Ep10: Business and A.I: a force to be reckoned with

We need to know what the problem is going to be that we're going to solve before we can start attacking a solution. And we have quite a few customers now who are excited to work with us because instead of just finishing the project, we'll go back and go, oh, this thing's now available. We reckon we could do it even better.

Large action model, I don't think that's been dwarfed. I just don't think people have played with it yet.

Hi, I'm Freya. Welcome to the Skynet Chronicles Business in the AI Age, the podcast that dives deep into the world where artificial intelligence meets business. Join us as we explore the latest trends, challenges and opportunities in this ever evolving landscape. From innovative startups to established enterprises, we'll uncover the secrets behind the successful integration of AI technologies. So grab a cup of coffee, sit back and get ready to unlock the mysteries of business in the AI age.

Foreign.

Chris welcome back to the Skynet Chronicles Podcast.

It's nice to be back. I feel honored to do two.

It's amazing. Great to have you here. So today I'm going to quiz you on all things about how we're helping companies.

Okay.

So I think if we start with a problem, okay, that's generally the place that we like to start with. A customer is either they've come to us with a problem or they say we want to use AI and we go, hold up, what's the problem we're trying to solve here? Tell us a bit about that process.

Yeah, so you're absolutely right. First, the most common thing we hear from someone who comes to us is we want AI. We don't know what we want it for, but we want AI because either they want to be seen to be using AI because it's a cool thing to have on your website, or. Or because they generally are interested in the technology.

We want stuff, it wants stuff and.

It want it should be shiny. Inevitably, there are few companies out there who are ready to go from where they are now to something AI enabled. Data condition is normally the biggest one. So if you're wanting something AI to help, but you're still running on a access database from 20 years ago within your company, then there are some steps that we need to take to take you from where you are to being AI ready if you like. So I think they're the customers. We enjoy those that want something, but they don't know what they want. And whenever we sit down with a customer, the first question is always, what's your biggest problem? Because when it comes to AI or automation generally, we need to know what the problem's going to be that we're going to solve before we can start attacking a solution. Where you see companies starting to go, oh, we want this bit of AI to do this and this bit to do that, and trying to model it all together, you generally see an output that's, that's not very good. So try and find one problem, the biggest headache, and we'll solve it with something. Automation or AI or the combination of both of them. Or a combination of both.

I mean, I quite like it where employees within a company have been playing with some of the technology. So playing with ChatGPT, playing with MidJourney, for example, so they've got a feel for what might be possible. And we go, okay, that's what you're currently doing. What's the problem you've got within the business? How can we use elements of these technologies with other things to create you something entirely new?

Absolutely. And that's what we do when we go out talking about AI when we're doing well, if it's webinars online or if it's in person sort of speaking events about AI. You can see light bulbs coming on around the room while we're talking, which is great. And it's not down to us being great speakers at all. It's because you're starting to talk about we're pretty good. Maybe it's that good practicing. Yeah, we are good at it now, but it's talking about to people about the art of the possible. When people know what is possible in terms of the technology now you can see them start to go, ah, but maybe if I do that, this thing or maybe if we, we think about this problem or actually I've got this problem that I've never been able to solve because I don't have the time, the expertise or the money. If, you know, if you perceive it to be an expensive problem to solve, you see people start to go, maybe I can do that in less time, less money, less stress. And that's the wonderful thing for us is being able to sit down and someone slowly start to hone in on a problem that they want to solve with AI and automation. But yeah, data readiness is normally the biggest challenge for us.

I was chatting with a guy who runs an excellent local company. I really enjoy their produce. I'm not going to name drop them at the moment because they're not yet a customer, but he was saying that they have a problem within their business about understanding exactly which products are sold in which locations and even which parts of the business are profitable.

And I'm like, let us in, let us in.

Because even where they don't have that accurate data, they've probably got enough information that we can build a model that would guesstimate to a level of accuracy that's far higher than they have now. And just that as a problem. That's brilliant. That's a very specific problem they've got. They've got data around them. Getting that data into a format that can work for us to solve that problem. Oh, it's heaven.

