Practical AI for Founders Who Want Real ROI

0:00My prediction is that within two to three years, every single one of us will have a digital AI twin. The only question here is not if we have one, it's whether it sounds like you and it actually represents you. The reality is there are some risks and as business owners we need to address those, mitigate those and make sure that we're not firefighting problems down the road. I don't believe we're in this B2B, B2C world anymore. It's very much this H2H, human to human connection. The key point there is human in the loop. That's crucial. That's the centre of the way we think about AI. And that really means AI enhancement, not human replacement.

0:37So we, collectively, as individuals and as businesses, have got to move to the top and add more value using the technology, not fighting it. Because as a battle, you are going to lose. Welcome to Screw It, Just Do It, the show for those who decided to screw it and just did it. This is the podcast that brings you face-to-face with the entrepreneurs and business rebels who threw caution to the wind and built something extraordinary. Good afternoon, everybody. My name is Rob Hanna. I'm the founder of KC Partners. We are a legal talent solutions and content creation firm. And I'm also the proud host of the Legally Speaking podcast, which is the number one podcast for law in the UK.

1:16And we are sponsored by an organization called Clio, who are a$3 billion legal tech unicorn business. and we've been taking the show around the world since 2019. I'm excited to be on this panel today where we're talking all things impact AI with impact, how founders are using it to scale smarter. And we have a star-studded panel who are going to introduce themselves from left to right now, and then we're going to jump into today's discussion. So please introduce yourself. Thanks, Robert. Good afternoon, everyone. My name is James Smith. I'm a tech entrepreneur and founder, and I'm currently CEO of a business called Critical Cloud.

1:52We're helping transform modern cloud operations for high-growth SaaS businesses. Thank you very much. Thanks, James. Hello, everybody. My name is Dominic. I'm the co-founder of CleverFox AI. We're an AI transformation agency based in sunny Brighton. We support businesses on their AI transformation journey right through from initial consulting through to design, build, and implementation. Thank you very much. Good afternoon everyone my name is Sabrina I'm the founder of 2Commerce PR so we build personal brands using and accelerated through AI growth. Former BBC One The Apprentice for any Lord Sugar fans in the audience and specifically how do you increase and amplify your brand online so that you can get accelerated across the competitions.

2:40Thank you very much. I'm Pierce former Dragon, a little bit of TV but the day job is I'm the co-founder and Exec Chairman of Implement AI, Read the Shoes. So we help companies implement AI, but we built a platform. So we basically supply digital workers, sales support, customer success, computer use, and analytics, organizations, all sectors, all sizes, from 2 million revenue to a FTSE 100. Fantastic. So as I said, very star-studded panel joining me today. There's a lot of hype around AI. So what I'm going to try and do as the moderator today is make this very practical and tactical so there's at least a few things you can go away and immediately implement into your businesses.

3:22So Piers, I'm going to come to you first. When founders talk about being AI first, what are the key design principles they should adopt from day one to actually stop being involved in the legacy models they're probably already using? So two things really. One is if you're really starting from scratch, you should be really thinking about AI first. Now, what does that actually mean? So that doesn't mean creating some fully autonomous organization. We're not quite there yet. What we say is you've got to think about where is the technology today? Where is it reliable? Where can you create a meaningful, measurable ROI?

3:57What processes and workflows can you actually automate in a way in which you know they're actually going to do what they're supposed to? So you can play in the kind of R &D space where you're using tech that's not quite there yet, but it is potentially a huge ROI. we play in the box where in that jagged technological frontier depending what you do where you are you use technology that actually works so if you're already got a business start to automate it but don't try and fall the ocean start small think big but I'm telling you now you've got to start thinking about whether you are one million revenue up to 10 million in revenue wherever you are you've got to start thinking about how do you move towards AI first because your competition, either the incumbents or the new entrants, the founders, they are going to be AI first.

4:44They're going to have 20 to 30 % of the people you've got in your business. So you cannot compete on margins. So that's the way the world's going. The question is, who will get there first? Because catch up is going to be very difficult. Love it. Just out of interest, who in the audience would say they have an AI first business at the moment? Put your hand up. We should put our hands up guys. Good stuff. Okay, a few of you. I mean, we do AI and we haven't. I'm quite surprised anyone actually says that today. Yeah, that's interesting. Okay, well, James, I'm going to come to you because you have scaled multiple tech ventures over the years.

