Phil Davenport: So before we dive into the what one thing, Scott, Richard, talk to me about the journey or the moment of realization. that led you into your what one thing.

Richard Geary: for me, my one comes probably the same place the color of this beard comes from, right, reality, is experience. So I spent 25 years big American SaaS businesses working globally with huge companies. And it was always about how do you the kind of investment you're making in SaaS software, right? And we're talking some of these companies are spending 20, $25 million a ⁓ year on SaaS subscriptions. And ⁓ what I found was, We were a solution looking for a problem. And that for me was like a cardinal sin effectively. my, my one thing is always know your problem before you go and look for your solution. I've developed this idea of, kind of a framework over the years, which is ⁓ people process And I think if you can align those three things against whatever your mission or goal is, you can deliver exceptional results. Right. And I think the technology part is the last part. Like that should be the last part. And that's where the solution sits or the final part of the solution anyway. So I think if have the right people, you've educated in the right way, you've given them autonomy or enablement and you've empowered them. You then have processes that make sense in the context of what you're trying to deliver in reality as well. Like it's fine to write a process down, but does that process actually have an unhappy path and a happy path? know, step one, two, three, four, five doesn't always go in that order, right?

Hayley Baxter: If you're a business owner right now, you've probably had that thought, we should be using AI.

Phil Davenport: But then it's where do we even start?

Hayley Baxter: What tool? What problem? Is it even worth it?

Phil Davenport: And that's exactly most businesses go wrong.

Richard Geary: So the reality sits in that process piece. Then if you go from, from, you know, having the right people with the right processes, then it's about finding the right technology to enable them to deliver against whatever that goal is. So that kind of PPT framework for me has always worked really well over the years. I've kind of built around that and thought about how do you then do the work to understand those things and it's define, align and deliver. So if you bring those sort of.

Hayley Baxter: So in this episode, we break that down with Richard and Scott, whose one thing is all about starting with the problem and not the tech.

Richard Geary: the three and the three together, you get this nice nine box that you can really understand where you sit in your maturity, your completion, your challenges in each of those kind of people, process and technology areas. And then, and only then can you really define a solution. And I think solutions that are driven by people with different experiences, opinions and perspectives are genuinely the best solutions because you're working from a toolkit rather than having one tool, which is like my perspective, right? So I think solutions built by, you know, not consensus, but by good research aligned to people, process and technology give you the ability to truly solve something in the long-term in a very robust way. And what we're seeing with AI is almost the complete opposite of that. Everyone goes, right, here's a 250 person organization. Everybody gets copilot access, go for it. We're to be 10 times more productive, right? And the reality is they're not because people don't know the use cases to use. Lots of your most conscientious people won't use AI because they're scared of it, right? So I don't want to train it because it's going to replace me or I don't want to put stuff in there because I think that's going to leak. Right. So there's a, like a permission thing, an approval thing that needs to happen from leadership to actually get AI tools embedded into the way that people work. But then training is the thing that, you know, Scott and I agree very heavily on this. We have seen tens, if not hundreds of businesses now who roll out a license and expect miracles. The reality is until you train and approve those methods of working, you won't see those massive gains, specifically not across a larger organization. You might see it in individuals here and there who are kind of those champions or self-starters. So yeah, the one thing for me is find your problem, find your challenge, find your opportunity, and then go find a solution for it.

Phil Davenport: So to break this down for small business owner, maybe they've got 30 staff, maybe they've got 50 staff, maybe they've got 100, 150. They've got a lot of problems. Normally lead generation is problem one, grumpy staff, problem two, we could carry on. But how do they define what can be solved by AI? And how can they define what a good training provider looks like?

