AI in the Engine Room - How Is AI Impacting Business Operations?

Speaker A: And so there was an excellent opportunity for IKEA to go, right, that's eight and a half thousand salaries I can save. They went the other way. They said these people like our, uh, brand, they know how everything works. They know how, what, what the brand is, what the brand's about, what we do. And so they spent quite a lot of money training them to do kind of clientelling and design, interior design advisors. And it now generated 1.3 billion euros in revenue.

Speaker B: Hello and welcome to the marketing and AI podcast, the Map. I'm your host, Hal Kimmer. I'm thrilled to be with Nick Darby in person. Hello, Nick. How are you? Uh, all right. Slightly weird that, uh, you're slightly weird or just us sitting next to each other?

Speaker C: Both of those things are true.

Speaker B: Excellent. Well, today's episode we are going to be digging into the real engine room of businesses. Our, uh, colleagues in operations and logistics. And I'm thrilled to welcome Guy Meisel, he's operations director at Radley, to help us explore how AI is being adopted in his world. Guy, a very warm welcome.

Speaker A: Thank you. Thank you. Delighted to be here.

Speaker B: Now, I think it's right to say that to start with, your actual, your role is that you have probably the broadest remit of operations and it also includes IT and customer service. Is that right? I mean, absolutely.

Speaker A: So the fundamental for me is anything that goes wrong is my fault. Anything goes right is the commercial team,

Speaker B: uh, that old CEO adage of whose throat do I choke? And everyone just points at you?

Speaker A: Just easier for me, yes.

Speaker B: I'm well at. So I think, you know, I've seen some recent surveys that suggest that perhaps AI adoption is a little bit behind where it is in other areas in the world of logistics and supply chain, but that most logistics executives are very bullish, that it will significantly benefit their areas. I saw 90% in one study and about a third, uh, pinpoint AI based forecasting as the ultimate market disruptor. I don't know what. Does that ring true to you? Do you share that optimism or, uh, even already seeing those kind of things

Speaker A: coming through, I think there is a huge opportunity to benefit from the powers of AI. However, uh, there are still market areas where, whether it's the occasional imagination, whether it's the occasional flawed response that means that businesses can't hundred percent rely on it. I think it's very, very specific down to the business case. So where you are talking about. So obviously I'm talking about operations from retail, wholesale, e commerce, but on the grander scale of things, AI in Transport and logistics covers a whole kind of realm of things from something like dynamic routing, which is obviously a subject. The less miles you put on your truck, the less hanging around time you've got, the better off you are. And obviously people like Uber and Lyft are using dynamic routing as well. Yeah, freight and fleet management. Why don't we check on it before it breaks rather than at the point it breaks? So there's a really, really important there about proactivity in terms of maintenance, but also kind of uberization of freight. I don't know the numbers, but it's 30 to 40% of trucks travel empty in most places. So how can you fill them? How can you get that load to take you back to your original starting point? So these sorts of things obviously so much, uh, easier when driven by something that can amass the right levels of data. And AI does drive that, uh, output.

Speaker B: Is it an area? I mean, I tend to think that, particularly when you speak of agentic AI, that the success of that will very much depend on whether you have a robust and entirely logical human process for

Speaker C: the AI to build from.

Speaker B: And at the moment, the variances of. Oh, yeah, well, we cook that bit in the warehouse over there. I know it says whatever, baked beans, but actually that's where we put the Heinz soup. Just because of the way it was set up in SAP four years. Is that still a scenario? That's not uncommon and therefore a challenge to work through.

Speaker A: I think it's deeper than the process. Before you get to the process, you've got to get to the data.

Speaker B: Right.

Speaker A: So the actual informational layer within almost any organization, how clean is it? I may have worked for a company that couldn't spell black the same way. Yeah. So it could be blk, it could be blck, it could be a black. And until you've got a standardization and uh, a kind of rationalization of your data, you'll get answers with AI. Fantastic. Brilliant. But they'll still be wrong. You'll just get them faster.

