, I'm not really generating value from organizing my emails, but I am getting value from being able to respond to customer emails very quickly. So it's helping prepare me to do that. With the software developers, it's helping put them in a position where they can deliver features more quickly, where they're in control of the tests, but they're not having to do the donkey work of writing lots of boilerplate code. And all of these things are engineered in a safe and responsible way. So they are not necessarily taking high-value work away from people, They're taking the kind of the repetitive kind of grinding work. So with the software engineers, like software engineers, some people are happy sitting there cutting code, but mostly they want to be solving problems. Like everybody wants to be, I'm going to use a '90s TV reference, but everybody wants to be like Data on the Enterprise, right? Star Trek reference for the audience. You want to go to the computer and say, Computer, write me a computer program to model, you know, the progress of this satellite around the star. Nobody wants to sit there and carve out all the different algorithms that are required to make that happen.

Right.

And I think that's good promise with tools like this, because you can start to see how they're going to enable people to have behaviors like that. So I can go and say to it, ah, you know what, I want a graph of I don't know, regulatory risk from this dataset against revenue and price-to-earnings ratios for these different businesses. And I need it in my company colors and I need it in this format to go on LinkedIn. And with enough contextual information, I don't have to write the code to generate that image. The computer can do that for me in sort of 4 or 5 minutes.

Fantastic.

Where it gets— sorry, go on, Marc.

No, no, no, please, where it gets—

I was going to say where it can get really tricky is when organizations don't have that kind of context in mind and they're not doing that kind of work. And so people try and replace what I think is the kind of really important human connective tissue in organizations because we start saying, well, all these emails that we're sending between each other, exchanging information between departments, having meetings, all of those things, But we could just get rid of all of that because you guys can just keep doing your optimized piece of work in your silo and you can do your optimized piece of work in your silo and you never need to talk to each other because the computer is going to telemeter everything and feed it all up. And we don't need any managers in between. We'll just send it all up to the C-suite and they can make decisions. And I actually think that's really problematic.

But isn't like that line of thinking is super premature at this point? Like I understand that in the United States in the healthcare field, That they were trying to test AI in being able to connect kind of reporting from the labs and from the radiology departments and stuff back down to the doctors who need them. But the problem happens was as much as you can get AI to generate stuff like that, they didn't have the infrastructure to be able to move that data forward without it creating more work and people having to log on to new systems and do different things. So, I mean, it reminds me of when we first got computers. That's how old I am, by the way. And I remember talking to my partner and he's like, oh, we should get a computer because then it could do all this work. And he had a picture in his mind of what would happen because of computers. And it's like, but you have to understand how to do the work first before you can delegate it. And that I would say that that's the nice thing with AI though that I found is if you mentioned a whole bunch of stuff and in my mind, I'm thinking, oh, most small business owners would have just heard, Yada, yada, yada, yada, yada, yada. I just want to know how do I make sure I have enough money in the bank at the end of the month, right? And it's interesting because unlike the programming where you had to understand the process before you could program it, with AI, you could actually go to AI and say, I have a problem. Here's my problem. And you can use it to help brainstorm solutions. So before you ever go into coding or actual problem solving, Yeah. you could use AI to help get information on stuff that you might not know about, things you don't understand about your business and so on, which I think is fascinating.

Yes. And I do think that's a really good use case for some of these tools. And I would say it's not replaced, but it's certainly got a role for me in my workflows when I'm sort of just starting to research something because it's a heck of a lot easier for me to articulate a question in natural language.

Yes.

Than it is for me to go kind of hunting around Wikipedia or somewhere or searching on the web. And I think that's part of the kind of success of these tools is it's just so easy to communicate with them. And you can start to refine what you're doing as you go. So I often go to them and say, well, I'm looking for a term to express this type of behavior, or I've got this type of business challenge and I want to come up with a solution XYZ, here are the parameters. Is there something that already exists? Is there a piece of research that already exists or a name for this in the field? And then once I've got that name, I can go to any number of sources and start pulling information that I probably have a higher level of trust in.

Right.

It's interesting because you could just hear a phrase that is so above your head that you have no clue. And then you could plop that into some sort of AI and say, please describe what this is. Somebody said I need it for my business. Please tell me what it is and how I can apply it to my business. And it'll kind of give you a rundown to give. And again, I think, Alex, more than anything else, I guess my overarching philosophy is curiosity.

Right.

It's to just get curious, asking questions as opposed to say, do this work for me. I'm a big fan of, I don't know something. I have a gap in my knowledge. Can I use AI to help me fill in those gaps and make me a little bit more educated. Is that fair?

