Why AI is like a highly motivated intern

Speaker A: Hello and welcome to the Digital Lighthouse where we get inspiration from tech leaders to help us navigate the exciting and ever evolving world of digital transformation. I'm Zoe Cunningham. We believe that meaningful conversations can illuminate the path forward, helping us harness the power of technology for innovation, scalability and sustainability. In this episode, I'm delighted to introduce Steve Burrows. Steve is a global product delivery and technology leader. His current time title is AI Product Operations and Delivery Director. With a uh, 20 plus year career spanning Europe, the Middle East, Asia and the UK Blending consulting, product development and large scale digital transformation, Steve has delivered digital projects and services for major organizations including Bloomberg, Sony Orange and Emirates Airline. Launching global digital platforms, leading cross functional teams and driving complex mobile web and AI initiatives. His career is defined by a passion for emerging technologies from the early days of mobile and web applications through the uh, blockchain boom and now in artificial intelligence. So in this episode I'm going to be chatting with Steve about how to capitalize on the potential value of emerging technologies within larger and complex environments. Uh, we'll talk about what AI can and can't deliver and uh, get his thoughts on how roles, skills and jobs are going to transform in the future. So Steve, welcome to the Digital Lighthouse.

Speaker B: Thank you very much Zoe, thanks for inviting me on. It's great to be here.

Speaker A: Could you start by maybe telling us a bit about your career history and the journey you've been on, I guess to get to where you are today?

Speaker B: Sure, yeah. I mean you've given a good intro there in terms of uh, a little bit about my background but I guess I sort of summarize it as combination of three complementary kind of skill sets. So there is a sort of problem solving, consulting side of things. So I went straight out of uni into that space and spent a few years there and then since then over the last 20 plus years it's been product development. So owning actual digital products and services and doing the whole product strategy and bringing them to market and managing them in life. And then the third part is the actual delivery stuff. So I'm um, very much on the agile side of project delivery. So it's those kind of combinations of skills that's got me into this world as it is. Yeah, I've always been interested in technology and uh, yeah, I've kind of found myself at ah, the, on each of the technology waves over the last few years. So like you mentioned in the intro, um, just as applications and tablets and stuff started emerging, I was, I was helping companies build those and deliver those to market and then through the sort of web technology, boom, broadband, taking off and blockchain. Now AI. So yeah, I've always been passionate about emerging tech. Um, yeah, AI at the moment seems to be the logical home for me.

Speaker A: I think you've got this great mix of skills as well because actually it's one thing to know about the tech and about the kind of cutting edge or research tech, but actually once you get into a business environment, it's how that technology is applied is what's important. And the product design is much more about how the humans are going to use it than necessarily about exactly what the tech does.

Speaker B: 100%, yes. So I mean I'm not actually an engineer or a technologist by background, so I'm um, comfortable in those conversations. But they can't ask me to deliver a line of code. It would be a nightmare. But yeah, you're 100% right. It's not the technology that should be the focus, it's the problem to be solved. And that's where I think my career has been, as it has, because it's always been what is the problem we're trying to solve? Uh, and when I'm working with product delivery teams and engineering teams, they get quite annoyed with me sometimes because that is my mantra. It's what is the problem we're trying to solve. Trying to solve. So if someone's going off down a rabbit hole, of course, amazing technology or this incredible idea, my job is to bring them back down to what is the problem we're solving, who we're solving it for. So the technology has been different throughout those 20 plus years, but the approach has largely been the same.

Speaker A: Right. And in your kind of current industry, where have you seen the biggest impact of AI?