Speaker C: Right, yeah. Is there also a, a cultural dimension to this as well? Because in, in marketing where we come from, we'll just go, ah, it's the new stuff. We're going to try this and we don't care about the data and we'll just see how it goes culturally within supply chain operations. Uh, you know, when you're selling five items from the warehouse and you need to know there's at least five of those things in the warehouse, is there a cultural barrier to adopting, experimenting, or

Speaker A: is that No, I think there is an interest in moving all these things forward. If you can reduce the amount of walking your staff do, they'll like you. Because if you say, right, yes, today's job is to walk 10 miles around the warehouse. But if I then review what's where and uh, what the demand coming in is, I can put the products closer to a single point. And now you walk three miles.

Speaker C: Yes. And then is it a lack of join up between the demand part of the business, the marketing part and the supply part? We've got high velocity things flying out the door. Put it near where the lorry backs into. Is that communication breaking down? Could AI help with that? Join organizationally?

Speaker A: Potentially. Again, I go back to. The more data you put into an AI, the better the output is. And when you're talking about demand forecasting, if marketing have made a sudden, oh, let's put a 10% off things on to a, uh, to a particular product that's going to move faster. If the warehouse doesn't know about it, the warehouse is going to be surprised. So there's a communication layer rather than a data layer on that one.

Speaker C: Yeah, I see that quite a lot. And the organ, which is why I was talking about culture and the organization. If you can get into demand harmonization so that the, across your whole business you understand what the sales are likely to be in the next period and everyone can adjust for that, then that's great. But that doesn't seem to happen in. I've seen it once or twice in organizations and they've kind of fluked upon it.

Speaker A: Yeah. And it is, uh, I mean we are dependent on our customers. Our customers will do something. I mean it depends on what sort of product group you're in. So if you're in tins of beans and tins of soup, that's a relatively flat line. And then it's cold, we'll sell more soup.

Speaker C: Yeah, yeah.

Speaker A: But there's not a Christmas peak. People going out and buying cans of soup for, uh, I want to gift you a can of soup.

Speaker C: Whereas welcome to our family.

Speaker A: Yeah. Handbags. Currently. Yes. We know there are seasonal peaks. We deal with wholesale as well as our retail business. So we know that before our retail business starts to spike, we need the product into our wholesale partners so they're ready for that same spike.

Speaker C: Yes.

Speaker A: We know there's a build and we know where Black Friday sits every year. Thank you America. And, and we know when Mothering Sunday is. Absolutely. That's uh, a, that's a great opportunity to sell handbags but the nuances for bigger brands who have hundreds of thousands of lines. Yes. Huge data, huge uh, AI probably do it better than a bunch of humans with really, really big labor intensive spreadsheets

Speaker C: and is that I don't want to lead the witness here. We work with a, with a high end clothing brand and we used AI to predict what the order levels were likely to be. And they had flagship stores, they had outlet stores, they had high street, they had franchise and they had partners and all manner. And it was particularly difficult in high end apparel to predict what the demand is going to be because it's a one off piece by nature. It was something new. And, and so we wrote a very, a brilliant AI solution to this. And the merchandiser said there's no way we're doing this. And there were two things that, that, that that changed the, the attitude there. One was we explained where the calculations had come from. So it wasn't a black box of here's a number, get on with it. And the second thing which was more important was we let the human override the, the judgment. So it was a recommendation rather than a set thing. And when we went back after six months that some of the figures had been overwritten and they became less accurate but net because they'd left some of the predictions in there. We got more accurate predictions for that. The dynamic of human involvement. And everyone says human in the loop. And I kind of squirm at that. It should be machine in the loop and the humans in charge just the other way around. But, but is there are there things in that area that you can do tips and tricks to, to start to drive that adoption of AI within, within operations.

Speaker A: But it is absolutely the cultural piece. I was talking to the finance director of a, of a3pl yesterday who um, is, is they're dipping their toes into AI in various areas and respects and they don't want it to be seen as an opportunity to get rid of people. And when you're looking at merchandising exactly the same thing. This thing is going to do my job. Therefore I will not have a job. So the doing or human doing things better is not necessarily the case. Now if the data is able to be fed into a very competent AI human input into that or human support of that is very important. But as you said, you actually got better results when the humans didn't intervene.

