Nov. 4, 2024

Johann Van Tonder: Exploring the role of leadership and experimentation in successful AI mobilisation

Johann Van Tonder: Exploring the role of leadership and experimentation in successful AI mobilisation
Johann Van Tonder: Exploring the role of leadership and experimentation in successful AI mobilisation
AI Pathfinder for Private Equity Podcast
Johann Van Tonder: Exploring the role of leadership and experimentation in successful AI mobilisation

Summary In this episode of the AI Pathfinder for Private Equity podcast, Steve Budd engages with Johann van Tonder, CEO of AWA Digital (E-commerce CRO specialists), to explore the complexities of AI mobilisation. They discuss the high failure rates of AI projects, the importance of embracing failure as part of the learning process, and the necessity of identifying specific problems that AI can solve. Johann shares insights on how AI could transform the conversion rate optimisation industry, t...

Summary
In this episode of the AI Pathfinder for Private Equity podcast, Steve Budd engages with Johann van Tonder, CEO of AWA Digital (E-commerce CRO specialists), to explore the complexities of AI mobilisation. They discuss the high failure rates of AI projects, the importance of embracing failure as part of the learning process, and the necessity of identifying specific problems that AI can solve. Johann shares insights on how AI could transform the conversion rate optimisation industry, the slow but significant changes occurring across various sectors, and the role of leadership in building a culture of AI adoption. The conversation also touches on the importance of data maturity, the organisation of information for AI utilization, and the selection of appropriate AI tools. They also discuss the need for a thoughtful approach to measuring success in AI initiatives, recognising that not all valuable outcomes are easily quantifiable.

Takeaways

  • AI has the potential to significantly impact businesses
  • A high percentage of AI projects fail due to lack of readiness
  • Embracing failure is essential for progress in AI
  • Identifying specific problems is crucial before implementing AI
  • AI is not meant to replace human roles but to assist them
  • The transformation brought by AI is gradual but profound
  • Leadership should empower teams to explore AI without pressure
  • Data maturity is a common challenge for organisations
  • Organising data effectively is key to leveraging AI
  • Measuring success in AI initiatives can be complex and subjective

Links
Public version of Johann's hypothesis GPT
Post re AI metrics
Johann's LinkedIn profile and company website

AI Pathfinder Private Equity AI Strategy Meet-ups:

https://www.aipathfinder.co/private-equity-ai-events

Have a question or would like to suggest improvements, please contact steve@aipathfinder.co

SPEAKER_00

Hello and welcome to the AI Pathfinder for Private Equity podcast. I'm on a mission to lead private equity from AI bewilderment to AI empowerment. And I'm doing that by building a dedicated network that will piece by piece create a playbook for successful AI adoption across private equity and their portfolio. In building this network, I'm meeting and having conversations with some very interesting people. They may work in or with PE or provide a solution to their portfolio. What unites them is their curiosity of this emerging technology. And I'm really pleased to bring some of these conversations to life through this podcast. Today we're going to talk about the building blocks for successful AI mobilization, a theme for my next private equity AI strategy breakfast briefing in London on the 3rd of December. A recent report by RAN found that 84% of business leaders believe AI will have a significant impact on their business, but the majority aren't ready, leading to as many as 80% of AI projects failing, twice the failure rate of other IT projects. So how might private equity reverse this trend? Well, joining me today to discuss this is CEO at AWA Digital, Johan Van Tunder. Hello, Johan. Thanks so much for joining me this morning.

SPEAKER_01

Steve, I had no quick uh idea you're going to ask me that question. I don't know the answer to it.

SPEAKER_00

No, well that that's part of the discussion, isn't it? You know, I think that these conversations are for us to talk about our personal experiences and and you know those that you've got within your business. And so how are you mobilising? Yes, we're not directly talking about how private equity is actually going to do this. But in the session we have on the 3rd of December, there's a question will be just as open, and no one will have the answer. I don't think anyone has the answer to truly mobilize for AI. But also we've got the opportunity to perhaps reflect on other transformations, digital transformations as well.