Richard Geary: ⁓ so I think AI is very different to many other technologies we've come across, right? I've been doing digital transformations and huge rollouts for, for a long time. I think AI is the first technology where we provide access and then we encourage experimentation. And actually that's very different to any other technology you've ever rolled out, which was right. You provide access, you give someone the exact process they're supposed to follow and then know the exact result they're supposed to get. And so I think the challenge with businesses of nearly any size. is giving people an indication of the direction to go and how to use it, but actually making a space for experimentation. So you and I make it access to the same tool fill, but we may use it in very different ways, right? You're using it for email summaries and writing, know, prospecting emails. I'm writing it for meeting some, I'm using it for meeting summaries defining a good document structure looks like in an area that I don't know that well. So that experimentation, think, and how it fits into today goes back to the process piece for me, which is once you understand the process, you then start to understand the capabilities of the AI tools that are in the market and which one adds value in which kind of area of process effectively. you go about rolling that out very much depends on what your workforce looks like. Are you a workforce of 30 van drivers, 30 plumbers? 30 accountants, right? That's a very different approach to each of those things. But a van drivers is going to be, know, Gemini AI in Google maps, right? Route planning. It's going to be some kind of machine learning that gives the way to stack a van most efficiently, plus the route that's the most efficient to deliver against for time, for, you know, diesel costs, whatever. If you're a set of plumbers, it's probably going to be, how do I get quotes out quicker? I talk to a client, maybe I record that call. I then have an automated system that goes through that transcript, pulls out all of my standard pricing and builds a proposal that I just have to edit the draft and send. Right. If I'm a team of accountants, then actually, you know, the tools are there for them. Meeting transcripts, checking against, you know, legislation in ways that are new, doing research on what has changed since you last ran a chart of accounts for, you know, an import export business in the UK into Europe, for instance. Those, those three use cases are so fundamentally different. And the tools that you would use are so fundamentally different. There really isn't a one size fits all answer. And I think what companies are starting to do is just throw a license at it, be that chap GPT or copilot or Claude or, or Gemini and expect huge gains. And that I think is becoming a very, very common fallacy amongst businesses. You will see some uplift in certain places, but you won't see a, you know, entire organization be 10 to 20 % more efficient through just giving them a Gemini license.

Scott Evans: Thank

Hayley Baxter: You think it's amazing though that we live in a time where for a relatively small cost, you've got the opportunity to do that experimentation. Whereas, if you go back even probably like five years, if you were looking at implementing tech into your business, that's like a big project. It's cost heavy, it's resource heavy, and half the time you don't get it right anyway because you're trying to do 50 things instead of the five things you started out to get to.

Richard Geary: Yeah.

Scott Evans: Yeah. Yeah.

Hayley Baxter: There's something quite amazing about being able to experiment. think, what do you think about that?

Scott Evans: Yeah, I can go ahead if you want Rich.

Richard Geary: Yeah, I hardly agree. Yeah, I hardly agree. think the, if I look back 10 years ago, if I had the tools that I have at my fingertips today, I was, you know, running teams of a hundred, 160 people globally. can, I can see the efficiency we could have put into that. And actually what it would have given me the ability to do is delight and support my customers in ways that we couldn't do at scale at that point in time. Right. Um, so, you know, 160 person organization looking after three thousand clients. That could be 6,000 clients without even breaking a sweat because most of the knowledge work we do is quite repeatable once you understand the people, the process and technology that sits underneath it. so yeah, for 20 pounds a month, you know, I spend probably ⁓ reality, maybe 220 pounds a month roughly, but I'm a one person business behind the scenes that ⁓ is a team of four or five people 10 years ago. And so that cost versus benefit is amazing. Plus I'm not having meetings. I don't have to worry about from the mission. It all comes from a single point goes into my systems and they enact things for me. I'm a big proponent of human in the loop, never take an AI output and just send it out to somebody. I ⁓ the and then the human in the loop and lots of my things that I used to have to manually do are just happening. That is amazing as a business owner to be able to say that.

Phil Davenport: So from a granular perspective, a business owner that got 20 staff, know, Bob's widgets, loves a bit of AI. He thinks, you know what, I could do with a bit of AI. I think we need this in our business. It's gonna make things more efficient. It's gonna drive us forward. I don't want to pay 15,000 pounds for a consultant to come in. I'm not sure which road to go down even to look at consultants. I'm not sure what it's going to do. What practical steps can Bobby comes in at 9 a.m. this morning, makes a cheeky cup of coffee, sits at his desk. What step one, step two, step three, step four, step five to working out whether he's got a problem AI can solve, documenting it and starting forward on that journey. Does he just gamble and experiment? or does he start with pad and paper? How does he do that?