Speaker C: Yeah, but yeah, I think that in the AI world the conversation that's dominating at the moment is all about efficiency. How do we automate, how do we Replace people. How do we do that? Actually, I genuinely believe a better use of AI, or at least a, uh, balance is how can I use AI to grow if I'm doing one thing this year, how can I do 10 of those things this year? And we create top line growth rather than just because the biggest constraint on growth typically is an efficiency drive, because you hack away all of the resources and your ability to do stuff and the creativity and the judgment and the humanity that you just kind of shedding all of that, but you might be saving a few quid at the end of that. People are missing that point.

Speaker A: Absolutely. But the best example at the moment of the positive outcome of AI is IKEA.

Speaker C: Yeah, explain that.

Speaker A: 1 through 17,000 customer service agents in IKEA. So where's my bookcase? Which screw is missing? Can you please, et cetera, et cetera. I can't build it, all those sorts of things. So, uh, IKEA put in a chatbot called Billy after their brilliant bookcase, and it now deals with around about 47, 48% of all their inbound calls and interactions, which is great. And that's marvelous. And so there was an excellent opportunity for IKEA to go, right, that's eight and a half thousand salaries I can save. They went the other way. They said these people like our brand. They know how everything works. They know how, what, what the brand is, what the brand's about, what we do. And so they spent quite a lot of money training them to do kind of clientele and design, interior design advisors.

Speaker B: Right.

Speaker A: And it now generated 1.3 billion euros in revenue from those eight and a half thousand people.

Speaker B: Blimey, that's fascinating.

Speaker C: That is fantastic. What a great case study.

Speaker A: And it's over 3% of their revenue. That's a, I mean, it's a big number. Okay. We're talking at scale and we're talking about a company that's got the ability to do that level investment into its people, but at the same time, what a wonderful AI story that is.

Speaker B: Yeah, that's fascinating. It's the most robust version of actually if you can allow people to move away from the low value, time consuming, drudgery tasks into something that's really high value to your end customer. That's. That. That absolutely proves that idea.

Speaker C: It's customer understanding as well, isn't it?

Speaker B: It's.

Speaker C: Yeah, retailers, I mean, I'm going to make a sweeping judgment here for a change. Yeah, for a change. Retailers are horrendously lazy. So you look at something like, I need to increase the volume of, of, of sales in something they default to. We're just going to knock some money off it. There are so many different ways to grow volume, but over the last 30 years, we've always defaulted 99% of the time.

Speaker B: But every time we try something as well, we should get guy types done as well. Yeah, exactly. What that's done is it's trained people to wait till Black Friday, buy on discount all your Christmas presents and, um, what happens after that? Oh, you January sale, you know, it's. Yeah, yeah.

Speaker A: But the grown, some of the grownup retailers are going, no, I am not going to next. Doesn't really do Black Friday.

Speaker C: Yes, very. Yeah.

Speaker A: But once you kind of got the drug of look, huge sales spike. Even if it's a relatively low margin, you, it's very difficult to walk away from that huge sale because you'll be challenged on your like for likes. Yeah, but last year in November, we did £10 million. This year you're forecasting to £2 million.

Speaker C: Yeah, yeah. You've got the wrong metrics, you're driving the wrong behaviors. It's. It's a very old argument, isn't it? But yes, still, still goes around. No one's learned. No one's learned the truth of it.

Speaker A: Absolutely, absolutely. Absolutely. But I guess with AI, the other opportunities within that, uh, space, if you take somebody like Amazon. So we have to look at Amazon as, uh, a kind of leader in the field again, because of their size and scale, because they can afford to go. Actually, I'm going to trial this in one of my 872 warehouses. Other than. Yeah, I'm going to do it in this corner of my wire. So whether it's, whether it's robotics, whether it's, uh, I mean, the popular Amazon complaint is you used a huge box to ship. I saw a wonderful note, the sort of water tubes that you get for swimming.

Speaker C: Yes.

Speaker B: Yeah.