SPEAKER_01

I I you know, I think that's it. You spoke about the failure rate. Um, and I think there's um when I look so I've been obsessed with this uh since before the ChatGPT days, but of course, when we make this recording, it's almost exactly two years ago that ChatGPT was launched, and that just sort of put it on the world map. And so I've been um I've burned thousands of hours um uh on it since then. And when I look behind me, there is you could characterize it as a a trail of failures. So most of the things I would say that I tried to do didn't work. I can show you a lot of things that didn't work, and there's a fraction of the projects that I attempted to do that actually ended up delivering what I expected and more, and that's made it all worth it. So I think there's a there's a way to reflect on the failure in a different way. I I I don't think I I don't think there is a way to make progress without getting it wrong. That's the only way to make progress, has been my personal experience of the last two years.

SPEAKER_00

So embrace failure essentially and accept that that's part of the the journey.

SPEAKER_01

I mean that sounds like such a uh a cliche, doesn't it? Fail fast, fail often the Silicon Valley.

SPEAKER_00

Yeah.

SPEAKER_01

Um I I I I don't think you set out, certainly I don't set out to fail. I set out and I spend an entire weekend building something which I think is going to work, and then it doesn't. Just yesterday, I was up until late, much later than I should have been. Um, MacBook on my lap in the bed, uh, all the stuff that you know you shouldn't be doing. Because I was convinced I'd be able to figure out this one thing, it's still not figured out. And maybe by the end of this week I'll abandon it and it doesn't work. But what I've learnt, there's so much that you learn through osmosis, through what doesn't work, brings you a step closer, um, and it it triggers other things. So I don't think you're set out to go and fail, but it it's just part of the process.

SPEAKER_00

Well, look, with the with uh as as as we've spoken um before about generative AI, it's not it's a black box. So you you've got to experiment to know what it's good at, what it's not so good at. And there are different models out there that are good at uh and not so good at um certain tasks. Um I think that there are a couple of things here. You know, I I said at the top the word curiosity, I think you've got to have that um curiosity, and that's you have that in spades. But I think that what you've also got to have is a um you've got to be close to the problem that you're wanting to try and solve. And and so I've got a feeling that that's that's something that you're very aware of. Um there's maybe an array of problems, and you're you're trying to see how AI can help you solve some of those. And it might be a list of a hundred things, and you've you're ticking them off or crossing them off as you go.

SPEAKER_01

Yeah, and you're right, and new problems keep on coming up, and I think that's such a core point. Um, because I speak to in my work um many clients, organizations, um, and one of the questions that come up is how do we use AI? And I say to them, well, that's the wrong question. Um, what problem are you trying to solve? What are you trying to get done? And then let's see if AI has a role to play there. Um and and and if it does, then how do we integrate it into the process? It it does start there. Um, what are you trying to get done? What problem are you trying to get solved? I get the question a lot. So, what is it that you do with AI? How do you use it? And that's a really difficult question to answer because how do you use the internet? It's like that for me. You know, it's always on, it's just there in the background, and sometimes it comes to the foreground, but it it's it's just become the way I work. And it it's interesting looking at how it's changed my behavior and the way I do things. So, for example, I spend a lot less time on Google and a lot less time in email, which I'm thankful for, because the hours have to come from somewhere to invest in this. So I'm I think being more productive, uh not getting sucked into the email and you know, channeling the hours there. But the other one that is really interesting is I go on walks with um with my phone and with my AirPods, and I have conversations with the um with the AI, with in that case ChatGPT, and maybe we'll talk about models, but that's a good one to use for that particular use case. And you know, it's like talking to a mate, and when I come back after half an hour walk, I've got a transcript of what happened on that walk. I've asked some questions, I had some inputs into my thoughts, and I didn't have to stop every two seconds to scribble it down. It's all there in a transcript, and it makes good sense of my accent. So um, yeah, it's it's it's changed my behaviour in fundamental ways.

SPEAKER_00

Yeah, look, that change of behaviour, obviously, that we've got used to with Google over the years, um, of of when we go, well, you know, we still go to many for all sorts of reasons, jump straight to Google, and you've disrupted that behaviour to to perhaps go first to you know uh Chat GPT in that case. But the actual idea of of going out for a walk and actually having a conversation, and as you say, then having a transcript, so you can any nuggets you've got there to be able to uh pull on, I think is is really fascinating. I certainly haven't got to that point yet, although it's very much in my workflow day to day, is that when something comes up now, uh it's such a habit it triggers me to then see, oh, how might AI help me in this situation? It's just finding ways that it can help you even even more. Look, I wonder if we should rewind a little bit, because it'd be good to get a little bit more context on you, Johan, and obviously the your your business as well. And you've talked about I suppose your personal relationship with AI, but how is that impacting your space at the moment, your industry?