It is. So we're doing some work with a local charity, Little Miracles at the minute and they're a prime example. So they've got data in different systems. They've got an ERP system which is zero, I think from memory, which does all their invoicing. They've got Eventbrite that does their event management. They've got some very, very specialized CRM software that deals with all of their highly sensitive information. They're using HubSpot for some marketing stuff and Arcflux for some other email marketing. And that's because different people in the business had different needs. So they've spun up different software and it does its job. I want SharePoint to manage file storage. So you've got all of these systems and if you look at it as a day to day operation, it all works well because the person who deals with events uses Eventbrite, bit of HubSpot. The person who does sort of the counseling side uses the specialized software and a bit of some other stuff. So it all works well. But then when you sit back and look at it as an organization and you think, well how can we delve into how we can make things more efficient? How can we save the charity some money? Because they're a charity, you know, they're reliant on, on funding for their income. So how can we save them some money or how can we give them an optimized version of what they've got to help them go out and get more funding and work out where to target the grants that they're going to go for or the companies who are the right profile for them to attack for, you know, for donations to work with. Attack, definitely attack. And Michelle won't mind me saying that, but that's the thing. When you look at it as an organization, have you step back and you go, well I can definitely help with sort of the machine learning around so the non sensitive data. And it's very important to keep that very sensitive counsellings tied and the crux of what they do separate. But if you look at the business data they've got, how can AI help when you go, well, that's some machine learning. It's very simple to be able to, for us to look at it and go, let's put this thing on top of it. It will start looking at business Personas of previous donations and all the rest of that just to help in this area or whatever. But that data is all over the place. So then actually what you look at first is an automation step, which is let's start bringing some of this data together. If we can start building models where eventbrite data and HubSpot data are talking to each other cleverly and we've got some good vision on what that data looks like, and then we can start injecting other data from the organization as well, all of a sudden, then we've got something that's quite powerful when we put an AI model over the top of it. And they're very much on that journey and it's an exciting time for them. It's an exciting time for us to be able to do that. But I think that's a great example of data readiness being the biggest challenge, one that we're very comfortable overcoming because we've done it so many times now.

But equally we're guilty of too. So when we look at our AI board, how we pull together the information for those individual board members who are also AIs.

Have we talked about that before?

I'm not sure how much we talked about it, but if we just sort of think about the. The marketing data. Yeah, there's information all over the Internet about how different channels are performing. Nobody looks at it, nobody knows how to do anything about it. Well, there probably are people, but we don't have them. So we're sort of in the blind. But we've then been able to build that gathers that information, that feeds that information to an agent that is specifically trained to come up with insights in responses to the data it's seeing and it's taken that mass of data that's largely unknown down to actions that we take every month. Makes it simple. And when you apply that to the whole AI board, then we end up with a company that's running far better than you and I can run it because there's so many things we don't know.

And that's kind of what we set out to do with the AI board is we wanted to have these AI assistants acting as Board members. And that sounds all fancy, but really what is it doing? It's just taking the data that it needs from inside our business from all of the systems that we use some external data from the Internet to talk about so it can see the external environment in a way and it's combining them together. So it's exactly the same as we just described. We're talking about it as an AI board member because we use it as a strategic level to help us, but it's the same process. It's getting data together in a clever way and then putting something over the top of it AI wise to make life easier. It's exactly the same thing.

Then all we've got to do is actually do the actions.

It tells us all they're doing is the problem.

Just get distracted by the customer projects.

Absolutely. A podcast.

So talk to me about the survey of all of the individuals within a company that you are working on.