Speaker A: And a customer ordered hundred of those and each came in a separate box from Amazon. Slightly frustrated Amazon driver turned up with 300 cartons, all identical, and each one had got one rolled out. Pool float. Um, but AI supports there by optimizing the box size that you use by measuring the dimensions of the product. And it will. So a combination of kind of robotic arms, AI, uh, machine vision all helps to reduce the consumption of cardboard in which you wrap the box by looking at it and going, I can fit that in there. Obviously, when customers order two randomly different things, it is impossible for anything to cope with that. So you want a pole that's a Meter long and a teacup. Yeah. How do you get that into a box? So they will always take abuse for that sort of thing and retailers will always take abuse for that sort of thing because it is difficult sometimes to do that combination. But again, AI drives improvements in that sort of thing, which is, which is waste reduction. That's good. But at the same time, putting, putting stuff, uh, as I referenced earlier, putting stuff as close to where you need it as possible is critical. And of course somebody big like Amazon have a much better idea of what they're going to sell and when they're going to sell it. And obviously if you Click buy at 10 o' clock in the morning, five minutes later, the warehouse knows at some point today, before my last lorry leaves at 9 o' clock in the evening, I am going to have to pick that product. And then it's, oh, is there anything around there that I also need to pick at the same time? And so you're driving that productivity and that optimization by getting sight of those orders as quickly as possible. In the olden days, we sent a batch of orders to the warehouse once a day and you just wandered around and picked everything. Now you get those orders in a flow, uh, and, uh, it allows you to go, right, I'm not going to pick that now because it's a long way away. I'll leave that to the end of the day and hope I get another one that's also a long way away. And then we'll do them both together.

Speaker C: Right.

Speaker B: Uh, cunning guys. We also see an impact of AI on the ability to forecast with more accuracy than perhaps has been possible.

Speaker A: Yes, again, again we're back to that lovely chestnut of inputs. So somebody, the big retailers, the big retailers like Tesco have weather departments. Yes. To go, oh, what shall I send this week? Will it be umbrellas or will it be ice cream? And therefore that goes into the forecasting. The challenge is on the kind of, on the bigger scale, where does that input stop? So, um, do you want to take into account a local election, do you want to take into account the fact that there'll be more people going to the seaside because it's going to be sunny, therefore that shop needs more replenishment than the shop in the middle of the country that's not going to be near the seaside, all these elements. And that's where AI comes in again, because the more data you can put out, put into it, you are likely to get a better output from that. And we've seen some examples with, on, um, kind of really, really kind of enhanced forecasting from people like Unilever and Dan army, who really show kind of up to sort of 90% improvement on forecasting and forecasting accuracy by throwing AI at it. Obviously, again, companies that have got the scale to do that level of investment, but it's difficult for 100 million pound retailer or 50 million pound retailer to go, oh, yes, let's throw a million pounds at AI because all of a sudden you find you've spent a million pounds and can't necessarily turn that into cost savings.

Speaker B: Do you see, um, a future where that forecasting will become agentic? I'm thinking we had a guest on probably about this time last year, maybe a bit before, a guy called Gary McDonald, who was a very experienced retailer from Northern Ireland. And I remember the case study he gave at the end was literally, it came down the gut feel to a store manager who was being told to buy an extra couple of cases of Coca Cola. And he said, what the hell, I'm going to buy 10 or 12. And he shifted the lot. But it came actually down to the gut feel of the human to actually make sure the, uh, the, uh, opportunity was exploited to the max. Now you could flip that around. Say he could have still had 10 cases of coca Cola in his store him a year later that he never sold. But did you, you know, how far do you feel that that kind of instinct from experience is still going to be valuable? Or is that actually something that's a bit of a myth and we can just actually move to advanced artificial intelligence for planning?