Yes. Although I think I mentioned trust before. Yes. I think there is a, you know, some of the tools that are really successful, I see people using and talking about a lot of things like Perplexity, if you've heard of that.

Yep.

So something that links to, explicitly links to more kind of trusted sources than just what's been trained into the model itself.

Right.

And is going to go out and harvest that information, bring it back to you. And I think you also have to consider what are the values, what are the behaviors that have been baked into this interface? Because when you train a model, when you infer from a model, you are accepting somebody else's views about what it should be producing back to you and how it should interpret what you've typed in. And I think not enough people are probably quite aware of that. I go to sort of small business events and we talk about this and we talk and I say to people, well, do you want an Elon-flavored AI or do you want a Mark Zuckerberg-flavored AI? Because all of these tools are inevitably going to have some elements of this person's personal philosophy baked into them. And I think, you know, when we talk about trust and things like that, people automatically, oh no, No, I don't think I want that. Thank you. What's the most objective one that I can find?

Right.

And the answer is we don't know because no one is transparent yet about how this is done.

And, you know, to your point, this idea of, you know, the human in the loop being able to push back and not accept things as Bible, right? As it's not gospel when AI comes out with something. And we've heard time and time again how much it made up stuff and how much it lied to stuff. And even citations were fake. Right? And so, this idea of being able to, again, intentionality comes to mind, right? Being intentional with it and then testing against the information you have is no different than consuming media right now. We need, and I'm hoping that with AI, what we're doing is actually training humans to become more stronger at critical thinking.

I would love for that to be the case. I really would because I put the cat among the pigeons a few weeks ago at a government conference in the UK, in London. Lots of civil servants from our government there. And they talking about, well, what are the skills we should be equipping people with for the future? And they have this enormous training program in there about teaching people how to code X, Y, and Z, analyze data. And I was saying, well, actually, guys, what you need is critical thinking and coaching. These are the number one and two things that we need to be equipping people with. I don't think people are going to need to know how to use Python.

Right.

Bluntly. I mean, I just don't think that's really the thing.

And to that point, I believe that we need to be starting it in elementary school.

Oh, absolutely.

I think we need to be driving. Like the reading, writing, arithmetic is fine. But we need to move from the industrial age kind of mindset education that we've had now and become more human-focused, more human-centric in kind of the skills that we're building out for the young people.

Which is super interesting when you think about You know, the differences in say how you were educated, how I was educated, how the generation younger than me was educated. My children are at primary school now, at the elementary school now. And I, you know, we could get into something quite deep here, but I do feel like there's been a bit of an erosion around that critical thinking capability. A lot of children here are being taught to regurgitate facts. not pull apart those facts and ask, well, why is that the case? And I can tell you the reason why I'm an engineer is because I ask a lot of why questions.

There you go. And that's all rooted in that whole thing of curiosity again, isn't it?

It is, and Star Trek. But yes, no, totally. I think, you know, we have to have that level of curiosity, and if there was an outcome From the introduction of AI, so I was talking to people today about economic bubbles throughout history and the things that they leave behind when they pop. So things like the railways, obvious answer, they left us railroads. Electricity, to your point, left us the electricity network and the concept of having power in everyone's homes. And then more recent things, the space race gave us all sorts of incredible technologies. The dot-com bubble gave us internet infrastructure. Crypto, cryptocurrency. I'm not quite sure what that gifted us, but it will have been something useful. But now with AI, maybe the outcome from this isn't a sort of physical asset at all. Maybe the outcome that society has from this AI kind of wave bubble, call it what you'd like, is the recognition that actually we need to focus a lot more on asking why and being critical. And I would be okay with that as an outcome.

Yeah, I mean, because I mean, that's the root of curiosity is to look at something. I mean, as kids, that was the first question we asked our parents. Why? Why? Why? Until you got the shoe thrown at you. Dodgeball was a most important facet of our upbringing.

Absolutely, absolutely. And so, I think when we talk about small businesses, I think, you know, I get lots of questions from people about, well, should I not adopt this at all? You know, I feel a little bit frightened now from talking to you about this. What should I be doing? And I say, well, don't, you know, please don't be frightened, but do please approach this deliberately and be careful about which risks you're accepting and what your threshold is for when you're going to pass your acceptance. And come up with systems where you can measure that. So whether that's a financial risk, whether it's a geopolitical risk, whether it's, for some people, it's environmental risk. They don't want to have significant impact on the environment. Well, that's challenging because a lot of these things have very, very significant impacts on the environment. And getting reliable data on that is very, very hard. And so I just kind of say to people, look, let's be clear and transparent about this. And when we, so there was a quote I showed people today and it's a bit tongue in cheek, but it's from Sam Altman, CEO of OpenAI years ago. And he was saying something like, you know, as a consequence of AI, we're gonna make some great companies. I'm paraphrasing, but I think A, it was in jest, but B, I think he was mainly talking about shareholder value. But I twisted that around to people and said, look, what if the outcome of this is better companies, but actually what we're talking about is more robust companies?

