The Human Interface for Enterprise AI, with Ilann Hepworth at ShopAI [086]

Narrator: The Business of AI Podcast, exploring how businesses are using AI to build services and tools, transforming organizations and delighting consumers. Produced by UKAI and hosted by Tim Flagg, AI is our business.

Tim Flagg: Welcome to another episode of the Business of AI podcast and I'm delighted to be joined this morning by Ilan Hepworth from Shop AI. So, Ilan, welcome to the podcast.

Ilann Hepworth: Good morning, Tim. It's great to be here. Thanks for having me.

Tim Flagg: Yeah, well, great to have you here. And um, yeah, I'm looking forward to finding out more about Shop AI. But let's start off. Could you tell us a bit about your background? Um, how did you come to do what you're doing today? And then tell us a bit about what Shop AI does?

Ilann Hepworth: Yes, certainly. Shop AI established in 2024, but I've been working primarily in marketing services for the past 30 years. So I'm genuinely old enough to be pre Internet. Uh, and where analog technology started to really die out and the Internet of things started to take over websites and so forth. Um, and being in marketing services, it was a really blessed position to be in originally from a manufacturing point of view in print manufacturing, um, apprentice, trained as a photo lithographer and then moved into various aspects of marketing services. So without boring you with the details, for 30 years, being at the forefront of manufacturing technology, communication development, um, and either the right person in the right place at the right time or the wrong person in the right place at the right time, I've never really been able to figure that out, but it's certainly been educational and it's been fabulous to go on that journey as technology evolves. Back in 2024, um, managing director for a company and the spectra of AI, let's call it that, uh, shall we started to surface in the industry and I've known about it for a long time, um, specifically working with big corporate clients and understanding what best in class communication manufacturing looks like. AI seemed like a really misrepresented subject back in November 2024. So, um, my colleague, my co founder, Christoph Challand, based in San Francisco, who's ex Google, YouTube and Facebook, I contacted him back in 2024 and said, look Chris, this is my understanding of how A.I. is evolving. Um, whether it's generative or agentic or any other kind of manifestation in between. What's your take on that? And long story cut short, we decided that there was a different way we could approach business, that we could deploy AI that made it consumable for every Mary and John. Um, so that was the start of A journey, really? What are we going to do that's different? Why would we do it and why is it important? And validating that with Christophe, um, ultimately Chris said, look, I'm all in. Let's do this thing. Let's look at, ah, doing something that's not been done before with AI. Let's democratize the intelligence and let's bring it to where people want it. So that was in November 2024. And then essentially we created a business interface, business, um, intelligence interface across email, uh, web and video conferencing. Fast, um, forward to 2026, where we are today. Um, you know, we've gone from MVP to deployed product and a good range of customers in 12 months. So super excited about our journey and where we're going.

Tim Flagg: Amazing. Well, thank you for sharing your story. And um, you know, you say it's like 12 months ago, a couple of years ago. It's actually quite a long time in AI, isn't it? So. But kudos to you for having got everything. Yeah, it's impressive to call off the ground. So, um, tell us a bit more about where you're at at the moment with, uh, Shop AI in terms of who you're selling to and where you're seeing the most demand, uh, from those customers who want to get access to that democratized data.