Yeah, that was quite a cool project we're doing for, for a company down Hertfordshire. So we've, we were asked that the problem that they wanted to solve, let's start there, was that they needed to survey all of their internal people to score their external customers. So these are people who are bringing them business. They want to score them for things like how well they communicate as a company with them or how well how much business they bring them, how well they fill in specific forms at specific times, how good they are at administration. So they wanted all of this subjective information from their own people about the external partners. So it was a. We could have just sent out a form that surveyed them and then taken the results and put it in a PowerPoint presentation and that was it. That could have been the end of it. What they've wanted to do was something a bit more intelligent. So we've, we've created a, almost a dashboard for each person to log in and do this scoring and then we've taken the scores and instead of just presenting them in a pie chart, we've, we've done some data manipulation to increase the weight of one particular thing. If communication is more important at the time, for example, then we can, we've increased the weighting on that and decrease the waiting on administration and then built it up into be a report to give the sort of the end partner a score which we've then fed back up into their CRM system so that anyone who logs in can see that this, this company is a, you know, an 18 out of 20 because of X, Y and Z and detailed all the reasons now that's, that's all well and good, but we've then taken it one step further to enable, sort of empower the people within that business to approach the, the better scored customers and say, okay, we've got a bit of budget available to wine and die and take them out, you know, give them a day out or whatever. Let's take the 18s, 19s and 20s to say thank you. And that's a great thing for them to be able to do. And now they know statistically which companies are the ones to approach, then we've got the next bucket down, sort of the 17s, 16s, 15s who they can approach with a really good conversation to say, look, we, we really appreciate what you're doing, let's do some more of it. And if you could just improve in this area, that'd be great for us in our relationship and we could both do more business together. That's a great conversation. And then everyone below that, they can then approach if they want to, to say these are the areas we want to see improvement. Or they can say actually that that customer or that partner there, they're only scoring a 4 out of 20. We can see why do we even need them on our book of partners? Or can we take what they were doing and put it to someone else? Is it, you know, can we give it to someone who's already scoring an 18 or 19 rather than all of the hassle that comes with dealing with a 3 or 4?

And the great thing is if you look at those top performing customers or companies and you then establish a Persona for those.

Yeah.

And you then specifically target that Persona, you're effectively leveraging that survey information. At the end of the day, it's just a survey.

Yes.

But it's how you take that and convert that into something that can then be used to produce targeted information to specifically target new companies that you could work with that would perform at a similar level, then it gets really interesting.

It does. And this is where I really enjoyed the start of this project because we're nearly at completion now. But when we first had the conversation it was, oh, we need a form. And we think we might want to do some waiting around the answers, but we don't know how. So we've gone from that to actually a tool that's incredibly powerful for how they manage their partner relationships. And there, there may be appetite. They were talking about appetite for if they wanted to split the country geographically, for example, and have one partner deal with one area, one partner dedicated to another Great. If we get to the end of this process, we know where all these partners are, we might be able to get to a point where we can hopefully naturally associate those, those postcode areas with partners. And if not, like you said, you know, then the Persona to target. If you need a gap filling in a certain part of the country, you know what to look for in a partner because you know from the, from the data that what you're after.

Fantastic. So that's an example of a company that approached us with a specific problem that was very easy to define, not really that complicated to deliver. But the potential in helping that customer in the long term is far greater because of that data.

Absolutely.

When we see customers who come to us at different ends of the spectrum in terms of understanding of what they want to a sort of a finger in the air. Can you talk about how we help some customers establish the detail of what actually is they need before we even start working with them on an actual solution?

Yeah. So those sorts of customers that come to us and say, you know, we want to do something, we don't know what, but we know that we, there's things we can improve. So I think it goes back to what we were saying originally. We start by showing them what's possible. So we talk about the cube reports that we do for Cube, sorry the disrepair reports that we do for Cube or we talk about some of the automation that we're now running with little miracles or whatever the customer and the example is or some of the larger projects we're doing with, you know, with customers who are under NDA. So I can't talk about, but we can talk about the rough structure of what we're doing. And again it's opening people's eyes to what's possible and going and telling them the first thing we normally say is have you ever played with anything to do with AI? If the answer is no, then just go away and play his. Go and play with Gamma because if you can create a presentation within 11 minutes like I did the other week, then you'll understand how powerful this stuff is. It was a, I was doing a demonstration talk at a business networking club and it took me 11 minutes to put the deck together for my 20 minute talk. So go and play with that because once you realize how cool that is, you'll get the appetite for what's possible. Go and play with mid journey, log in, try and create, you know, something that you've, you're passionate about. Go and at least log into chat GPT and get it to answer some questions and then pay for the subscription and see how much better the GPT4 is. So pushing people away from us actually is what we tend to do first to go and play because once you've played then you'll come back with some ideas.