Speaker B: I guess my, my industry at the moment is kind of that consulting, product delivery side of things. So I get to work with a lot of different companies in different um, sectors and different spaces. So it's sort of the impact of AI. I don't necessarily. We've seen the M full impact of IT yet and the results have been mixed from my view. And the reason for that is largely because the companies are taking a different philosophical approach to implementing it. So you do have some that are, uh, some senior leaders read a blog about AI and how it's going to take over the world and how it's going to revolutionize their P and L. We need AI. Typically those implementations are not the ones I'm going in to help deliver, the ones I'm trying to fix in large way because they haven't had the impact that people expect and that's largely because of misaligned expectations as to what gen can actually do today. Um, the impacts where I have seen success is when people have got the right mentality and the right expectations from day one. So it's not going in with that. Uh, this is going to revolutionize my business completely. I can operate it with robots now and I don't have to pay humans. The successful ones are ones that have actually looked at how they work and mapped out a workflow and then thoughtfully plugged in AI into sections of that workflow where it's where it's strong and kept humans where it's not strong. So the impact it gets typically around efficiencies and time saving at the moment with generative AI and agentic AI. So that's I guess primarily focusing on the stuff that humans don't like doing but have to do. So the drudge work, if you like the repetitive time consuming stuff. So low level checking, documents, data entry, deep research, that kind of thing. So AI being used to get humans started on their workflows, but then humans coming in and actually finishing it and putting that I guess quality layer at the top of it. That's where I've seen the most impact. Right.

Speaker A: And the judgment and oversight and um, so I suppose, uh, like a question I'd kind of like to ask as a follow up is let's say I am a CEO and I've read a blog post about AI and I am excited about it and I want to make sure that I'm taking advantage of it. What should I be doing before I go out to the market and try and find someone to revolutionize my business with AI? What are the steps I should be doing internally or what, what do I need to be mapping out or planning out?

Speaker B: I mean at the risk of annoying you in the same way that I know my teams that I work with, it would be what is the problem you're trying to solve? Um, CEO person that's read the blog. So I would be understanding their business right now. What is it that they're trying to do? What is their purpose? Why do they exist? Who are they existing for? What's the Personas? What do their Personas need? Understand that in detail and map m the workflow. So it's similar to what I was just saying there about pinpointing the parts of the workflow where AI can help. So looking for the drudge work sections, I guess it's a standard product development approach that I would take with that, with that company. So again, technology agnostic, it's all about the right problem solving, uh, user centered approach to working out where to use AI. That's probably where I'd get started.

Speaker A: And I suppose that's actually just such a good explanation of why reading a blog post or listening to a podcast, it's hard to get started just from that.

Speaker B: Sure, absolutely.

Speaker A: To implementation. And it's hard to even discuss where AI can best be applied without discussing a specific problem. Um, have you got any um, any examples you can give of like this is the best use case that I saw and I was like if everything was like this use case, you know, UK productivity would be through the roof.

Speaker B: You know a difficult question. I mean one of the more successful implementations I've seen is in the financial services space where you can imagine investment bankers and um, financiers around the world spend an awful lot of time trying to dig through tons and tons of data in order to create investment thesis that they can then act upon and invest in or divest from. Uh, the most successful ones I've seen have shortened analyst workflows massively from, you know, taking days of time to create an investment thesis simply because there's so much documentation to review and absorb and interpret to using an AI machine to spend, I don't know, one or two hours crunching through this data all on the basis of a natural language prompt. So it's very easy to interact with it. So that's the most successful implementation I've seen. It's still kind of early days, but that was a lot more that it can do. But right now you can see the results of uh, that just from seeing the analysts responses and saying how incredible it is. Um, so yeah, particularly pleased with that implementation, let's put it that way.

Speaker A: And I think it kind of allows you to pick out the kind of common factors that successful projects are going to have. So like lots of data, already established processes like you say you've not changed what the output is or you're not trying to like re engineer everything at once. You're saying here's the thing, we know this thing delivers value to the business, but we can now do this thing in a totally different way.