Ilann Hepworth: Yeah, so that's a great question. And from my point of view, I wanted to really leverage what I've been closest to for the past 10 years, which is wholesale and independent convenience and that sector in particular, and, um, I guess the fragmentation of intelligence and data that exists within that sector. So prior to, I guess, launching Shop AI, we reached out to, um, our first customer, um, and spoke to a chap called Tom Gittins, who's managing director of the wholesale group, and sat down with Tom and said, look, Tom, what are the challenges that you've got as a business when it comes to siloed data, democratized intelligence, actionable insights, all them lovely words that people like to say. What's your core challenge, though? And Tom validated, really what I suspected, which was in isolation. He has a number of systems which are all excellent systems and they all work very, very well, but they are disconnected systems in many ways. That is, they require a username, password and a graphic user interf to extract intelligence from it. So sitting down with Tom and saying, well, look, how do we work on a proposition that combines all that, um, that doesn't require software installation or training or it if we can deliver something. Tom, that was agnostic in that regard, that didn't require installation or training, but did connect your data, how would that look for you? And if the model works for Tom, how many, and the wholesale group, how many other businesses out there have a similar dilemma? And I call it kind of 80, 20 rule. 20% of people use these systems, and 80% of people find it difficult to use them for various reasons. But if that's common for how many other people? And the truth is, it's kind of common for us all because our intelligence is trapped in silos. It's on, um, Excel sheets, it's on PowerPoint, it's on a Google Drive, it's on a Microsoft drive, it's on a CRM system, an ERP system. Every business, I believe, not just the sector that we work in, suffers from similar dilemmas. Um, and validating that, deploying it, proving the point and scaling it, is where we're at now and what we've done.

Tim Flagg: Yeah, it's really interesting to hear you talk about the importance of data. And I, for the businesses that I've worked in and with, I've seen that exact same challenge. Often you'll find the databases will be legacy databases that have kind of accrued over the years. There'll be mergers, or there'll be new divisions that have been brought in, and then you've got those legacy databases. And what you'll end up with is this huge mishmash of data that no one really has time to clean up. Occasionally, you know, if you're moving to, like a big new platform, like on this Cel Salesforce or something, then you'll go through the process. We're trying to clean up the data in order to make it valid. But, uh, I suppose in the age of AI, two things I wanted to get your thoughts on. One, how important do you think it is to, um, clean up that data before you start deploying any AI tools and solutions? And secondly, is AI maybe the solution itself? Um, I know there are tools out there, you've probably got some tools as well that can actually help businesses to go through and clean up and, uh, better structure that data.

Ilann Hepworth: Well, you're dead right, Tim. And there's actually three or four different, uh, I think, perspectives on that, um, or angle, should I say. So first and foremost, yes, your data needs to be in a format that can be understood by any system. Um, so having your data clean, having it ready to digest, to do something with, uh, is a cornerstone without shadow of doubt. But how you federate that data, uh, after it's been cleaned, who has access to it, at what point is equally important then the interrogation of that data and I guess this hallucinogenic angle that comes from AI conversations and whether these systems are hallucinating on the data that you might have clean, you might have federated it correctly and you might have deployed it. But you got to be real careful around the whole hallucination part of the proposition. Just because it's a 2 in your database, does it mean that the end user is going to receive a 2 or a 4 or a 6? What's the reasons for receiving a 2 or a 4 or a six? It's about the transparency of the answer as well as the quality of the data. As an individual, you know, certainly working in a business, you could say as a broad brush example, the business. A business. And uh, I'm just being very generic just to paint the picture, Tim, but a business has given me a tool called AI and they've said that I can extract data and get answers from it, of which I will act upon. So I do so and I will act on them and I will deploy them. But what governance is involved in that deployment? What governance is there in the delivery of that information, irrespective of whether the information is clean? I asked the question, it gave me an answer. So therefore I'm validating in saying that it's the truth. But where's the audit trail to that answer? Where's the audit trail to that data that validates that it is true? I think it's a multifaceted answer to that question. It's not just about I've got clean data, therefore I can. I think it's a lot about the governance of delivering that, federating that, uh, auditing that I'm being able to prove it.

Tim Flagg: Yeah. Uh, that takes us on to another sort of area of the internal liability, uh, for that, uh, data as well. Because within any organization there's normally some sort of governance framework that says if something bad happens in the factory or in the office, there are rules and regulations and there is recourse. And this is the person or the entity who's responsible. I suppose the challenge with AI as it's sweeping into organizations is where does that, uh, liability sit? And you know, as you're sort of saying it's, it's, there's a sort of a liability there around data. Ah, the, the ownership of the data, the decision making on the data, the providence, but also the quality, maybe something about the ethics of that as well. So, um, it's kind Of I suppose it's balancing the need to make sure those guardrails are in place with also what you were talking about at the beginning in terms of democratizing data. Because we want people in the organization to have access to, to all these tools to do great stuff, but we also need to give them an understanding of the governance and the ethics of using that data as well.