And occasionally we got a customer come to us and they've got a very, very specific idea and it's super detailed.

Yeah.

And it's potentially not necessarily a multi sprint, potentially a year to 18 months long development project. Really big packages of work. How have we worked with those customers to get to a point where we know exactly what's going to be delivered and how that's broken down into its component parts?

You see those are amazing because if you think how far AI technology's come in the 18 months since ish since ChatGPT became a thing. 18 months ish. So we were working with it for a long time before that. But since it became popular, if you think of the development since then, what we often find is we enter into a project that maybe is a year, 18 months in the life cycle. By the time we get halfway through there's new technology available that didn't exist or we didn't have access to because it was still a white paper at the start of the project. And we have quite a few customers now who are excited to work with us because instead of just finishing the project we'll go back and go, oh, this thing's now available. We reckon we could do it even better. Might not cost you any more because we're still on the development road. Do you want us to try it out? And that's always, I don't think we've ever had a no to that. So that's quite an exciting way of working for those longer projects. But it's the benefit of sort of those projects as well is that we've been around long enough to be able to see big companies and small companies, both from our own careers and with the customers that we've worked with. So we can bring ideas to the table from the start. Not just technology wise, but sort of actual operational ideas. Have you thought about this problem? If it's a manufacturing company, we know where the, we know where the general pitfalls are and the general strain points are. We've got another, another customer who I'm not going to name yet, but again they're a manufacturing company who want help with build lists. So they've got different products that they're going to build or different products they're already building and they struggle with managing the bill of materials for those products and how to manage that through. Now you've come from a manufacturing company, you've done fair few years in a manufacturing company in your career. We've got Ollie who's you know, head of delivery and he's done that sort of thing all his, all his career in, in that same company. So we can offer expertise in so many different career and so many different industries that it's, it's getting what's possible in terms of tech and also what other companies are doing or have done or should be doing in those industries.

And that particular project we're focusing on making sure the specification and the screens meet the customer requirements because the customer is very, very specific about. I want to see this data on this screen in this position and none of the other tools give me that ability. I want to be able to choose make versus buy for any of the individual components within a bit of materials and change it from one time I build it to the next time I build it.

Absolutely.

And that capability is difficult to find in off the shelf software. But making sure we understand exactly under which circumstances which things are going to be acceptable.

Yeah.

What those screens will look like will mean that when it comes to actually building the application there's going to be no doubt whatsoever because all of the heavy lifting will have been done at that specification phase. And it's, it's almost the antithesis of some of the development where we do where it's a super short sprint knock out a proof of concept iterate from there to. With this type of project the customer needs to sign off on exactly what they want. It's almost back to traditional software development except we're able to use the tools in order to accelerate that process and to potentially produce some of the interfaces automatically rather than having to do it all traditionally.

And that that bill, bill of materials customer I see I've talked about as the Pokemon model before because they all they sometimes they want the to use the raw material in its first form. Sometimes they need to use the evolved material and sometimes they need to use the next evolution and they might flip and change that for that one part because they might want the raw unproduced steel in this time but next time they've got manufacturing restrictions so they actually they need it worked already bought in so they can just do the final step and sometimes they want the full and fat Charizard instead because they Pokemon. Yeah, totally.

So I'm old but you're a kid.

You'Re not too old. For Poke. Are you too old for Pokemon?

Too old to care about Pokemon?

Okay, fair enough. No, that's a. But that's one of the problems is how do you, in manufacturing, how do you decide whether you want something raw, something slightly finished already or something finished bought in to produce the end, the end model? And that's one of the exciting challenges that we're working on with, with that guy, to give him a system that can do that.

Superb. And then we've got the other end of the spectrum where a customer says, I want to be able to analyze videos, I want to be able to produce the results of that quicker than anybody else in market.

Yeah.

Want to check for these things. Can you do that for us?

Definitely.

Technically fearless.