5:19So how can SaaS companies specifically and digital agencies integrate AI without disrupting their existing cloud operations? So I think there's a couple of things to think about. The first thing is to look at the processes that you've got going already. Typically it's a similar approach to getting started with automation so mapping out the processes and the workflows as Piers said is a great place to start and as part of that really understanding the data that you've got and where that data sits and kind of looking at how you're processing that information what's available to you. So that's the first thing I think the piece of advice we just heard is don't try and do everything at once.

6:01So the next stage, once you've got those workflows mapped out and understand what data you need inside them, is to start small. Pick an area where it's quite low touch to get involved, maybe some reporting elements, maybe kind of some customer engagement activities, but something that's not too mission critical while you kind of get used to kind of what you're trying to transform. And ultimately, it's something that you can measure, right? once you start to deploy some machine intelligence and machine learning into those areas, can I see what it looked like before and looked like afterwards? And then the third thing after that is think about how you're going to scale that out.

6:41So the big perspective, you know, wave migration planning, so taking pilot after pilot after pilot and moving that into production one at a time, you know, is normally a good recipe for success because you're not changing everything at once. Yeah, I love that. And I would just sense it's habit. Once you start dipping your toe in and you're seeing traction, it becomes more of a habit, becomes more native, becomes more confident, I guess, in terms of... I'll ask a question for you. Who's confident using AI at the moment? Put your hand up. This is another one we should put our hand up. Okay, interesting.

7:14Alright, James, thanks for sitting there patiently and Sabrina coming to you in a moment as well. Please, can you share with the audience here what AI governance is and why it's important and how business owners sat in the room can use it to enable safe and effective AI adoption. So that's my question isn't it? Yeah nice easy one. I'm Dominic not Jeff. So yeah we're gonna be hearing over the next couple of days about all the wonderful things that AI can do for our businesses and I'll be sharing some of that during my presentation tomorrow. The reality is there are some risks and as business owners we need to address those, mitigate those and make sure that we're not firefighting problems down the road.

7:54So an AI governance framework enables you as a business to do three things, to adopt AI safely, compliantly and confidently and you need all three, it's really important. So on the safety side, as business owners we've all seen the news stories of data security breaches and your IP being leaked by rogue AI or a member of your team using AI irresponsibly. So as part of a governance framework we have some tools and resources and protocols in place that prevent that from happening. So for example you might have a AI model register so everybody in your team knows which tools they can and can't use and in which way.

8:37You would also have an accountability matrix so again everybody in your organization knows who is responsible for which data flows and which tools. So that's the kind of safety side. On the compliance side, it's a bit of a grey area because we don't have a UK AI Act as yet. We were promised it by the Labour government in the spring. Funnily enough, they haven't got around to it. So they're promising that by, I think, May of 2026. But we can't sit around waiting for them to do something useful. We have to be proactive as business owners right so when we design our governance frameworks for our customers we are aligning it with the ISO 42001 standard and also the EU AI Act so when the UK legislation comes through it will probably be somewhere in between the two and so we need to get those frameworks in place now and so that when the UK legislation comes out we can just adjust our AI policies and our and our government governance frameworks the third part is crucial and that's confidence.

9:38I hear this from business owners all the time, you know, we've implemented a couple of AI tools, the team start using them, and then they get stuck, and then too embarrassed to ask how, you know, call their boss and ask how to use it, so they quietly stop using it. And then, as a business, you've wasted the time and the money you spent on that implementation, and you're not getting the benefit, the productivity amplification that you should be from those AI tools. So there is a training and upskilling part to governance right it's really important that you have a register of who has had which training and what that kind of skill level is because if your team feel confident then they feel empowered and actually you'll be getting the full benefit from from your AI transformation yeah no couldn't agree more and kind of builds on what I was saying for in terms of those those habits if you've got the infrastructure you've got things in place people feel more confident you take more action more things can happen Sabrina something I'm passionate about something you do super well I want to move to now is sort of mixing AI and personal branding because you've developed what you call a personal brand operating system so what does that actually look like in practice again for the founders in the room and how can they use AI to scale their online presence without losing their personality I love this topic a quick hands-up is anyone actively wanting or trying to build their personal brand as a founder amazing so most the mistakes that I certainly did myself at the very beginning was I went on to something like ChatGPT, I typed in exactly what I wanted as the output, and I got something that was half decent, and it was just noise.

11:12And right now, especially if you go onto LinkedIn, you'll see so much noise. And when we look at using something, and I'm using ChatGPT as a very simple example, is that we kind of start using it without any kind of training, and that's kind of bringing on a new employee into our team without giving them any background into the business or who we are. So we specifically brought out a personal brand OS operating system. So what this is is a really intense document that you could probably start doing tomorrow. And you start feeding it everything about yourself, your tone, your authenticity, what's your story, what kind of things do you like, and what kind of things do you not like.