Speaker A: So I'm thinking of all my merchandising colleagues, past and present, and I don't want to upset them too much, but again, Unilever got a 30% increase in sales in key markets from letting AI take over the forecasting. I think it was in ice cream. Wow, 30%. I'll take that all day long. And on a grander scale, if you look at somebody like Zara. So Zara is a whole different ball game the way they've done it, because AI is fundamental to the business. They have built the business round AI, which is obviously very different from a lot of us who. Yes, well, let's try a small agentic opportunity or something like that. So, I mean, Inditex, the parent company, everything Is run through AI so they keep their residual inventory below 1%. Now, it still seems relatively big number, but when you think of all the sale activity that you see, that is for other retailers to drive out residual inventory and then there will be a Summer sale that will be four weeks long. But if you're only holding 1% of that as residual engine you don't need to do a four week sale, a flash sale in a day and it's gone. So that has made a monstrous difference. I mean that's one of the reasons that they are, they are the poster child for retail and inventory management, buying management. But it's all based on a spine to the business built around AI. So absolutely. Forecasting in 98 different countries at the same time for the same products, 200 plus online markets and eight brands. But they all run that same core functionality that enables all of these aspects. So it's the forecasting, so it's the merchandising and it's everything else. So they are absolutely a leader in that field.

Speaker C: An interesting thing you said about Tesco that they have, they consider the other factors that we all know uh, uh, come into play like the weather. I remember being at Clark's and we all had to contribute our ideas for projects for the next, for the next year up and coming and they all get ranked and, and I put in one idea which was why don't you know weather has such an effect on uh our physical sales. Why don't we buy some airplanes and seed or disperse clouds. So we're in. It was uh, clearly a joke.

Speaker B: Yeah.

Speaker C: And they sat around for about five minutes to both. This is a great idea. It's the major determinants of our office. I am just joking.

Speaker A: Yes.

Speaker C: I also that the Unilever project was my project that we did the insights for it. Voice of the consumer. It started off as. And that's exactly the use that we put agentic AI you know clumsily and eight years ago, nine years ago. But we had individual agents. One would advise on weather conditions, one would look at tv, one would look at trends and social things. One would look you know and, and allowed. Well it gave us a huge advantage at the time but that kind of thing is, is commoditized. Now you could create one of those agents a day and it would allow the smaller players to level, uh, level the playing field with the bigger players because they would. You don't need a Tesco weather department now you just need an agent that you could build in half a day and it gives you. And it talks to the other agents and says okay well the net weekend outlook here for sales in physical stores in whatever looks and it is some guidance. But my view is that AgentIQ can become an advisor to that human judgment and accountability In a way that just wasn't available previously.

Speaker A: Absolutely. Uh, and I think equally to enhance the adoption, being shown the possibilities that it can provide to you as a merchandiser, as a retailer, as a business. That's where the kind of the starting point is. There is still a concern for those who get to use it. Will it take jobs away? Absolutely, there is that genuine concern. But as a business, and as lots of businesses, we have a lot of people sitting around doing spreadsheets. And that's where the opportunities lie. I didn't employ somebody to just do spreadsheets. That's not their job. Their job is to tell me whether this is right or wrong or we need to fly that instead of putting up a boat. But they spend their lives in spreadsheets. That's where we need to take it. Because I'd like them to be able to look into that data and make better decisions, rather than almost creating the data by V and X, looking up hundreds of different fields in hundreds of different tables that sometimes they'll get it wrong. Oh, look, all of a sudden the lead time on this product is 8 kilos. Okay. I don't understand that.

Speaker C: Yes, yeah, yeah.

Speaker A: Where you do get challenges. But it's absolutely. We employ merchandisers to look at the trends, to see what's going to happen in the market. So try and take them out of the spreadsheet.

Speaker C: We did a project with a European multi brand, multi market organization and they employed, uh, one person, a whole person. Her job was in the morning to download the trading figures from Amazon and then to look for patterns, trends, anomalies, opportunities, and it's oceans of data. So she would spend the whole day looking through the specialties, looking for kind of four or five actionable things. Two problems with that. One is humans are terrible at doing that kind of thing. And is this more important than that? Or, you know, have I missed something? It's just terrible. The bigger problem was it took her from nine in the morning until five o' clock at night to actually do the analysis. There was no time whatsoever to pull a lever in the business or make a change or do any action or anything. So she then came in the next morning at 9 o', clock, repeated the same. There was no action that was coming out of the. It was kind of busyness, but it wasn't outcomes.

Speaker A: And the challenge again for business is looking at that level of data and what is the important thing? Is it the outright sales number on a sku?

Speaker B: Yeah.