Technically fearless should be our tagline, I think. Yeah, absolutely. So we, again, we know because we know what's possible now and because we're so up till now we've had a finger on the pulse of what's coming. We enter into discussions some of the time knowing that the technology we need is coming and we're very honest about that with customers. You know, can I have a chat bot that does video and, you know, live video, so it looks like I'm actually chatting with someone, an avatar, live. And we think, well, no, not yet, but we know it's coming so we can start working on something to give you sort of a live interaction chatbot in text. And we know that the technology is going to be here within the next 12 months to be able to deliver you something along those lines. And that's where it's exciting, isn't it? So we are as absolutely right, technically fearless, because we are. We know what's coming down the line from, from some of the superb AI companies out there.

And I think the amount of people working in the AI space now.

Yeah.

Is increasing year on year. The amount of budget that's being spent on the development behind these capabilities, it's exponential.

It is.

So the capabilities that are coming available that are going to be created, they're in reach. And you can see how on any of these sort of, if you drew a line of a product roadmap, here's what we have now, here's what we had last week, here's what we're going to have in two weeks time. And then you see the frequency of these releases where just last week there was three significant releases and that's without even five, if you count this stuff from Amazon and Meta. In one week, five huge releases from five different companies, all with new capabilities.

And if you think about only eight weeks ago, we both ordered those Rabbit devices.

Where's my Rabbit?

Where's my Rabbit? So those effectively mobile phone personal devices. Because we just need to explain what Rabbit devices are before people go down the wrong road.

But they are about 4 inches wide. It's quite good.

They're about 4 inches wide, but they've, you know, they do a lot more. I was going to talk about vibrate features, but I'm not going to. So those Rabbit devices which were released at CES, CES, announced at CES, they had 10,000 units available and they all went within a day. A day? Something like that. And we were lucky. We. We got there fast and bought two, which I'm very excited for them to arrive. But that's been dwarfed already. Like just the conversation about that was so new and exciting because having a handheld device that had onboard LLM and.

On board large action model, I don't think that's been dwarfed. I just don't think people have played with it yet.

Do you think that's what it is? I think there's been so many releases of excitement since then that it's kind of just gone quiet. Maybe you're right, maybe it's delivery.

But the capability that's there is beyond what's in the current available models. That ability to perform a set of actions, to not need an app store to work regardless of effectively any operating system. You just tell it what you want to do and you can train it in there. I think they've called it the rabbit hole.

They have called the rabbit hole how.

To perform a series of actions. And then those sets of actions as a scenario, if you like, are available to everybody else.

Yeah.

So not needing to program it for your own needs, being programmed based on anybody training for anything. And that being available through a large action model. Yeah, I think that's going to be.

I think you might have turned my mind on that. You're right. Because the rabbits haven't been delivered yet. It's gone quiet. But I shouldn't assume that means it's been dwarfed. But there have been so many like you've said, announcements since then that it kind of. It gets almost fatiguing if you're not careful. Because there's so much change coming so rapidly. If you're not got your finger on the pulse, then you could be working on a project that's using technology that was cutting edge three months ago, that's now been replaced.

Yeah. And I think having the ability within those builds to swap out, okay, you don't need to use that one anymore. Use this and it's going to give you that extra ability. That's critical.

But that's where our dev team are beyond any others that I've ever met with this stuff because we don't just develop for what's out there now. We develop in a way that assumes that that stuff's going to be replaced quickly so we can swipe it out, swatch it, swap it out if we need to. And that's critical development work when we're talking about projects that have a lifecycle of more than about six, nine months.

And that ties in really nicely to the way we like to work with customers on an ongoing basis where there's effectively a profit share or recurring revenue for us and in exchange we continue adding in that latest capability. So where it's an ongoing relationship with a company where we've already built something that's breathtaking, we can continue to make sure that is as good as it gets all the time. Because we're being paid to be part of that journey with a customer. We're sharing in their profit and in their success. Love it.

I think that's a win or we know it's a win win because customers keep coming to us and talking about it where, you know, instead of charging somebody a big amount, fat amount to build some, an app or a software development for them, we charge a lot less up front. And then we, we sort of split, as you've just said, we split the, into a monthly, a monthly charge because we want to be able to keep offering them the best tech that's out there. And if they do that, then they know that they're going to get that software houses that charge, you know, x amount up front, build hand over and then charge you a, a service fee just to make sure that the thing doesn't go down. There's no appetite there to develop.