SPEAKER_01

Yeah, so my industry being conversion rate optimization, CRO, um experimentation. So um, if you're not familiar with the industry, it's um understanding the world of the customer and then forming ideas, theories about how to improve the world of the customer through what you do, and then testing those ideas, A, B testing those ideas. Um, and so we do a lot of data analysis, which is uh a good application of AI, and that includes both quantitative and qualitative, so numbers, but also words, what people say. In fact, I tend to think that the words are actually more important than the numbers, and and most of us are sort of stuck in the numbers, but the the words, the qualitative, what's in the mind really explains what you see in the numbers. And AI is is very good at not doing it, not replacing. I think that's a a key theme that comes up in all these conversations and whatever you read on AI. Most people will tell you that, and I agree with that fundamentally. It's it's not here to replace you, it does certain tasks really well, and it you know, it's recognizing what tasks it does well, but you are firmly in the driver's seat. For me, it is a I don't like the metaphor, but there is this popular metaphor of um of it being a junior intern. Um and and maybe that's one way to think of it. So your question how it's changing the industry is an interesting one because I you you'll get different answers from different people in the industry. I'm firmly in the one corner, and I believe it is changing the industry in a way that we'll wake up in two or three or maybe five years from now and look back and go, oh, yeah, it the world looks very different. But it's happening in slow-mo around us. So I'm old enough to remember the days when the internet just became a thing. I I left university just as the internet became, you know, people started getting dialed up in their in their house kind of thing. And if I I'd become comfortable with the internet because we had it in our labs, I looked at this and I saw the headlines that this is a fad, people buy from people, you're never gonna buy on the internet, it's hype. And right now I'm having deja vu. I'm hearing those same dismissive statements about AI, but I've got that same feeling as a 50-something-year-old that I had as a 20-something-year-old that this is changing the way we plug into the world in such a fundamental way. I have no idea what the future is going to look like, but I know it's very different. And so I think it's severely disrupting our industry. Most people just don't know it or don't want to know it. They choose to kind of think it's not really happening and dismiss it as as hype.

SPEAKER_00

And do you think that I mean you're you well, you're you're definitely uh much more of a leader than a laggard. Do you think that that's had an impact positively on the business in terms of how you're um you know you're seeing positively the the the what AI could do?

SPEAKER_01

You would think when you when you hear me speak about AI that our organization is you know completely run by AI and it's transformed. That's not how it's turned out. I'm I think we've into it's a similar story to what I said earlier. We have found a lot of ways not to do things. Um so we've walked into a lot of whole walls, um, and there is some really valuable applications that have come out of that. I think far less than you would think. What I find really surprising is how the the lack of enthusiasm about this around me, and and I socialize with other people in the bubble, like yourself and friends who are equally obsessed with it. But it is a very small bubble. And what I've recognized increasingly is that most people firstly aren't part of that bubble, don't want to be part of that bubble, and look from the outside in on people within that bubble, thinking, you know, they they just hyping it up. And so I I've started toning down what I say about AI deliberately in my organization because I don't want to I don't want to force it onto people. And I, you know, there are I think it's a good microcosm. There are there are people in the organization who embrace this and are really curious about this, and that's great. Let them figuring out, let them figure it out in the same way that I'm figuring it out, but I don't want to force people along who don't want to do it. And so it's an interesting mixture of things, and what ends up happening is that I've I think it's fair to say I've abandoned any idea of enforcing this top-down. And what's happening now is I continue to be openly, publicly passionate about this, I continue to share what I do, and I accept that most people won't want it near them, but those that do, it'll make a difference in their lives, it'll make a difference in their projects, it'll make them more productive, and so there's a bit of a mixed bag. Um, and I think that will change, but that's where we are now.