11:52Then you also get into the compliance side of things as well. And once you have this, you're able to start scaling. So when you start putting out content, it sounds exactly like you, but you're not putting out noise. The second level to this is creating one for your brand. So all of your brand guidelines, how would you onboard an employee? And what you can start doing is integrating this into your team. So you can say, okay, team, we're going to put together the ultimate document, one for the brand and yourself for your personal brand, and they can start speaking to each other. So then when you start looking at agents and more no-code AI tools, you're able to deploy this into organization without having to start back at the very beginning.

12:33Love it. Very practical, tactical. Hopefully you have learned something in the first part of this panel session. If you have, write it down. Biggest tip I can give you for the day. Okay, now we're going to move to scaling smarter. So balancing speed and strategy. Piers, back to you. Lots of leaders ask about ROI. So how do you measure the value of digital workers in a way that boards and investors actually buy into it?

13:00There's two dimensions to this, right? So there's measuring the way you used to do it. So I'll say this, you've got three vectors in your business, okay? You've got velocity, variability, and value. So if you're low to medium value, humans, they can do that pretty well, but they can also do high value. AI's are good at low to medium. You're not getting AI to sell a private jet, right? Velocity. Velocity is about data, volume. AI is very, very good at that. It's like a fighter pilot tracking 200 targets. Humans can't do that. That's AI. Then you've got kind of variability. If somebody's doing lots of different things, AI is not very good at that.

13:36It's very good at linear, clean, we call them, clean workflows. So if you're high velocity, data, low to medium value, low variability, automated AI. If you're high to medium high value, high variability, low velocity, humans. But the point here is everything in the middle is a machine human partnership. You're augmenting yourselves and your human resource and you measure ROI in two ways really. One is how do you measure it today? How can you increase the ROI? But also what we find is this is, I'll give you an example right, we can talk about real examples. We've got a company recruitment, construction recruitment, their problem The problem is 200 ,000 people in that database, they've got four people on the phones all day, every day, they can cover 5 ,000 a year, nine to five.

14:25AI calls 3 ,500 people in three hours, has 650 conversations, and books 50 jobs. Their problem in their business now is not the flipping ROI on their pipeline, it's how do they now, second order issue, how do they deal with that deal flow in their business? So you've got to start thinking about what's ROI as it is, but also you're going to have to re-engineer your business if you start implementing this technology another example we have a very large national pharmacy chain using data you've got a lot of data if you're not recording everything your business and I mean everything powwows meetings video calls phone calls record everything because now you can pass it you can analyze it so we have a national pharmacy 30 million prescriptions a year what AI can now do in terms of ROI is your known knowns your dashboards your known unknowns things you go and look for in your data your unknown unknowns things you don't know you need to know we sent an AI looking into their data and they found if they pick 83 different conditions take out one which is contraception you can get blood pressure test contact those people book them in the pharmacy when they go back in to get their prescription we know when they're going to go back here because it calculates it.

15:41That's a 400 million pound business and it found a four million pound revenue opportunity with a 75 grand cost. That's it. That ROI is off the chart. So don't think about the current ROI, think about what is the art of the possible if you implement this technology properly. Yeah, love it. And thank you again for giving such clear crystal examples. And it just shows the possibility of what actually, if you take that approach with an open mindset and take action on this, what you can actually achieve in your businesses. Okay, James, coming back to you, Critical Cow blends AI tooling with human-in-the-loop engineering.

16:16So how do you see the balance between AI automation and human expertise evolving? Yeah, I think the key point there is human-in-the-loop. So that's crucial. That's the center of the way we think about AI. And that really means AI enhancement, not human replacement, ultimately. So we're taking really smart, capable, deep domain expertise and using AI to really enhance the capabilities and the efficiency. So let me bring that to life a little bit. One use case, for example, in our world is we're taking millions of logs and millions of data points from different cloud systems, bringing those into tooling that allows our engineers to spot trends and patterns super quickly.

17:02That's not possible for a human to do across a data set of that nature. But what AI is not very good at is the reasoning and judgment about what it does with those decisions after it's kind of identified the patterns and the trends. And so we need a human in the loop at that point to look at the output from that information and then make an educated, context-aware decision around that. So that's one example. But the second one is taking that deep domain expertise. We're training fine training models with lots of documentation and lots of history and context around how they can make decisions in cloud environments based on response to failure, as an example.