Ilann Hepworth: Yeah, for sure. And like any good business will have rigid policies based on data, uh, um, privacy deployment, usage and so forth. They're no different I guess, when you deploy an AI system. But what I'm really trying to I, um, guess explain is the governance around using the data is the governance, but the real time usage of that data. How is that? Are we sure that that's correct? Can we audit that in real time? Are we acting on the right information at the right time? Are, uh, the right people acting on it? So it's not the whys and the wants and the do's and the don'ts and the metaphorical sledgehamme hitting an employee over the head and saying this is the policy abide to it. What I'm saying is in real time, you're confident that those technologies are actually serving the person at the right time with the right information. How do you check if that information is valid and responsible? And if somebody does something based on that information, who's, where's the ownership lie in that? Because the AI has said this is the right information. I've, uh, followed the standard company IT policies, but it's not about the governance of me using it. It's about is this data the right data? And therefore I acted upon it. Was it a good or bad decision? So there needs to be a real time audit trail for this stuff. And two plus two equals four that you can audit. But it's never given in numerical format, is it? It's given in language interactive format. You ask a question and it responds back in a narrative. And there's a lot of work I think, that's going on around as well as data. How is AI communicating and how is it affecting us socially? And how do we use those two things together to deploy professionally or personally on the intelligence that it's given us? How do we make that actionable, safe, secure, authentic, um, all those things. It's a big melting pot really Tim.

Tim Flagg: And so how do you do that? Because I think I kind of understand a little bit more now that actually the challenge that businesses have is about getting the right data to the right person to make the right decisions in A way which is efficient and building that productivity. Um, and you've talked about that and that sort of real time audit. So how do you actually use the technology within an organization to be able to enable the right data to get to the right person and doing those kind of audits?

Ilann Hepworth: Um, well, let's take a step backwards before we kind of, you know, say how shop AI does it. Um, I, I think it's really interesting that you as an individual, when I say you, you know, we, everybody, um, from a technology point of view, somebody usually sends me an email, let's say, and hi, Elan xyz, can you just do this? I really need help with that. Uh, the communication comes in via email. Then I use all the tools at my disposal and the systems, etcetera, to extract the information, to then respond back in email. So I moved from one work stream, which is email, one train of thought which we all work in predominantly. And then I go to different portals for information. Each one of those portals is a graphic user interface. Uh, and this is another thing that, uh, I want to challenge is that as an individual system I need training, I need access and I need authenticating to use that system. And then I will extract data from that system, multiple systems compile that and deliver it back in an answer to whoever's asked me for, um, the question, so quite simply, why do we do that? If intelligence, if AI is so clever, why can't I just ask it in my own words, in my own conversational style, in my own sentence, in the work stream that I'm um, in, which is email, why can't I say, dear, uh, AI, I need this, this and this. Can you go away, sort it out, thank you very much and forget about goes into the other systems, it navigates the graphic user interfaces, it gets to the data and it pulls that data back to me in a way that I understand it in a language I understand? I think that's where technology should meet humans. We shouldn't have to go find the answers, we shouldn't have to leave our work streams. We should be able to exist where it's convenient for us. And the convenience at the minute, like Olympia, is email, text message or any of those kind of real time things that we're doing. So if I can communicate with data in my own language, in my own way, using natural language processing in sentences, and I can ask, well, I need these numbers for next week, so I need to report to Jeff and Jeff, I'm going to be in a load of trouble. Please help and you type that as I've just said it, and something understands the context of that and retrieves the information relative to it. That's how it should be. It shouldn't be a series of education, dropdown, filtering, extracting, downloading, recombining, redeploying. So I think it's more about getting AI to meet you where you need it to be, not where it wants you to be. Not visiting a. Another app or another download or another install or another widget connected to something.