SPEAKER_00

Is there I look I think it's hard to predict who in an organization is going to be, or the skill sets that are required to for people to either be really curious or not, or or you know, shy away. So it's not predictable who you think is going to adopt the technology. But just from you know, you're uh the leader in your organization, and and when when um people talk about successful transformation, uh mobilization, a lot is mentioned around the culture and the the leadership, you know, the leadership, um, and and it's got to start from the top. And so from what you're saying there is that you've maybe tried to mandate uh an approach, but that that's perhaps backfired, has been pushed back. So you're now well I can imagine very visually, um, as you said there, you're sharing what you're learning. How do you you then do you incentivize those that are coming up with some really cool things, new ideas, whether that's efficiency's productivity, or whether it's how you might be able to develop your product? How are you sort of encouraging those people?

SPEAKER_01

I mean against the idea of incentivization um in this context because I think it's again it creates the wrong sentiment around this. And you know, it's like incentivizing people for using the internet. Um I so a couple of things. Firstly, I've as I say, I've for now stopped trying to enforce this top down. I have a bit more empathy with people throughout not just my organization but other organizations as well in similar positions. I think there's skepticism for good reason when leadership tries to impose this on people. Because what's the narrative that we read about? You know, and this comes at the background of mass layoffs around the world. And very few, I mean, the tech industry has been particularly hard-hit, but very few industries have not felt this. And so I think you've got to read the room and be sensitive to that. And so a leader coming in saying, uh, you know, use AI, adopt AI, the subtext there is because I want it to replace your job so I can have a higher margin. Um, I really think that's part of it. And I think that's just human, that's just natural. I perhaps would have reacted in the same way. I must be clear, nobody has said this to me. This is me reading between the lines um and connecting a lot of dots that may not even exist. But I think that's one thesis that I'm working around. So, what I'm doing, you're right, is I'm I happen to be obsessed with this. I happen to use it in a way that I think opens up wonderful opportunities for everyone in my industry, and so I make that known and I demonstrate that and I show that and I talk about it. Um and um some people want more of that and they naturally sort of migrate to that and others don't, and that's where we leave it. I think one of the lessons, and there's actually a study on this, one of the lessons for me has been you've got to put the tool. What I think leadership should do right now is enable, give people access to the tool or tools, and then get out of their way. Let them figure out how to use it. Because as you said earlier, they know what problems have to be solved, they are working on the front lines every day trying to get things done. In fact, they're probably doing it already. You might not know about this. This also came through in the study, is I don't remember the percentages, but there's a significant percentage of people in the workforce who are actually using AI and the bosses don't know about it. Again, I think the reason is as I explained earlier, uh, and I understand that. So it's happening already. And and I think what leadership should do is just enable that, not even encourage it, because the moment you start putting too much weight behind it, then you know that that fear I think naturally arises. And so I think if you're going to encourage it, I think you've got to be really transparent about your agenda there and reassuring. And people, people I think need that. And and then the last thing I'll say. So the one is give people access, let them figure it out, because they will figure it out. Number two, I think you've got to put some guardrails in place, call it a policy, if you will. I've seen a number of policies now, and some of them are just so restrictive, the lawyers had so much fun. You may as well not have a document because you're just, you know, killing it. Um, so have a policy that is sensible and you know allows people to get stuff done. It's more like guardrails, I think of it. And then number three is to orchestrate it somehow. And I don't know what the role of leadership is in that. I think it should be done at the lower ranks. And what I mean by that is, you know, work groups get together every week or two weeks or month and just talk about how they've used it, what they've learned, what they've done, share some of the insights. And I think in most organizations, lead top leadership should not be involved in that for reasons that I mentioned earlier. Unfortunately, that's just the reality, is because of where we are right now following a lot of layoffs, following the narrative around AI, people see this as a threat, and you've got to do whatever you can to kind of um give people reassurance around that. And if it means removing your presence from it, then that's what you do.

SPEAKER_00

Yeah, really good, really good points there. Um, and you know, a lot of this is around change management, isn't it? But as a uh as a business and as you're looking forward, are there other aspects and a key one here is around sort of data and infrastructure that you're thinking we need to we need to be more mature here, more sophisticated. There's these foundations that um also need to be be looked at uh in terms of uh I suppose preparedness uh as a as a business as a whole. Yeah, definitely.