Richard Geary: not a gambling fan. think you can make very educated guesses. So I would say the number one thing to do is probably invest the bucks a month, 20 pounds a month, whatever it is, pick a tool and then ask that tool questions. So actually you could ask that tool to interview you about your business so it gets context. And the most important thing in AI is context. It will give you a really good output that a junior slash mid-level consultant would give you for. 10,000 pounds. If you were to spend two days sitting in a room talking about your business to an AI tool that you had set up correctly so that you're not sharing that data, you're not teaching the LLM based on your responses, I'm a big proponent of like, be secure in these things, but you could spend two days and take more than 10,000 pounds worth of value from those two days, I would think. It will give you the use cases, it will give you the possible outcomes, it will give you the tools that you should look at. Always question it, right? Don't take a first response. Always give it a hard time. But it can do those things for you in ways that we have never had the technology access to do before.

Phil Davenport: So Bob, he grabs his coffee, on to Gemini, AI, whatever we fancy, ⁓ and of them can get a monthly. I'm a big proponent, probably the same as yourself, getting a pro license, because then at least you feel a bit warm and fluffy that if it goes wrong, you've paid something to get some support. And is very first prompt, how can AI help my business, or ⁓ ask me about my business? How do they go down that journey? And at what point does Bob then need to say, do you what, might be worth getting a consultant in now? can see this value. This is where I'm gonna start to talk with an expert.

Richard Geary: Two things for me. I think the first question is always not necessary question, it's a statement. You are an expert business consultant. You are going to help me figure out where AI can impact my business. That's your role, that's your task that you're giving AI in this instance. You then want to give it context. You want to give it any constraints and you want to give it what the output needs to be. I would suggest that that output needs to be a pretty verbose interview process where you give it. all of the context you can ever think of. Because if you sit in isolation and try and think of, these are the things my business does, you'll miss 40 % of what you do. I also think there's probably a pre-step to that, which is Bob needs to truly know what everybody does in his business so that the detail he's giving the AI talk is correct. we, as leaders, we sort of get loftier and loftier, right? And more and more detached from reality. ⁓ truly understanding reality, use the tool to interview, use the tool to give you suggestions. use the tool to then question those suggestions very heavily. And then as far as getting to a consultant perspective goes, I think for me, consultants are becoming a way to shortcut the outcomes, not necessarily always improve the outcomes. Often someone can get to expert status, right? But it takes time. If you bring in an expert, what you're doing is you're, you're producing the time to value effectively in what you're committing. 20 pounds a month, you might pay that for six months and Bob has the time to do this. Bob can work on his business and figure out what it needs to look like, or Bob can bring in a consultant next week. You can then do that in a month. So I think consultants for me are about time, not necessarily access to information, expertise and outcomes anymore.

Hayley Baxter: So the question I've got with all of this, because it all sounds great and there's some really tangible things that you've shared there that you could go away and do. It would be quite fun to just get ChatGPt in a room and just, you know, talk about everything that's going on in life. Not that I do that at all. But like, how do you make it not overwhelming? Because... The reality is as a small business owner trying to grow, and even if you're not trying to grow, even if you're just trying to, you know, just keep going, there's like a million things you could look to fix or a million problems that you've got. How do you take it back to this is the right one to start with so that it doesn't become overwhelming?

Richard Geary: I think it comes down to money or time. Most of these outcomes come down to what's costing you money or stopping you making more money. What's costing you a lot of time or what would free up time to go do something else that can make you more money. Like as a business owner, really those are the two things you trade with, right? Money and time. So I think when it comes to the tool, its output and part of the original prompt should be, I want you to help me understand what will be the biggest priority and have the biggest impact on my business. using the values of time and money.

Hayley Baxter: So you would even use the AI to do that bit of thinking with you too. That's really interesting, isn't it?