Speaker B: Definitely, definitely. And as well as that it's having a realistic expectation about what it can and can't do. And what it can't do is often more important than what it can do. So one of the biggest learnings for me when I got involved with AI a few years ago is how Many humans it takes to actually do AI, to do it properly, I should say anybody can sort of spin up a very quick agentic AI, but to do one that sort of powers businesses or powers business decisions, you have to do it right. So you can't just trust the AI bot to run with things. I like to say you treat it like um, an incredibly motivated and very keen intern. So think about the kind of tasks or responsibilities you'd give that human being in your business. You're not going to give them strategic decisions to make, you're not going to give them control of your finances, you're not going to put them in positions where they are having to make really critical business decisions. You use them to sort of very carefully explain what you need, you check the output and you put guardrails around it all to make sure that if they do go up track, you can spot it. So that's again, that's what AI can't do, is equally important as what it can be.

Speaker A: It's so interesting that I thought that description was brilliant because I've heard of the concept of thinking of AI as an intern, but actually once you add in like highly motivated, it's the intern that comes in and they're like, oh, I'm gonna, I'm gonna do everything while I'm here. And actually while that's a great characteristic, like a great human characteristic, there are anyone who's been in that position, you know, there are downsides to this as well. And then you throw in some like massive overconfidence in their abilities and uh, and boom, you've got AI and you're like, right, okay, so how do we um, curtail actually not just what can you do, but how we curtail what you do is yes, how you channel

Speaker B: that massive enthusiasm in the right direction.

Speaker A: So are there any other kind of expectations that when you meet uh, teams or maybe senior leaders who don't have in depth knowledge, are there any other expectations you find people have that you have to kind of recalibrate before you can start work on something?

Speaker B: It is mostly that one around expectations and assume that the machines can solve all of your problems and you can divest your entire workforce and save a lot of money? That's the main area. I suppose another area that you have to get people's mindsets around is I come from an agile delivery background and I see so many projects that are set up agile but are not actually agile. It really m does help to have an agile mindset when you're working with AI. Because pace of change is so incredible that you can set out a uh, pathway in good faith and then find out a few weeks, a month later that actually there's another bit of the technology that's just evolved that's even better than the one that you've got. So you need to pivot across so that you can continue to build the quality product that you want. So it's sort of, I guess encouraging these senior leaders not to think that they can have a road map for the next 12, 18 months of AI deployments and stick to it. Exactly. And waterfall it all out with detailed requirements and have a team commit to delivering. On this particular day you have to have a lot more of uh, an agile, a pragmatic agile mindset and be willing to pivot and learn as you go. So it's super important that your delivery expectations are equally as aligned as your product expectations.

Speaker A: Which is why it's so important to come back to what you said at the start. Right. Like what is the problem? Because I was going to say the problem doesn't change and actually that's not true. Sometimes the problem does change, but when the problem changes, you absolutely more than in any other circumstance need to switch your delivery up. Right. To make sure you are meeting what the new problem is. Um, and then working iteratively rather than fixing. And I think, I don't know, something I'm seeing a lot at the moment is that uh, it's almost as if because there's a new technology or some new words we're using for technology, people seem to have forgotten a lot of the lessons because I feel like the lessons of agile delivery were hard won, um, and so impactful when, because I'm sure you're similar to me that you've kind of seen the transformation from development teams not running Agile, you know, as agile was invented and introduced and became more popular and actually seeing the impact of that, the idea that people are suddenly saying oh no, it's fine, like it's AI. So I can instead like you say, just have one developer will write replace entire business with AI into uh, ChatGPT and suddenly it'll done.

Speaker B: Yeah, meet the new CEO of it's Chat GPT. That's not going to work. It's not going to work.