Speaker A: Is it the margin on that product? Because you might have Sold loads of them. But if you've only made threaten Hapney, no huge benefit. What's the highest margin product that you've sold? What are you trying to do with the stock that's in an Amazon? Are, uh, you trying to clear down a selection of products so that you can put new lines in? So there's a whole different bunch. That's where benefit comes in. He's looking at it from that perspective rather than just throwing numbers in the air and trying to work out what they mean.

Speaker C: Absolutely. That's the human dictating what the outcome should be. Machine go and do the donkey work and come back and find the five things that I should be doing today and then you can action. And that's very good. Yeah, yeah. I remember there was a. I hope I'm not betraying too much here. There was a CMO and we sat around in the, in the room and she said, um, we just need to double our sales. Or it was something ludicrous like a. I'm never one to think too small, but it was a. Just a ludicrous statement. I was like, it was easy. We'll just halve our prices for everything. And they were silent. And she went, yeah, but that would ruin the margin, wouldn't it?

Speaker A: That wasn't the question you asked me how to do, to create more money. Yeah, absolutely.

Speaker C: And, uh, it feels as though that is what an AI agent would do. Oh, okay. You've set me a task. I'm going to do this autonomously. I've got the guardrails of just fulfilling the objective and then suddenly your business collapses because you're selling things at a loss or, you know, whatever it is, it's. The human context needs to be in there.

Speaker A: Yeah.

Speaker B: I was fascinated to hear that you still acknowledge that spreadsheets form the absolute bedrock for the logistics and supply chain professional. Because the closest I had to collaborate with my colleagues in logistics is going back probably about 15 years when I was at L' Oreal and we'd had this very serendipitous and joyful moment where which of all consumer bodies had done a product test on a body cream and announced that the l' Oreal one was the most efficacious. And so suddenly there was this huge spike in demand. Everyone was scrambling to get this on posters. And then everyone turned and looked at the logistics team in the corner of the open plan office and said, have we got product for this? And of course, they were like, nothing. We haven't got nearly enough to meet demand. In the uk, so they had to scramble, go around all the warehouses in Europe. And I remember at the time just seeing them all there with all these spreadsheets. But I think then that since then I've had this, this situation where I've, I've had to try and use AI to replicate just budgeting spreadsheets for myself and being quite, quite surprised by the number of errors that the AI was coming. But obviously Nick is going to say, well, that's, you know, an idiot driving and it's a crap in, crap out. But nevertheless, I became very, very sensitive at that point that you have to super double check everything. At least I do. That you're asking the AI to do to make sure that the numbers do indeed add up. And I don't wonder if that experience is something you've had to make the hard way as well, or at least been aware of.

Speaker A: Yes, I mean, there are plenty of stories of AI, uh, having a moment. I was with a finance director yesterday who will remain nameless, and she said, well, I was using Gemini yesterday and I asked it that she'd identified a particular item in the final budget and so went to Gemini and said, uh, Gemini, can I capitalize this or can you capitalize it? Can I capitalize this? And it came back in all caps. So it was asked the right question. Yeah, it's 100% accurate response. How was that? It wasn't. It is. It is. It wasn't the way that the question was intended. And that's where the new skill set is, is. Okay, can I ask it the right questions so that I get the right responses?

Speaker C: Yeah.

Speaker A: However, there's the possibility of a sea change. Now, again, for the big players, Uber is the latest one to come out of the box. So in the last four months, uh, up to the end of April, they've spent their year's allowance budget of tokens on Claude.

Speaker B: Wow, that's interesting. I think maybe it was when we last spoke. I've definitely heard this issue coming up more than once, though, around you burnt through your tokens and now we've got a pause or, you know, that's where the money's going to get generated for.

Speaker A: Uh, absolutely. And what many of us haven't really clocked yet is there's a cost to each query. Sometimes if we operate within Copilot, it's kind of hidden because it's part of the license fee and we're not coding with it. We're just asking dumb questions or trying to build up relatively straightforward agents on customer Care and this sort of thing. But those using it for coding are ah, suddenly finding it's rather expensive, but are now challenging that they can't determine whether the four and a half billion they've spent on AI is actually generating a return. And that was from the COO of Uber. So we're talking about huge numbers being spent and the concern is if these companies are go to market at a trillion dollar valuation and you've got somebody who really using it and brilliant it's doing. I think 15 or 20% of the hot fixes on the Uber platform are direct AI. No human intervention or something's not doing what it's supposed to do. I've spotted it, I'll fix it. Fixes it. If they're going, you can't prove that there's a business case here for me to spend four and a half billion. That may slightly undermine the IPO situation. I'm, um, not quite sure.