Richard Geary: 100 % Yep. If you think about it, by the time you've spent, let's say you spend eight hours over a month with the tool, it's got eight hours worth of context that you've probably never boiled down to the way that you're giving it to that tool, right? Cause most people work in their business, not on their business. It's the reality. And so you've invested eight hours into working on your business. That tool now has all of that context. It can go through and prioritize against any measure you want. It could be that actually you've got a really unhappy staff. Right. And you want to improve their experience. You could give it that and say, that needs to be the measure for priority. I don't want to lose my 30 van drivers. All right. so I think it comes down to asking the right questions, but that tool, because it's read everything on the internet ever has the ability to think like any business consultant you name. So you could say act as a senior KPMG consultant, right. And it will do that. Or you are actually is a better way to do it. But yeah, I think you're taking a really well read in turn. You're giving it a detail about your business, which makes it a collaborator to you. And then you're asking it to, you know, be an expert in certain areas and prioritize things in the, in what is important to you. Then that makes it effectively an expert colleague. time, I would also ask you to challenge my point of view. As I said earlier, those opinions, expected experiences, right? variety there is really interesting, but I would say these are the things that I think matter. challenge me along the way to let's get to the point where we agree what matters for my business in my context, in my domain expertise.

Phil Davenport: I really like the idea that Bob sits down, gets his cup of coffee and he says, okay, we've got four hours, whatever GPT. Interview me about my business then he says okay I've got six members of staff member of staff one is gonna walk through and start answering questions now So interview them for half now interview because I think that that that exercise alone as you say that's gonna produce those systems it's gonna produce those and Then you can start to start to actually say right. Okay. Now you understand the business Now what suggestions can you make about AI? I suppose the concern is how much can we trust AI in this decision process? How much do we believe those answers? Do we then need to get an external group of people to have a look at the case studies or the journeys that we can follow with it?

Scott Evans: Thanks.

Richard Geary: Let's go.

Scott Evans: Potentially, mean you it's kind of acting like a second brain, isn't it? Because you spent all of that time with it now You've given it all the context giving it all the information and it's that second brain now. That's got all of that detail there It all depends on the questions that you're asking it. So you think your question was can we trust it? Yes and no, so I know that's like the worst answer to give but you you can trust it if you're asking it the right questions and you've given it the right information it's got the right context on your business and you haven't missed anything out or your staff haven't lied when they've interviewed them. And then it's no because AI, like Richard said, is trained on all of the data across the internet. So it's trained on like your academic white papers, it's trained on the Reddit threads and all the horrible keyboard warriors that are in there as well. So it's trained on absolutely everything. So you can't trust it 100 % of the time. And with tech as well, you can't trust tech 100 % of the time because there's always going to be glitches. There's always going to be with it. ⁓ what you can do is you like we said before, you can give it as much context and information and as Richard said, context is the most important. ⁓ then as long as you're really specific and prescriptive in what you're giving it and what you're asking it, you should then in turn, get a trustworthy output from it. ⁓ And when you kind of go into the depths of ⁓ using as well, you start talking about like, having knowledge bases and uploading documents to it and master documents and all these things as well. And what the AI can do then is draw information from those documents to give you a more trustworthy answer off the back of that as well. So that's kind of how I would see it. I know if Richard, you would agree with that.

Richard Geary: I do, absolutely. think there's a secondary element to trust, which is don't trust it. Like, you trust a consultant you've never met before who walks into your business and tells you to pivot and sell smaller widgets than larger widgets? The answer is no, right? You're the domain expert. You know your business, you know your customers, you know your market. always question it is number one for me. And then I think there are tips and tricks you can ⁓ to any of prompting or interviewing process with AI that allow you to get a better feel for. being able to trust it full stop. And those are things like don't cite a singular source, right? If you're telling me that there's a use case, I want multiple use cases that show the same thing and have the same outcomes. I don't want a one source reference to an obscure law, right? I want three independent sources. know, we've all seen the very early on lawyers getting caught out of just made up court cases and stuff because they just trusted chat GPT. So think for me there are more advanced prompting techniques and more advanced collaboration techniques with AI that get you to the point where the output is somewhat sanitized you can trust it more. And to Scott's point, as you get, as you give it more and more context, you know, you can ask it to focus more on what you've given it and less on just going out to the worldwide web and, you know, finding that Billy Thornton on Reddit who, you know, sells Gator toenails at a thousand bucks ⁓ a right? Like that's not, it's not a use case. ⁓ probably relevant to the business of accountancy or van drivers. so I think giving it that curation, giving it the challenge, ⁓ making sure you are giving it a hard time. Like you can be really rude to AI. It does not care. Right. It's not going to, it's not going to take it personally. All it's going to do is try and make sure it's ⁓ what you want it to answer. One of the things that, yeah, one of the things you'll see is like hallucination and bias are the two big things that come up whenever we talk to businesses. Right.