Tim Flagg: Yeah, no, I love that vision of a sort of human interface rather than a sort of machine interface. And, you know, you're leading to some of the tools that people learn. There's the business intelligence, business visualization tools. They do take a long time. Even Excel. Right. I mean, there's a. That's a sort of the standard that everyone uses. Um, but that takes time to learn. So, I mean, do you foresee a world where people won't have to use those products at all, that they can just go straight to this sort of human, um, agentic interface almost where they're just having that come in in the morning, have a conversation, and then the agents go and take care of everything and come back and talk to you, tell you the results in your own language? Is that sort of how you view it?

Ilann Hepworth: Um, yeah. So agentic workflows, agents, per se, um, you know, and executing workflows, they're fantastic tools. And they're like dominoes. You set them up, they perform a task and they save a lot of time. But at no point do those agentic workflows reset the dominoes or recombine the dominoes, or change the flow of the dominoes in order to adapt to the request, that's not the initial request that created the workflow in the first place. So I do believe, and it is fact, that, you know, you can create workflows and the tons of technologies out there, that this is a process that takes us two hours. Let's create this agent and it'll take us two minutes. Wicked. You know, that's great. Um, and they work very, very well. But that's not really what I'm talking about. What I'm talking about is imagine four domino lines which all intersect and then ultimately crescendo at the end with the reveal of an Eiffel Tower. You know, if I'm speaking in kind of visual metaphors, uh, if I ask to see the Eiffel Tower at the beginning of the day and I push the agents and the agents go away and do Everything and it takes three minutes and it comes back. It's an agentic workflow. There's no documented evidence of how it got to show me the Eiffel Tower. And one day I want to come in, I don't want to see the Eiffel Tower, I want to see the Empire States Building. So how do I use technology, you know, to get that information? It tells me how it got there. Uh, so in answer to your question, do I believe that the future of anything is natural language, conversation, education, information, and doing that fluently in any language, in any dialect, with any nuance of a up Siddhi meduk, you know, any kind of those little things that exist in language, then yeah, I do think the future of technology should encompass that. I think we all should be able to communicate with an interface of some description, speak to it naturally like we would speak to our spouses, our parents, our children, and it come back with educated information that is that you can iterate on Tim. So you get the first response, but then you get multiple responses and any iteration from those responses and it's all done at a place and time that requires me to do nothing other than talk or feel or think. You know, it's um. I think the graphic user interfaces are fabulous things. I really do spent a lifetime working with them. But in a world of tomorrow, if what you're asking me is do I think they're going to exist forever? No, I think in a world of tomorrow things like wearables and holographic heads up displays and real time tracking will be the future of interactions between computers and humans, moving more towards androids. But I don't want to freak everybody out and get into that conversation.

Tim Flagg: Yeah, yeah. Still a few years away. Um, well, I love that we got ayokmed in that last interaction. I m haven't heard that phrase in a few years. But that's a proper Nottingham phrase, isn't is, but.

Ilann Hepworth: And um, um, I use it purposely because if we're asking for intelligent information, the intelligence that drives the delivery of that information should be tolerable. Of the things that I'm not very good at, like speaking, like using language, those things are not important. If you have to use a graphic user interface, you're dropping and dragging and you're scrolling. But if you're in this world of tomorrow that you're referencing Tim, well, it needs to understand that Aupme duck is a really nice welcoming hello there. How are you doing? Um, there's a thousand of those things, isn't there? It's not just a Nottinghamshire thing, but technology should be tolerable of natural language is my point.