SPEAKER_01

Yes. I I think all of us are being caught um in uncomfortable positions in terms of data maturity. You know, we we're a we're a data organization. We we we we make sense of data on behalf of clients. And I think even we have discovered weaknesses that and and and it's it's natural because it's the kind of weakness you can only discover when you're trying to do things that you previously there was no reason to even think about doing that. I'll give you a good example from my own personal um life, and then maybe we can transfer it to business. I consume a lot, uh read a lot of books, um, peer-reviewed articles, uh I read a lot, I listen to a lot of podcasts, um, and so and then I'm I'm a note taker. And I've digitized m most of this now, um, and that's been part of the obsession for the last two years. And the reason for that, and this goes back ten years, so I've got, for example, meeting notes going back ten years that I've scanned. And you know, upload it. Highlights from my Kindle reading, all of that is in a repository. Podcasts, transcripts from it are in a repository with highlights. Because when you deploy AI on top of that, you're able to firstly find something at a needle in a haystack by searching awkwardly for an adjacent phrase. You know, you can't quite remember what it is, but you sort of remember the context. AI will find it. But more than that, AI will has the ability to process all of that, find dots that you would not have found, and then connect those dots and surface it to you and say, here are dots that you may have forgotten about not even being aware of. Here's how they connect it. Um and then you go back and you say, well, can we connect it in a slightly different way? And a minute later you've got, you know, so you so that I think it's now my highest ROI activity is just diligently making sure that everything I consume or want to consume is in there. And I use ReadWise for this, um, and Obsidian, and um AI plugs into both of those, and I use various AI tools for that. But it's that's one of the things I said earlier that it's always on. That's one of the things that's always on. Um is it because and I speak to it. It's like I talk to my notes. Oh, the other thing that's in there is I'm a journalist, so I write a journal, and I've written this for uh many years. All of that is in there as well. And I I'm able to have fascinating conversations with my journal, and it's it's really helped me from a personal growth perspective. Um, so I can't remember what the question was. Oh, it was about data. So I I think what that means for organizations is we've got so much data floating around, and and it's about just starting to kind of pool it. And this doesn't have to be very fancy, you know, start with a Google Drive, or you know, just it's just about organizing these things in a way that you uh can ultimately, if you're not doing it right now, expose it to uh to a machine.

SPEAKER_00

Yeah, there is so much, isn't there? The conversations that we're having every day. Um, whether that's um prospecting, whether that's customer conversations, and obviously you've got to be sensitive to to recording those um and what you do with them, but uh there is so much across the business that could be captured and we could be learning from. Um I think that um you know I'm on-person business, so it's it's highly manageable for me at the moment to be recording conversations and doing as you I'm definitely not as efficient as as you bringing all of that together in one one well it's uh an extended brain, isn't it? Really?

SPEAKER_01

Um to be able to they talk about the second brain, yeah.

SPEAKER_00

Yeah, absolutely. Um, one thing that um uh Google's um uh notepad or notebook LLM has been really interesting where you can just on the fly drop stuff in, and I know you know I suppose in in a similar way that's that's what Chat GPT can do. But they just created a nice interface that you can drop all sorts of stuff in there. Podcasts, I think you now you can put in there, and you can start to to query, or you can even create your own mini podcast that will help you understand a little bit more about you know it the the information that you've uploaded.