Speaker A: So how do, like, what are the kind of signals we can look at? Say someone is bringing ideas to us and saying this is a great idea, we should use AI for this. This is a great idea, we should use AI for this. What are the kind of signals we can look at to see whether something will, you know, where to invest our time and efforts. You know, as a leader, uh, I'd

Speaker B: say there's probably a few areas and again at the risk of now annoying your listeners, I would be saying what is the problem that you're solving with this new idea? If they can't answer that question in a way that links it back to a user problem and uh, a defined outcome, not an output, an outcome, then that's a signal. That's a big signal to me that the business doesn't have the right delivery culture in place. Other signals I wouldn't see data being presented forward to say we came up with this hypothesis to fix this particular problem. We launched a lightweight solution for it could be as simple as a tactical prototype or something just to validate with a real life user from your Persona group. If I'm not seeing that kind of behavior in a company and product delivery function, that's another signal for me that perhaps uh, AI is more of a distraction than it is a solution. It's usually a signal that you've got someone that's got the mindset of this is incredible technology, how do we use it as opposed to got this problem to solve, how do we solve it? Yeah, it's a tough thing to change particularly for senior stakeholders because they're the ones that read the blog. Um, they're the ones that picture on the website is the leaders of the business, they have to be seen to be on top of things. So it's tough for them to remember that they're not the user, they are not the people actually using the product necessarily and they should let the user dictate the direction more.

Speaker A: Yeah, although I like that way of thinking about it because it actually gives you something concrete to uh, rather than replying no, you can say great, I love your enthusiasm, you know, go away and bring me it back with the problem, you know the well defined problem and some data around, you know, proof of concept and then, and then we can talk about it. I think that's uh, constructive on the people side as well as maybe setting out your boundaries.

Speaker B: Right.

Speaker A: On the, on the tech side, um, have you got any thoughts on like worries people have I suppose about AI that maybe are uh, not the right things? Because I think we've talked there a bit about what ah are the right things to be worrying about. So are there any things where you're like actually there's a lot of talk about this but. Or a lot of people worry about this but that's something that we either can worry about later or it's just not in reality going to be a problem.

Speaker B: Yeah, I mean we're not short of media articles and social media clickbait saying humanity is doomed because of AI and we're all going to be reporting to the bots in a few years time. That's obviously an extreme example at uh, sort of the ground level there is this sort of undercurrent, um, of nervousness about the impact on the job market, particularly sort of in that layer of kind of office white collar workers that could get replaced. There's a lot of concern there and I suppose I've reflected on this a lot. Should we be worried? I suppose to an extent we should think about it, but we shouldn't panic about it. It's not something in my opinion that is going to happen in the relatively near future where you lose entire swathes of the working population. We are some way off that in terms of development. Uh, as we Pascal Borne's book recently and he talks about five levels of AI and you've got sort of the base level, level one. It's all the basic kind of, I guess algorithmic rule based automation that we've had for years and years. And then at the top end level 5, which is the fully autonomous system that it's no longer your intern, it is your CEO. We are so far from that. I think if it's done right and AI is rolled out in the right way, in the right um, kind of cultural, philosophical way, you will always need humans in there because AI can't replicate. And it's arguably it may never be able to replicate what humans bring uniquely, which is creativity, empathy, feeling. And how that actually applies in the business world is important because we can think about a goal that we've got, we can think about ways to implement it, but we can also think what's the impact on that beyond that pure decision? AI struggles with that. So if I give an extreme example which isn't a real one, I'm making this up. But if you gave a really bad prompt to an AI bot and said your job is to help us clear the stock of product X from the supermarket shelves, AI would quite happily go, right, well let's price it at 0 then the shelves are empty and the AI has succeeded in its goal. But the knock on effects are that the company's lost money. A human being wouldn't do that. So at the moment AI can't really think in that sequential long term impact way. So there's always going to be a role for humans in a good AI deployment. So I guess the short answer to your question is we should be worried and keep an eye on it because it does have the potential to change the way society runs and economies run. But it's a long way off from being replacing humans fully.