Tim Flagg: Yeah, well, I think it's that point you made at the beginning as well, around the technology coming to you rather than you to go to the technology that makes so much sense. And maybe it's because technology hasn't had the dynamic, ah, agility for one of a better way of putting it to actually able to adapt to you before. So that's why we had things like spreadsheets and erp, because that's just the way that it was able to be made for the masses. But now with, um, the systems we have in place, the AI powered tools, the whole thing can be personalized around you. So we're often looking at how we can use AI within, um, learning and education to personalize education around your learning style. I suppose there's an analogy here with what you're talking about because you're actually using it to uh, be personalized and very tailored to the individual. Regardless of what other language, dialects, phrases, background learning style, all of those things, it becomes very personalized.

Ilann Hepworth: It does. And also, you know, you and I are knee deep in technology and AI and you know, we, we understand it deeply. Um, but there is a whole raft of the population out there that when you say the words AI, it goes completely over their heads. You know, they might think of how they can use it to create a story, a poem, a video, an image or, you know, or anything in between, but they won't necessarily think, well, how does this help me work in business? How does it translate to what I might need it for in my working life? So I think we have to be careful there because if you look at it from that perspective and you go into a business with another tool, the first thing that's going to happen is 80% of people are going to drop their heads and say, here we go again, here's a new system, here's a new thing that's going to make us all great. That's going to. And that narrative has been done to death, haven't it, by a thousand people. So why is that different? When somebody then says we're going to use AI, the knee jerk reaction is, well, that's me out of a job. That's kind of the narrative of the thought process. Should I say that? Most people come to it instantly and I don't think that's true, I don't think that's true at all. So I do think AI could be really useful in business. But you've Got to meet it where your people are. Ah. You know, and that's why I think email in particular is a really interesting channel to AI, because everybody knows how to write an email. Everybody gets asked via email to do something and everybody responds an email once they've done it, nine out of 10 times, unless it's a personal conversation. So why can't technology, as an example, just meet me where I am in email. You know, I get, I get a request in, I forward it on, I put my own little blurb on it. Hey, up me duck. I really need some help on this. Um, sort it out for me tomorrow, will you? Tar. You know, done. Off it goes. Send. You know, that that should be interpreted by technology based on who I am, where I work, what I do, what I expect, and then the deliverables from that and it should give it to me in a way that I can understand. And that's, that's, I think that's the important thing here, Tim. It's how do you deliver that technology? How do you federate the information in a way that the end users can get it in English, in any other language, in a data and style format that suits them?

Tim Flagg: Is that the solution which you're providing now for your customers? You're starting to provide this, um, email product service, um, that you just described.