SPEAKER_01

I think the the other point you make there is um an important one about you know the different tools, and Chat GPT is the leader. Um for whatever reason. I looked at the the graphs just a few days ago, and it's by far um leading the pack. And Claude, which is an interesting one, is is nowhere almost. And I I think that's just from my experience and what I hear from other in this bubble, that's probably the most impressive model uh around at the moment. It beats GPT-4.0 uh hands down in a number of of use cases. And then Notebook LM, which you mentioned there, is perfect for it was it was designed by I think one of the one of the leaders in that project is a former journalist. And so with that background, what they the product they designed is made for pulling out little nuggets out of transcripts. It's actually made for that. So it's if I if that's my use case, I would go to Notebook LM and not GPT. Um if I want to do coding or writing, I would go to Claude, not GPT. Um it has a GPT tone of voice. I think everyone is aware of this. It is terrible, right? It is it's got a horrible style of um robotic way of writing and talking. And there you've got to, if if you want to, if you want to write with it, you've got to give it a lot of guidance. And even then it struggles. Claude is on the opposite side of the spectrum. Um, but GPT, where I use it a lot, is with the the custom GPTs, which is like a little um program that doesn't involve programming. You just speak to it in English. But it it's a task that you repeat quite often. So I'll give you an example. Uh there are many examples, but in my business, we do a lot of hypothesis writing. And you know, it's the kind of thing you hear hypothesis and you think back of your science classes and you have a yawn. Um and and most people, because of that, don't writing a good hypothesis, don't really, it doesn't really come natural to them. And so people really struggle with this, but it's a vitally important part of A-B testing, of the experimentation cycle. If you mess that up, the whole thing is messed up. And so I've created this little custom GPT that helps you to write a proper hypothesis. And it shouts at you. It goes, you know, it's it's not because GPT, the all these models really, I think they're trained to be nice, to be polite, to, you know, uh you you see this in the way that they respond. They they apologize and you know, they're friendly. This one is trained to not be that. It it knows that to be helpful, it's got to push back. And so there's no way that, and I stress test this this now for months. There's no way you can use that GPT and not have a decent hypothesis. Um, and so that's you know, that it's the kind of thing that I have daily use for that, and so I've got a few dozen of those where I'm in GPT all the time, just for that reason. One more that I'll mention that hasn't come up in this conversation is perplexity. And if you talk about um search, so you know, Google, I don't think ChatGPT is a good replacement for Google. That's the wrong word, but it's it's not made for that. But perplexity is made for that, and I think it's underselling the value of perplexity to um characterize it as a Google replacement, it's much more than that. But it's that's become my default search in most cases, and it's a far better experience than the 10 blue links in Google.

SPEAKER_00

So um, we're coming to the end, and this this can be quite difficult, and we may have touched on it already, but um actually measuring the success of these experiments, and and you could say, you know, for your business of how you're progressing with the technology. Have you got defined KPIs around this?

SPEAKER_01

No, there's um there's an interesting post that I saw, and I'll send you the link to that as well. Um, just recently, um, and somebody in the space who listed KPIs, I think that can easily backfire. And I've I've seen that happen elsewhere. But I think we we measure what we can measure because it's measurable, and the things that really truly matter often are invisible, let alone measurable. And so again, if I had to reflect on my own life, right? So looking back at two years, I have spent evenings, I've spent weekends, I have invested a lot of personal time in this. Um if I if I were tracking that, if I f you know is that an do you track the input? And then how do you justify that input? Because as I said earlier, the you know, the the the failures are more than the than the successes. Um and and really the thing that I get from it is hard to even express, let alone measure. I'll I'll try and express what I get from it. After 20 years of doing what I do, frankly, I was getting a little bit bored. And this has given me new excitement. It has given me a new quality of life that I didn't think I would have. It's energized me. It's completely impossible to measure that. But then you can talk about productivity, and I think there have been a lot of productivity gains for me, but but there's that's also how do you measure that? Because there's there's this notion of use AI to do things faster. I find there's a lot of that. But what I also find is using AI actually takes longer, but you get a better quality output. You're able to produce something far above the standard that you'd be able to do on your own, or perhaps even with a team. And so I'm happy to invest the time doing that. But how do you measure that? And so I deliberately don't put numbers on it because I think numbers mess it up. And I think that's true not just for this, but for many areas. I know that's not the answer you want, but that that that's honestly what I feel.

SPEAKER_00

Well, look, I think that's um the um, yeah, the energy, the motivation um that the this emerging technology has given you um is yes, it's really, really good to good to hear. Um, and I think that um if we just reflect on this conversation at the beginning as well, we don't have the direct answer of how to uh mobilize for AI. I think that you've given some fantastic um tips and and describing your your journey, which ultimately is about your mobilization um from a personal and and from a business perspective. So Johan, thank you so much for for sharing being so honest with with that journey. And as you as we've uh talked through, there's at least a couple of uh links there um which we'll put into the podcast notes. So yes, thank you so much for your time, Johan. Steve, always good to talk to you. And I I recommend to uh connect with Johan on LinkedIn. Again, I'll put that link into um into the podcast notes as you're you're sharing you know your learnings uh with everyone, which is which is great. And obviously check out AWA Digital. Um but for now that's it. Thank you for listening, and hopefully we'll see you on the 3rd of December for our next breakfast briefing. By the way.