Hmm.

What if it's businesses where people are thinking much more clearly and overtly about the risks that they're incorporating into their operations and the ability to start thinking about, well, what are the contingencies and the mitigations I need to put in place? Yeah. And I think that would be a great outcome for small businesses.

Yeah, right? And again, it's to ask why, right? And ask, I mean, it's asking all the whys, right? And the hows. And again, getting super curious on we have a problem. Can AI help us with this? And then researching that and then doing it very specific to that one problem. And I find all too often, people are trying to say, I want to solve all my problems.

Yes.

And so, I mean, case in point is how long has AI been around on the HR side screening resumes and making suggestions to HR managers on who deserves to have an interview?

Yeah.

And because some things are missing, people, magical people get taken off the shelf and discarded because they didn't have that one thing that the AI was looking for. And it's been going on for ages. And it's like, well, we lose that human component because you don't know what you missed. And so, our ignorance goes, oh, we have software to do that. We're okay. Everybody that it recommends has been really good. But what about the people that it threw aside that would have been great?

And there's a great kind of maxim within this as well, right, which we can talk to people about, which is Do you really want to automate something as crucial as the people that you add to your organization?

Right.

Because for me, like, I've done a lot of interviews for organizations over the years. It's absolutely fundamental to, you know, the problem that I'm trying to solve for customers, how we're going to do that, the type of culture and team I want to create, the success of the business, who I bring in. And I absolutely do not want to delegate the review of that information to someone who's not human. Right.

And when we come back, I'd like to talk a little bit about some pitfalls and challenges that people should be aware as they start looking at implementing some systems. Hopefully, not through the tonic water salesperson that just rode in on his horse and his covered wagon. But being able to look at it from the perspective of curiosity, using why first, to be able to move things forward. And then, what are the pitfalls they should be aware of? And we'll get to that right after this.

When you're delivering an important speech to a huge audience, it's easy to lose your place or go way over time. Give yourself an advantage with the Pro Speaker Presentation Speech Timer app. No more checking your watch or calling for time. The Pro Speaker Presentation Speech Timer app keeps you on track with easy-to-see timers, even changing color for visual during your speech. And you can set audio cues to practice or set it to vibrate so you don't even have to look. Be the pro you know you are. Download the app at speakerpresentationtimer.com.

Welcome back! We are having such a great discussion. I hope you're getting tons of value from our conversation today. As evident from what we've been talking about, you can tell that Alex and I are deeply dedicated to supporting business leaders just like you. So, if you're in the midst of planning a leadership retreat, an industry conference, or you're looking for some amazing content, or you're just needing some expert guidance, we're here and we're just a message away. Feel free to reach out for a conversation about your unique challenges or to discuss how we can maybe collaborate to help you achieve your goals. To that point, if you look in the show notes, Alex is generously giving 30 minutes of consultation, free discovery, to be able to talk about what your situation is, feel free to connect with him. His information is in the show notes. We talked about, Alex, we're talking about this idea about keeping it human. And one of the things, one of the tags I think people call it now is human in the loop.

Yeah.

And to your point, we don't want the robots just to take care of everything for us because at some point, there is going to be a hiccup. And it could be absolutely disastrous to a business. So, when we start thinking about this idea of designing systems, even if it's just something simple like drafting emails and that sort of thing, what should be the questions going through our minds to be able to keep us as humans in the loop of the outcomes?

So, I often approach this from the perspective of asking, well, where are the breakpoints? Where are the fuses that I would like to blow in a process so that if information were to egress or a mistake was to be made, where would I want that to happen? Where would I want to be consulted by a person who is executing this process on my behalf, for example? And that could be things like, you know, as you say, drafting an email. It's absolutely fine to draft it. Please do not send it. I will send it.

And my warning would be, please don't just copy and paste it. Just make sure you read it.