Ilann Hepworth: Yeah, 100%. That's exactly what we do. Um, so I believe that it's the only proposition in the marketplace that does it in the way that we do it. So, yeah, smart mail. I have three pillars, really, which I think are important to deliver information that's in the office, out of the office and office to office. So I think if we in the office, I think emails are, um, a really great tool. Like any other tool, we get requests, we can craft questions, we can ask for help. Um, and from our point of view, we wanted to be able to do that in email, but without having the nuances of installation. Do we need another email client? Do we need to install something locally? Well, in our world, you don't. There is no another email client. There's no widget to install, it's simply an email address. So it's intelligenceopai.uk um, it's prefixed@shopai.uk but essentially, as an individual, I can email that address, I can craft a question, it will go off into the big wide world and it will come back with the answer relative to the question that I've asked. Um, that's a really easy way to get people used to working with AI. If you can write a question, you know your job, you know what you need to get, you know what you need to do. If I'm in something that's familiar. Email, Outlook, Google Mail, Apple Mail, any of those. And we're not asking people to change their habits of creating things well, all they are actually doing is just sending an email to another address. And that address is business centric. We provide this white label SaaS model which means that they don't even know they're leaving. The business insights@your business.uk, would be a good example. And they're just sending an email and they're getting a response. But the, you know, we talked about hallucination, Tim, didn't we? And I think it's really important to pick up on that because one of the things I quite often get confronted with is, you know, large language models and you know how they can do very, very similar things? Well, they can, you know, they do, they're excellent and they get better every six weeks. You know, um, the fabulous. But do you not think it's a bit like chasing the dragon, so to speak? Whereas let's go for the latest 4.7 or the 5.5 or the 3.2 or the 8.6. Chasing the latest and greatest is fabulous if you want to utilize it for high end graphics. But the reality of the language models is that a version of those things two years ago would probably be more than enough to answer a question you have on email based on any business data that you might have access to. Using the power of an LLM to do that language piece I think is really important to make it relatable natural language processing, the text that it sends you back, the narrative that it provides, I think that's really interesting and it's important, but we shouldn't and we cannot rely on them not to hallucinate, we shouldn't and we cannot rely on them to not run out of context. So if I'm a small, medium sized business and I've got databases that have got 60 million rows of data and I use an LLM and I point that LLM through some magic technology, through some wonderful configuration to all that data, what about the cost of running it, Tim? What about the token costs of those things as well as the hallucination costs? Before it gives you an answer, it's got to search its data and then it's going to give you that answer back. There's a whole cost base that's involved in that. Is a large language model the right thing to use interrogating massive amounts of data? Probably not. What we've chosen to do is look at it completely in a very, very different way. First of all, you've got to decouple the costs as much as you can with large language models. Then you've got to look at how you recategorize that data. So in our technology, when, when somebody sends an email and they ask for help from AI, it's ephemeral is the AI. It only exists in the thread of which they create. So they send a subject, they send a request, and when it leaves their outbox, an AI gets attached to the email, to the singular email, not to the outbox itself, not to the computer, to the email itself. We embed a little AI in that email and we give it a toolbox and we point that at the data cortex of the business. Those two things have a conversation with each other. The little AI that's embedded in the email says, ilan wants me, uh, to answer a question that two plus two is four. Could you tell me if Elan can have that information? Could you tell me if we've got two plus two and can you validate that the answer is four, please? The main AI that's in charge of the cortex gives the little AI the numbers, the database numbers, and says, yes, Elan can have this, these are the right numbers. Go away and give it back to him. Then the little AI in the email crafts, it's got tools, it crafts the answer on the fly, ephemerally. Then it delivers that back to me as an answer, but it's been fact checked before, it's actually come back to me. We know that it's powerful and we know that it's right because it doesn't use large language models to get the answer. It uses SQL databases to get to the answer. Does that make sense?

Tim Flagg: I think so, yeah. I just wanted to maybe, uh, get a bit deeper into some of the technical, um, understanding of how you're actually doing this, because I think some of our audience, um, will be fascinated by what you're doing there with the emails going back and forth to, to get these answers, but break it down into, um, how you're actually doing that. It sounds like you're using, um, sort of, you're building a platform that's quite agnostic to models, um, but under the hood you're able to then use the best type of models which are out there. Be interested to know whether you're using open source or whether you have some preferred, um, sort of tools. That you're using and then the bit you said where you're actually running the queries, but back and forth between the um, user interface and the database sounded a bit like you're using some sort of federated data platform. So again, be interested to hear how that works. I know, you know, go into as much detail as you feel confident.

Ilann Hepworth: Yeah, no, absolutely. Well, all of it. Um, so where the jam in the sandwich really, you know, so if the bottom layer of the sandwich is the person and the top layer of the sandwich is the data, we sit in between that. Um, so the inquiry that goes out of email, that goes to a designated address, rules based rbac, we know the people, we know the department and we know what kind of data and structure that they would have access to. That's a profile of a person. Like we all have profiles within our different departments. So that question comes out, we know it's coming from that person. The ephemeral nature of this is the USP really Tim, because we work within the email itself, within the actual embedding of the email. And it goes off to a server. The server's got connections to all the data. Uh, the data is um, extracted from the SQL database and given to an agent within the email that has a toolbox of SQL and LUA and various other little tools that it has a little agent that sits within the email. It's only crafting numbers, it's not crafting sentences. It's been asked a mathematical equation, it's using SQL A versus B. And then it's coming up with the answer that C, then it's validating that answer and then it's crafting the response back to the human. And that's the simplest way that I can describe that to you. And every nuance of, of language or graph or chart or document or anything else that goes back in that response from the AI to the human is included as it comes back to the human. So the reliance on large language models, Tim, for us doesn't really matter because that's not doing the heavy lifting in our world. Um, a large language model, a version three of a large language model could communicate with me quite naturally. Two years ago, that's all I wanted to do in email is just structure the sentences for me, not the data.