Scott Evans: until there's humanoid robots.

Richard Geary: Businesses care about consistency and quality ways that individual users don't, right? I'm to care about creativity and fun and all the things that you can play with AI and get. But to get consistency and quality, you need to understand how hallucination and bias work. So bias is based on the fact that it is trained on the internet and the internet is inherently a white westernized view of the world because that's where the content comes from, right? So it will inherently be biased. It's much better now than it was two years ago, right? If you asked for a doctor and a nurse two years ago, you got a white middle-aged male doctor and you got a young blonde nurse. That was the reality. I did that experiment again in a course last week and actually we got two different races and almost a similar age, which is unheard of. It's still male doctor, female nurse, right? But that is the reality. There are far more female nurses than there are male. So that's where bias comes from. Hallucination is a more interesting one. So if you ask a very specific question, so Phil, you say, I'm Phil Davenport, what's my favorite color? Right? It won't know that answer unless you've answered that in written form somewhere in an interview, right? But it will make take an approximation. So it find answer A and answer B and it will try and extrapolate between those two things to tell you what Phil Davenport's favorite color is, because that's what you specifically asked it. Now, if I've got access to this picture, for instance, of you right now, it's probably going to guess your favorite color is somewhat on that purple scale, given the logo behind you and the top you're wearing, right? So it's going to infer from what it doesn't know to give you an answer. And that's where hallucination becomes dangerous because it's done with confidence. And so giving a hard time of like, okay, how do you know my favorite color is purple? Right? And it will then give you its thinking or approximation or sources and you can challenge those. It's no different for how do I pivot my business to sell more small widgets than large widgets, right? It give you answer A, challenge answer A, get to B, C, D until you are happy that you have enough sources, enough information, and it's understood enough context for that truly to be an accurate answer. Then you can go in action.

Hayley Baxter: So it's really interesting, the whole conversation, there's lots of things to think about. So what we're saying is to summarise then, we start with the problem, not the tool, and then we build the structure around that. But actually, before you even get to that, you can use the AI tool to help you define what the problem is to begin with, and where you should start hundred problems that you've got to solve as a business owner, which one is the right one to focus on first, that's going to help give you back either time or money.

Richard Geary: Yeah. Or increase over those two things. think, interestingly, there are two, there's a two-stage problem there as well, right? So problem one is, I know I've got a problem, I don't know what that problem is. Right? So your first purchase is a tool that helps you define more clearly what the problem is. So you're answering a problem, a problem like 0.5 before you get to problem one, which is like, okay, problem one is our van drivers are often off sick, they have inefficient routes and we don't stack those vans well enough.

Hayley Baxter: Yeah. Yes.

Richard Geary: Those can be the three problems that come out of the research phase of what is my actual business problem.

Hayley Baxter: Yeah, it's really interesting. So Phil and I have started a little experiment this recently about where we are implementing some of the what one things that people have shared on our podcast so far. we try to implement it over 30 days and see what impact it has and share different things from it. We're halfway through the first one. This one, I'm really excited to do this one and where takes us. ⁓ If have enjoyed listening and want to get in touch with you or find out more about what you do and how you can help businesses, where do do that and how do they connect with you?

Scott Evans: We can head over to our website, wearefoundry.ai. They can also follow us on LinkedIn as well. both have personal LinkedIn, so that way we push out a lot of free information and value, we think. And also the AI Foundry also has ⁓ a site on there. When you onto ⁓ the website, can also join our free community, ⁓ which is to, ⁓ we say for business owners, business leaders, but ⁓ it's open anyone who really wants to learn ⁓ about AI. questions, not feel scared about asking questions either, and hopefully collaborate and get answers with people who are kind of like-minded in a safe space. you can head over to there and we'll be in there as well ⁓ to push as much free ⁓ to you as possible ⁓ we'll see you some of our events that we're going to put on as well in the coming months.

Hayley Baxter: Well, thank you both so much for joining us.

Richard Geary: Yeah.