Speaker B: Yeah, slightly.

Speaker A: It is an amazing tool. If you ask you the right questions and um, build agents in a sort of sympathetic way, you will benefit from it hugely. Is it the resolution? There's a lot of stuff on LinkedIn. I am a one person, 100 million pound company. Everything is done via. Yeah. Okay. Your customer service would be interesting to

Speaker C: witness and your bank account that you've entirely fictitiously made up just for clicks. But yes, let's hold our uh, skepticism about some of those.

Speaker A: Absolutely, absolutely. Apologize.

Speaker B: And our next guest is content creator. Exactly. Johnny entrepreneur with his 100 million pound business. He started from his bed sitting slough, age seven.

Speaker C: Yeah, yes.

Speaker A: Uh, yeah, absolutely. But there are other kind of different areas. So slightly more in the kind of marketing field, somebody like Adidas is supporting their design functions. So they've uploaded 150,000 shoe images into their bespoke LLM. Um, different angles, different shapes and then you can 3D print off the back of that and all of a sudden you can generate new ideas and design and build and really speed up that go to market process with the support of the design team, with the support marketing team and everything else. And it, it's things like that that will also sadly again enhance speed and speed to market. I think it was Speedo who went from idea to product to launch without creating any samples because everything was AI and 3D rendering and everything else. So there are companies now who are going, yeah, let's not bother with this creating samples. We know what the material is, that's fine, but the design and everything else just flows straight through. So there is a Lot of change in the retail design brand world and yeah, absolutely, a lot of it's going to be driven by AI.

Speaker B: That's fascinating because we're already living in a world of fast fashion, but it sounds like it's only going to get faster and faster in terms of.

Speaker C: I remember when we were at Clarks. Clarks. Have they, what, they got their samples done in China and obviously sleepy old Somerset is where they did all the designs and they were one of the first people to have a 3D printer kind of 10 years ago. And they would. Instead of having to ship the design to China, China made it. The samples came back six week delay, eight week delay, whatever. No, I don't quite like the eyelets or whatever. And it would go back and they had two 3D printers at both ends and. And I was amazed by this. And uh. But it didn't print it in kind of the 3D. The plastics that it can do. It was in. It was in clay. It was in not, uh, clay.

Speaker B: It was not resin, was it?

Speaker C: It was heavy resin, very heavy. And uh, I remember showing a very senior, uh, someone on the board of the UK company. I was like, this is just amazing.

Speaker B: And.

Speaker C: And this lady picked up and went, oh, no, it's a bit too heavy.

Speaker B: Yeah,

Speaker A: not hugely flexible.

Speaker B: Yes, exactly, yeah.

Speaker C: There's no holes for my foot to go.

Speaker B: His laces a bit.

Speaker C: Yeah, yeah, yeah. You have to get over the imagination gap sometimes. But yeah, the technology went well used. It shortened that from like eight weeks to four days or something. It was a remarkable innovation.

Speaker B: Guy, this has been absolutely fascinating, but as a seasoned forecaster, it would be remiss not to ask you how you see the next six months or year unfairly panning out as the signal goes down and we've lost him. But no, you can see you are still there.

Speaker A: Yeah, I think, I think there will be more adoption. I've been speaking to a couple of companies recently about almost a try before you buy solution. Because the issue is there are a lot of companies out there who just are, uh, not quite sure where to start. It's all lovely. And yes, if I could write a check for £100,000, I could get something really clever in AI. But until somebody stands in front of the FD or the CFO and goes, right, give me £100,000 and I will give you a reduction in residual stock, I will give you two days faster to market, I will give you reduction in air freight costs. What's the trade off and the difficulty for a lot of Businesses, I don't know. And you don't want to go into that conversation and go, two heads, because that's not right for any business. So the ability of a company to go, right, just give me a bunch of your data and I'll give you an answer. Ask, uh, me five questions about a pile of your data and I'll give you an answer. And then if you like that answer, let's talk about a, uh, commercial partnership or something like that. So you've got something concrete to share with a CFO or with a leadership group. And it might be in the supply chain space, it might be the forecasting space, it might be the retail space, merchandising, whatever. But you've got to find, for some people, it's very difficult to find the payback because, yes, we may knock 20%, 30% off our, uh, customer care team, but that's a third of a person. Well, what do I do? How do I do that?