Tim Flagg: Right, okay, that's interesting. So now I can get it. So actually the bit you're accelerating or augmenting is the processing of the business data into natural language. Um, that's your unique area. And then the LLMs are uh, being used, but Only in so much as you're then making um, the data and that the output, the information which has been gathered more understandable and more readable in a, in an email format. So I think that's clearer now. Um, just, I'm just realizing in terms of time we need to start to kind of wrap things up. So let's get your thoughts on what are the things that you're most excited about as you sort of look forward over the next year. Uh, so what are the technologies or the changes that you see happening which you're most excited about?

Ilann Hepworth: Um, multimodal technologies I think are really exciting. Uh, I think vision based technologies and that collaboration between not what we can say, not what we can type, but what we can see, the environment that we're in and the inclusion of intelligence within uh, that instance wearables, glasses, vision based technologies, the embedding of, of AI into those things I think is a fascinating subject as we move forward. Um, where we are today, being able to democratize intelligence through email, a web app and a video conferencing call from a shop AI point of view using Meta SQL. That's the secret sauce, if you like, that we've created, that we own. Um, is definitely best in class today, but best in class tomorrow I think is a whole different ballgame. It's how do I use AI in real time based on the environment, based on the conditions, based on the individual dilemma. That's a combination of environment and data. What's to stop us from providing real time health and safety analysis on site with glasses, with advice, actionable insights and recommendations whilst I'm doing it, you know what, what's to stop us from asking for a quote from a plumber, you know, or a builder or an architect who can wander around with glasses on and build all those quotes, all those dependencies, all those answers in real time just by what they can see, not go away and create, create something, create, report and deliver it back. So I think the, you know, the hybrid of um, this multimodal visual AI piece is certainly coming out as an alarming pace. So whilst we're all still trying to get our heads around how to democratize intelligence and provide it in a non scary way and make it human first, you know, there's already a freight train coming at us where it's, you know, it exceeds that and then some.

Tim Flagg: Absolutely, yeah. Well you've touched upon a couple of uh, kind of really big themes there that we'll be tracking over the next few uh, months and years. Um, amazing. So Last question then, Elan. How can we find out more about Shop AI and uh, follow what Shop AI is up to?

Ilann Hepworth: ShopAI website. Um, drop onto the website. There's a ton of tools on there, uh, that helps help you understand is AI right for my business? Do I want a human first approach to technology? Do I want an automated approach to technology? Do I want to reduce headcount? Do I want to retain people? Do I want to combine that intelligence and do I want to move forward? There's a whole bunch of tools on there that help you make the right decision based on how you want to deploy AI as a business. Um, so yeah, shopai website, have a look on there. Uh, failing that, there's not very many Elan's in the world. I L a double N. You know, you can find me on LinkedIn, drop me a note and I'm more than happy to sit down and talk to anybody about, um, kind of what the art of the possible is.

Tim Flagg: Amazing. Well, thank you Elan. It's been great to talk to you about that interaction of bringing data to life and allowing humans really to get the uh, data brought to them and in their language and in a way that they can use so that it really empowers those humans to be able to. And I think that really is a very valuable use of technology is to empower humans rather than taking away. So I like the way you framed that. Um, and it's fascinating to hear also about the way you've been approaching, um, working with different organizations to empower them with um, these email based systems. So thank you for talking to us today and for sharing more about Shop AI.

Ilann Hepworth: Thank you very much for your time Tim. It's been great to talk to you.

Narrator: Don't miss the next episode. Subscribe now. UKAI is the trade association for businesses across the uk, tech and non tech, large and small. AI is our business. Find out more at UKAI Co uk.