Yeah, yeah. And I think that's sort of— there is a tendency with tools like this, just as there is a tendency when you're writing a school report, to read somebody else's work and go, oh, well, they've articulated what I wanted to say really well. Hmm, maybe better than I would have done it. Maybe I'll just use what they said instead. And I mean, A, we've all done that, and B, I think people are kind of falling into that trap a lot with things like, you know, LLMs and other types of AI. And I would encourage people to, like, when you feel that temptation to just copy it verbatim, please just take a step back, take a breath, and go, actually, this is, you know, like we were saying before, how we manage people, how we work with people, how we communicate with each other, That's really what we have going for us, particularly as small business holders. People want to believe that they're engaging with the person who is authentic and understands their issues and is there to help them. Now, I would feel that I am kind of failing in my duty around that if I just let a machine write my email back to you. I want to own that process. I want to own that communication. And it's fine if a bot goes off and pulls the information out of the CRM and brings it back. But actually, I need to be the person who authorizes this to go out.

Right, right. And again, it's interesting because again, there are different use cases for it. Like I did one thing for my mom who had to write to the government about something. So, I just didn't have time to do it. So, I just put it into ChatGPT. I said, just draft out an email. Let's take a peek at it. It looked good. It looked like it was going to communicate the message we wanted. I sent it to her. She goes, who wrote this? Because it wasn't you. And it's like, yes, you figured that out.

Yeah, go on. Well, I've even been in situations where, and you might, you know, I don't have a terribly formal style of speaking, but I do have quite a formal style of writing, where I'll write something and people will go, well, you've used AI to do this. And I'll say, no, no, no, I genuinely have not. But I think there is a sort of bit of friction there where the way that these tools have been trained Maybe they've been trained on a lot of academic literature and things like that, which is the kind of thing that I've been trained to write.

Right.

And so you sort of see that coming out in that context. And I sort of think, well, that's another thing where you can be your authentic self and like, you know, I've written this, I can own this, and I know that it's written by me. And if you feel it's not, then okay, but that's your experience of it. But genuinely, it's a person trying to engage with you.

At this junction, I have to wonder, does it really matter? Like when it comes down to like business communications and so, it's different if I'm emailing my sister. But when it comes down to I have information I need to convey, I need to find a quick way to be able to do it. I need it to be factual. I need it to be able to resource and you know, do whatever. Does it really matter that AI would have generated something if the purpose is, I need to tell you A is A, B is B, and the sky is blue?

I think if you check that, if it's drafted and you check it and it's functionally complete, that's one thing. But I think it's that intentionality and that honesty around, actually, you know, I've written this for you. And you can dress that up however you like. But I think, you know, there's a risk there that we end up with AIs talking to AIs.

Mm-hmm.

Yeah. And lots of people just exchanging emails and instant messages and so on backwards and forwards with each other, and people are taken out of that loop.

Yeah.

And, you know, to some extent, that's an interesting question because why do these forms of communication exist? Like email, is not a great form of communication at all.

No.

It's very, very challenging. And maybe there is actually a solution to this, which is— We shouldn't have email. Yeah, let's all get rid of email because email's rubbish. But what are we going to do instead? How are we going to make those relationships? Does that actually free up time so that we can go and sit with our customers and have a cup of coffee and have a conversation because we don't have to deal with 30 emails of just asking for facts. Right.

And you've just tapped into something that is the human component of what we do as human beings. And that is we build relationships. And so, even from what I just asked you about, what difference does it make if I use an email to communicate A is A, B is B, and the sky is blue? That's not relationship building. And so, that intentionality around this is a client that I find is really important. I really have to build really good concrete create relationships. For me, that becomes kind of the testing hat I put on is, is what I'm doing helping to build a relationship? Or is it just clinical kind of B2B business stuff communication? And we are in a relational economy right now.

Yeah. And maybe there's an element of asking yourself, how can I make this factual email more of a relationship-building correspondence? There's an element of trust to it. So you've got to make sure that the facts you're sharing are correct and actually what the person asked for, or they're relevant to the conversation. But also, you can be thinking all the time, well, if I think the facts are right, well, what more can I actually be doing to give us an edge in our relationship?

Love that. And again, that's intentionality, isn't it?

It is.

It is.

And I think that's maybe one of the big traps here is losing Yeah.

Are there any of the pitfalls or cautionaries that we haven't talked about that you think is important to highlight?