Speaker B: Which bit we're losing

Speaker A: sitting down, have the legs. Um, so, yes, that's where I see the market opening up for smaller businesses, for bigger business. There's just going to carry on throwing huge sums of money at it. And I'm sure the IPOs will be massively oversubscribed. However, when you see, uh, a big global brand, for example, Starbucks, Go. We've spent huge sums of money on our, uh, AI automated counting solution to count stock in stores that went in mid-25, 2025, I think, and they pulled it last week.

Speaker B: Wow.

Speaker C: Wow. That's a big decision, isn't it?

Speaker A: Yeah, it's not very accurate.

Speaker C: It's not as accurate as the minimum wage person who just goes around with a clipboard, you're saying, and a scanner. Yeah.

Speaker A: And it was making multiple mistakes. It was struggling to identify one product versus another product. So it was actually overstating stock in some stores. So therefore we've got loads of that in that store. I won't replenish it.

Speaker B: Yeah. All right.

Speaker A: We run out of soy milk or whatever it is. So, yeah, nine months, the whole thing has been scrapped. Now, combination of poor execution and maybe poor tech.

Speaker B: Yeah, it's never one or the other underneath that. To your point right at the beginning of the show, you know, it's the quality of the data in is. But nevertheless. But that was one of your consultancy projects, was it?

Speaker C: No, no.

Speaker B: You to work with Starbucks, didn't you?

Speaker C: I did, I did. But in 2001.

Speaker B: All right, not guilty. Not guilty.

Speaker A: But. But, yeah, I mean, that's like m The Uber statement.

Speaker C: Yeah.

Speaker A: That's big.

Speaker B: Yeah, that, yeah, that's very big. And your point five minutes ago about burning through tokens and then suddenly realizing that you have to pay a lot more for usage. That's, that's a wake up call that every business would need sooner rather than later.

Speaker C: I've just seen it's not just Uber, it's other uh, major companies have woken up to the fact that, you know, they need to reel this in.

Speaker A: Yeah. What's my bill? Because.

Speaker B: Mhm.

Speaker A: To encourage the, the enthusiasm and use of AI. Let's get everybody a license. Yeah. And everybody's doing it. But some people are booking their holidays on it. And you consume a token in effect. Word.

Speaker B: Yeah.

Speaker A: And the output of that is when you do multiple level queries. So I'd like to buy a shirt and then the question will be well, what color of shirt? All right, so a blue shirt. But you pay to go back to the beginning every single time. So every time is going. It will start charging you with I want to buy a shirt. Um, all the way through that. So if you've done 10 queries.

Speaker B: Yeah.

Speaker A: And got to the bottom of it and it's a blue shirt with short sleeves and a button down pocket. And, and, and it's less than fiverr, then. Well, you just added a quid to it. Yeah.

Speaker C: Spend a five of finding it.

Speaker B: Yeah.

Speaker C: Brilliant Guy.

Speaker B: This has been a fascinating conversation. If people want to follow more of your. Your work or get in touch, what's the best way? Is it LinkedIn or.

Speaker A: So LinkedIn is the best way to get hold of me. I am guy misel on LinkedIn. So um, yeah, please feel free to contact me. I've always got an opinion.

Speaker C: You certainly brilliant opinions on LinkedIn. I always, I get excited whenever you. You posted today about tariffs. Yes.

Speaker A: Uh, and our orange friend.

Speaker C: Yes, yes it is, it is well worth following. It's brilliant.

Speaker A: Brilliant.

Speaker B: Thanks so much again Guy. Great to speak to you.

Speaker A: Appreciate it. J. Thanks a lot. Bye bye.