Speaker A: And actually perhaps that's something else to kind of, um, you know, share with people when they're maybe bringing ideas to you, is that we often don't really understand how we're making decisions as human beings, or at least not in an algorithmic way that we can explain to something that is not a human. Because, like you say, another human will have a lot of the implicit understanding that we all share and make decisions. Another example I heard was, um, clean this room. And that. Clean this room. You know, an artificial intelligence could easily throw out a baby with, uh, you know, along with trash or whatever, because we've got all these implicit understandings of what terms mean. And actually it's an opportunity for us to break those down and say, well, what do we really mean when we say this? And actually, what is an algorithmic instruction and what is a human judgment instruction? You know, like cleaning a room? What are the things we want a human, Even if we were asking a human to do it, actually, how can we make this less ambiguous or clearer and essentially solve the problem better?

Speaker B: Yeah, absolutely. And that kind of room cleaning example is good because if you approach that from a, we've got to clean this room. How can we use aic, map out the workflow again of cleaning that room. You can identify the drudge work parts that humans don't like doing. They don't like the Mockingbird. They don't like the tiding up that they might like looking after the baby whilst it's being cleaned. So that's how we would apply AI, Map the workflow out, look at where it can logically pick up. Now, the time that you, the significant time that you've freed up from the human being, they can now be looking at, how do we use this room when it's clean? How can we build more rooms? How can we do the creative side of things? How can we drive value from this room instead of spending time and mop. That's the right implementation.

Speaker A: And so just finally kind of the, the flip side of that, we're saying that jobs might not be lost, or at least not in the short term, but they will change. So what if you are worried, uh, about your own job or worried about how, you know, kind of at all Stages of engineering. You know, as a junior engineer or a senior engineer, an engineering manager or cto, what do you need to be worrying about in terms of your own personal skills?

Speaker B: I would say get comfortable with AI right now because a lot of these people that are worried about their job perhaps don't have, uh, the full understanding of its capabilities. So I recommend dive in play with agentic AI. Understand, actually if you can, how AI works. I'm not talking about the actual coding level, but if you understand how AI and generative large language models, sorry, um, do their work and how they sync, uh, how they process, not think it will help you to understand their limitations, where it can be good and where it can't be good. And if you can get that ground level of understanding, it will help you and it will not only calm you down, but you will know how to use AI in your job to become better at your job. So even though somebody isn't telling you to do this, you're able to all of a sudden get rid of the drudge, work from your day to day, get an AI agent or an LLM to handle that for you, under your control, allowing you to do the more high value tasks. So I would recommend people jump in, learn about it, how it works, its limitations, and apply it to your world. So I wouldn't sit and wait to find out what happens. I would jump in and try and shape that for myself.

Speaker A: And I love the idea that just by going in and playing around with it and seeing what it can do, you'll be building up this implicit knowledge in this human way that we do as humans, so that it will just shape how you view challenges and what solutions you come up with. Uh, and it's kind of magic, isn't it?

Speaker B: It is, it is magic, definitely. Especially the sort of emerging agentic AI space. It's very easy. You don't have to be a coder these days to create agents. Have a play around, just get stuck in, see what you can do, see what magic you can create.

Speaker A: Oh, uh, fabulous. Well, thanks so much, Steve. Um, I think that was, uh, just an extremely useful and precise summary of a kind of almost workbook, I think, for leaders to start thinking and to set things on the right track from the start and while iterating.

Speaker B: Absolutely.

Speaker A: To go forwards. Uh, so thanks very much for joining me.

Speaker B: Absolute pleasure. Thank you for having me.

Speaker A: This Digital Lighthouse episode was edited by Steve Folland and produced by Patrick Anderson. The theme music was written and recorded by Ben Bailow. A huge thanks to our sponsor Softwire for their continuing support from the inception of the show in 2019 to the present day. If you love the podcast, please let us know with a rating and review on your platform of choice. We're always looking for feedback to ensure we're making the best show possible, and if you'd like to take part, please drop us a line at thedigitallighthouseoftware.com you've been listening to the Digital Lighthouse with me, Zoe Cunningham. Thank you for sharing your time with us, and stay safe on this wild technological ride we're all on.