Oh, and I've spoken with so many business owners and that's the model that they use within their dev houses where it's, we do this particular delivery, customer comes to us, they say, oh, I want it to do this now. Well, that'll be that much. And those models are a struggle in the current time because if businesses are struggling to find the money, they might keep going with that technology that they already had and the dev houses start folding up because they don't have the work coming in. Whereas if you're sharing in the benefit that you're delivering to a customer, you're giving them these latest greatest capabilities. You're making sure that their business improves with the times rather than being left behind and you share in the success of that company. Makes that.

How much, how great of an incentive is it for us for our customers to succeed by doing it that way? If they succeed, we succeed. And that's why we tend to avoid those conversations where someone says, I want this thing building and I want to take it and have it ourselves. We'll go, oh, yeah, we'd rather not. Thanks.

There's a place for it, isn't there? But in some instances it's given us the ability to prove off a concept and then parts of that concept we've been able to leverage in other systems. But it's not as much fun handing over something and it disappearing. I'd far rather continue to be involved with a customer in the ongoing growth and use of that application.

Yeah.

Both because it appeals to me, I find it interesting. But also because it's profitable.

Yep. And it's like we said, it's a win win. It's good for them. It's good for us.

Awesome. So what's our perfect customer?

Perfect customer is one that can admit that they're not in a good state of whatever it may be, technological state or data state or whatever, and comes to us knowing that and is a happy to take advice on how to get to the next stage. There's nothing worse for us than sitting with a customer, like we said earlier, that says I want AI and having to persuade them that they're not ready. So a customer who knows where their faults are in terms of technology or data is open minded to solving those problems first and then kind of knows the problem they want to solve as they move into sort of the AI readiness area. I don't care how big or small a customer is in terms of any metric you want to give it. I don't care how long they've been in business. I don't care where they're based. I don't care.

I care where they're based a little bit. I used to say 30 miles of Peterborough apart. Unless they're changing the world or we like them. And I still kind of like that.

We tend not to market or go proactively towards customers outside a certain area. But I don't. If someone contacts us and says they're based in the U.S. absolutely not. We'll still speak to them.

That's what they meet the other requirements.

Oh yeah, absolutely. Absolutely. Probably should. Yeah. Okay. If they're based in embargoed countries. Maybe not. Let's just put a quick wrapper around that. But the only things I care about are what I've said. Plus, are we already doing work for their competitor? If we are, then we're not interested because we. Your phrase. We don't. We don't.

We're not selling guns to both sides.

To both sides. So that's. That's the biggest stopper for us is if we're already doing something with a competitor or someone who could be deemed as doing work in the same space. Yeah, different.

Not the direct. Same solution.

Absolutely. We don't care if they're the same industry, but if they're doing the same thing, then we won't touch them with a barge bowl. Yeah, I think that. Yeah, we're quite open to, I suppose, industry as well. We don't really care which industry they're in, as long as it's not something that's morally questionable. I wouldn't talk to a company who distributes pornography. I wouldn't speak to a weapons manufacturer. So, you know, anything that makes me morally uncomfortable, then we wouldn't touch. But we're in a great place because.

We'Re in a privacy.

We can say no and we can say go away, or we can politely say that it's not our, you know, not our wheelhouse.

Yeah. We did previously use. I'm sorry, we're at capacity.

And the scary thing about that is we're always at capacity.

Yeah, we just increased the capacity.

We have such a great amount of customers. We have a great development team and we've got site for more developers that we want to bring on board because of what we're doing. We attract the good people who want to do something interesting. So we're always at capacity. We just have to increase the capacity. But, yeah, you're right, capacities. That's going to be an interesting challenge for us over the next 12 months. Is going from where we are now to, I think, an order of magnitude growth and being able to deliver that.

I think an order of magnitude in a year might be painful, but doable, though.

That's what we're here for. Without AI board, we can do anything.

We just need to leverage it and find some people that we could show.

We've got Super Meg now.

Super Meg to the rescue.

Yeah, absolutely. We can do what we want.

Fantastic. Well, Chris, I think that's been a great session.

Good stuff.

Let's leave it there and we'll talk again soon.

Good to see you. Cheers.

Thank you.