17:49But we don't want to let those agents go away and just make those decisions without having safety net around them. So again, when an agent makes a decision that it wants to do something, there's a human in the loop that will sense that decision. So we're using the humans in that instance to kind of add the safety net at the moment, but also to keep adding additional context onto the capability to give it more awareness and more context. So that as we kind of move forward in this AI journey, you know, I think reasoning and judgment is something that's definitely in our future with AI. but we need to get there in a safe way at the moment.

18:27So for us it's about kind of supercharging humans and not replacing them with a capability. Yeah, now again, thank you for being so clear and giving hopefully some examples all of us can relate to and again go away and hopefully take action off the back of today's panel. James, coming to you, we're sticking with real examples. Could you give me maybe one or two examples of custom AI automation or systems that CleverFox AI has actually built that entrepreneurs with limited AI knowledge can easily understand? Have you got my name written down as James on your sheet? I'm Dominic. I'm not offended, but no worries.

19:02Yeah, of course. Actually, Piers has already alluded to one of the use cases, but I'll give you a similar use case in a different sector. What we want AI to do ultimately is a few things. We want it to help us monetize our dormant data. We want it to free up our teams from doing the tasks they hate. and we want to be able to automate repetitive tasks, right? So the AI voice agents are really powerful for this in terms of lead reactivation. So we're selling these into the estate agency market. Every estate agency has a team of people relentlessly calling their database of customers to see if they happen to be planning to move house.

19:39They've got a load of listings they need to sell or rent. It's every estate agent's least favorite job. They normally get the junior guys to do it. They end up resigning because it's a miserable job. Why? Because 99 % of those calls are no, go away, go away, why are you calling me? But every now and then, every 500 calls, someone says, yes, I'm a landlord, I'm selling my portfolio of 20 houses, let's have a call. So to give you an example, we've got a customer with seven branches, they've got 40 ,000 people on their database, and they've got three full-time employees, so eight grand a month payroll, and they're just hammering this list.

20:14Now, as Piers says, you can only do$80 a day or whatever that is. We can now run an AI voice agent to do 1 ,000 calls a day or whatever number you want. It doesn't get upset when people tell them where to go. But when that 1 in 500 customer actually says, yes, I am interested in a call, that voice agent is able to have a proper conversation and connect that call to the relevant person. So it's freeing up the team. Actually, it will generate viewings which the human team can go and do. That's the in-person stuff that we should be doing. the person receiving the call doesn't care if it's an AI or a person it's just no go away whatever I don't care who's called me and the estate agency crucially in their mountain of data there's a few nuggets of gold in there and typically with a human bashing away an axe at that mountain of data you can only unearth so much the AI can go through like a robotic power beast through that mountain of data and extract every last bit of gold yeah and that's really exciting.

21:11Have I got time for another one? Absolutely, please. So another really powerful thing that we do a lot of is automating legacy workflows. Most businesses have run their business in the same way for the last 25 years. We've been in the SaaS era. So I'll give you an example in the veterinary sector. If you take your dog or your cat to the vet, that workflow has remained the same for 40 years. If you've got an 11 o 'clock appointment at 5 to 11, your vet is frantically scrolling through the stuffy on CRM looking for who's coming next what is it an operation is it a vaccination that's manual processing of data which is very inefficient so of course now a click of a button the AI will produce a two paragraph relevant summary saving the vet from scrolling through ten years of clinical history then at the end of the consultation your vet sits at the computer tries to remember what was said typed in manually into the computer this is what happened blah blah blah again now we run an audio recording through that consultation we capture all of the data not just what the vet can remember that is unstructured data the transcript we then take that unstructured data which the ai turns it into structured data into relevant fields which you can then search across and by the time you get back to the car with your cat or dog you've got an email bang hi mr class great to see you today automated email betsy's had this vaccination see you in six months time so the patient gets better care because you've captured all of the data.

22:34The vet saves about two hours of boring admin every day. And the vet group is saving, in this case, they've got 200 vets across 40 locations, two hours a day per vet. That's 1.7 million pounds of salary time saved. Unbelievable. And what I love about this, we'll all see deadlines, data is a new goal. But these are practical examples of how this data is transforming businesses. And it's just really exciting. if we can get our heads around this. So again, if you do have questions, we're hopefully going to be allowing a few at the end. Please do get them ready and we'll try and come to you. Now, Sabrina, coming to you.

23:11Many founders fear AI will make them feel less human. We talked about human in the loop and Pierce was talking about clobbering with human. How do you advise leaders to keep authenticity when scaling with AI? When you're looking to build your brand as a founder and you use AI to accelerate that process, you're ultimately expanding your team. So providing that you create your personal brand operating system, you're already showing your story, your values, your tonality, and what you stand for. And then you're creating the values and you're scaling this out. My prediction is that within two to three years, every single one of us will have a digital AI twin.