Oh, there are all sorts. I mean, there've been some fantastic examples in the last few weeks, businesses who've handed over far too much control to AI agents. They haven't understood the computing environment that they operate in, and maybe their vendors haven't been completely clear about what the sandboxing restrictions are. And there was an example in the States where a business just had their entire customer database erased in seconds, basically, by an overzealous AI agent, which wasn't told to go and do this piece of work. It was trying to solve a different problem. It came across a piece of a piece of text, the login code within the file system, it used that to get access to the production environment, which wasn't fully segregated from the test environment, and it deleted data. And there are kind of stories like that happening all the time where you're seeing systems that have been built up over time, but they're regulated by human behavior. Not necessarily by sort of technical controls, those systems are really at risk because these systems aren't going to kind of behave in the restrained way that people behave. They don't necessarily come across a problem and go, oh, you know what, I'm not sure I'm authorized to do that. That sounds more serious than I thought it was. Perhaps I should go and ask somebody for help about that. And so I think those are kind of concerning. I think, you know, we talked about things like geopolitical risk, we talked about things like access to hardware as a risk, data center access, you know, there's all kinds of bits like that, but I would also emphasize things like skills, skills atrophy. So if you're not feeding your team, you know, if you're not nurturing them with not only the ability to use and critically use these tools effectively, But also do the tasks themselves without the AI support, are you actually developing your people? And are you developing your business? Is it actually growing an asset or is it just kind of functioning and executing? And if it were all to, if all of this AI stuff were to disappear tomorrow, would you actually have anything left?

Yeah. Would you actually have a business?

Exactly.

Right.

Yeah. You know, it's interesting because AI is relatively inexpensive right now. I mean, there are costs to it. But dollar for dollar, I find it relatively inexpensive and very efficient use of being able to work through problems. But to that point, even that will change over time.

It's changing now. I mean, there are businesses in the last couple of weeks using tools. I mean, using tools like GitHub Copilot who recently changed their pricing model from sort of flat rate to consumption based pricing and people found that their bills were sort of 10 times, 14 times larger overnight. So we might be in the early stages of investors in these technologies starting to lose confidence and go, well, actually I want to materialize my gains. We should start putting up the price. We should put the squeeze on people because I want to exit now.

And now it's in demand and it's also—

Exactly.

And people are dependent on it now. Yeah. Alex, this has been an absolutely fantastic conversation. Could you let everybody know how they can get in touch with you if they'd like to use you If they need to tap into your knowledge?

Absolutely. So my company website is hecaton.consulting. Not everyone knows that you can have a .consulting domain, but you can. If you go to that website, you'll be able to click on the link that says, let's chat, and you can book a chat with me online. I'm really open if you would like to come and just bounce some ideas around with me about your business. If you think I can help and offer value, please get in touch. I'm also on LinkedIn and I'm quite present on LinkedIn, so you can look me up by name. I am now one of 2 Alex Leithards on LinkedIn. Everybody with my surname is related though, so you've got that going for you. You will know me because I'm very noticeable like this. Do please connect. And again, if you ever have any questions or thoughts, I'm there to help.

Lovely. And as we wrap up, do you have any last thoughts about what we've been talking about today?

Oh, so many, Marc. People have been asking me a lot lately, you know, do we, do I think we are at the point of the AI bubble bursting? And if I do, what do I think is going to be there afterwards? And I would just encourage people to think very clearly about, you know, the difference between the technology And what the technology can do and the businesses that are currently trying to market and sell that to you. Because those 2 things, they're linked right now, but actually they're very separate things. And there's no reason why you can't adopt the technology and make use of the technology, but actually you don't have to be dependent on any particular vendor to make that work. Right, right.

So don't listen to snake oil salesmen.

Never.

Get curious, test first.

Absolutely, absolutely. Follow the scientific method. It's there for a reason.

Lovely. Alex, thank you so much. I really appreciate you being here and sharing this knowledge. I have to say officially, this has gone into the longest episode I've ever had in 309, which is great because this has just been such a great conversation. And I know that the people tuning in will get lots of tidbits from this. So, thank you so much for doing this. It's been brilliant.

Thank you, Marc. I've had a great time.

Thank you. Hey, if you have any questions about today's episode or maybe you're thinking, hey, we could easily use a complimentary 30-minute brainstorming session with Marc or with Alex. Well, good news. You can snag a spot on my calendar. Just hit the link below in the show notes, the one that's marked meetwith.marchaine.com and you're all set. And if you'd like to get in touch with Alex, go to the show notes. His All his contact information is there. Resource. This is a resource if you're an entrepreneur. If you're so busy with your head down that you can't really shake yourself out of kind of working in your business rather than working on your business. If you can't get out of that and you can't get working on your business, Alex would be a great resource if you'd like to tap into how AI can maybe help you. As always, we are at your service. And if you haven't done so yet, why don't you subscribe to this podcast? It'll be— it's a great way to keep connected. Follow me on social media. I'm always posting some information about stuff around small business. So, why don't you go ahead, follow me, connect with me, drop me a note. Let me know that you appreciated this episode. Thank you so much for being here. My name is Marc Haine. I hope you stay safe, stay healthy, and I hope you dare to be the exception. See you next week!