23:54And the only question here is not if we have one, it's whether it sounds like you and actually represents you. Because there's so much noise right now on the internet, and there's so much content, hopefully you're putting out a little bit of content too. It's either going to create an AI version of what's already out there, or you're able to create what's there. So imagine that you, everyone's going to voice now. So when you have an idea, something happens, maybe you had an interesting talk with a team member, or you feel super motivated on a Tuesday morning, you voice note this then my team take that voice note we put it through something trans transcribe in AI and there's a ton of automations and we have an AI agent who's able to take that content duplicate it in the way that I want to be shared based on everything that I've previously done over my last thousand posts so that it sounds exactly like me so going into AI I feel if we're scared around the authenticity then we're actually fearing AI and it's the fear that's stopping us behind.

24:56So AI is not here to replace us as a personal brand. It's here to accelerate the values, providing that we give it the values and the story that we'd like to. And that's the biggest mistake, I think. I also think anyone in the next six months, like this is your six month window, you have six months to go across. Because all the current creators and founders who are already building their brands will just accelerate with AI because they'll be able to leverage what's already there. So without stepping in now, you're going to go not only massively in competition with the big names already, but you'll also be attention competition with the AIs that are coming up and creating online clones.

25:33So if you have anything to take away from today, and I do talk about this in my keynote tomorrow, it's how do you use the next six months to double down, create those online versions of yourself, leverage so that you're so far behind otherwise in six months that it will be really, really hard to catch up. Yeah, no, really well said. And again, some great examples. So we're going to move into the third part of the final part of the panel now. Just to recap, we started with from hype to impact, making AI practical. Then we moved into scaling smarter. I think we had tons of examples across the panel there, balancing speed with strategy.

26:07We're now going to talk about future-proof growth with AI. If you did learn some nuggets from the last piece, write them down now and go and take action on them. Piers, so if you had to redesign workflows in a very heavy, slow-moving industry, you and I have law, finance, healthcare, where would you start to unlock AI's value the quickest? So everyone, say in law for example, everyone's fixated on the regulated activity, financial services, diagnosis, law, whatever it might be. We kind of ignore that at some extent. There's lots of companies trying to do that, raising billions of pounds. But every organization has very similar workflows at the end of the day.

26:45It might be sales, marketing. It might be HR, support. It might be customer success. In law, it might be compliance, onboarding. And the point was made, actually. Look at the workflows in my bottom left box in my two-by-two matrix, where the tech works and there's a decent ROI. And the key is this. Think about a pyramid like that, right? and it's been filled since the first industrial revolution with technology and they were tools now the tools the intelligence can use the tools and it's filling that pyramid so we collectively as individuals and as businesses have got to move to the top and add more value using the technology not fighting it because as a battle you are going to lose and the point was made about you know i I always find it dangerous anyone ever saying, it can't do this, it won't do that, it's a human to do this.

27:40Because I'm telling you, within five years, it will be better than us, better, cheaper, faster, smarter than us at every cognitive task. The only question is, how do you implement it to free yourselves up and your colleagues and your team to add real value, not deliver stuff? Yeah, love that. And I always talk a lot about collaborating, collaborating with technology, collaborations, domination. You give a great example there of actually how you could do that and the impact it would be. And probably get in touch with you at implement AI as well. I've got one here, I've got robots. So within another 10 years after that, they'll be caring for your parents, right?

28:14Yeah. So the embodied AI, they're coming as well. So cognitive labor will go away and eventually physical. The question for you is, in that world, how do you carve out the values that you can actually hang on to? Because it won't be that straightforward. Yeah, something to definitely go away and think about, folks. Okay, James, coming back to you on the end there. With a background in cloud and security, what do founders need to know about keeping AI adoption safe, compliant, and sustainable? That's a good question. So I think there's a couple of things to think about, but it all boils down to one thing, in my opinion, right now, and that's visibility, getting observability across what's going on.

28:55So what are these tools doing? you know we talk about kind of when you're using an LLM model there's things called tokens how many tokens are being processed governs how much these things are going to cost so getting visibility of the cost getting visibility of where your data is not so much about where it's located but where it's actually being processed we've heard a lot about kind of security and compliance and one of the things that sort of catches people out is you're feeding kind of information into some of these tools whether it be a chatbot whether it be ChatGPT through the browser, that information is going off somewhere to be processed, and it may not be in the region or location you quite expect it to be.