Why Data-First Thinking Is Killing Your AI Strategy

00:00:06 Mark Smith
Welcome to the MVP show. My intention is that you listen to the stories of these MVP guests and are inspired to become an MVP and bring value to the world through your skills. If you have not checked it out already, I do a YouTube series called how to become an MVP. The link is in the show notes. With that, let's get on with the. Show. Let's kick things off with a fresh voice from the heart of Copenhagen, where AI meets real world impact in the Microsoft ecosystem. Today's guest is someone who is not just riding the AI wave, he's helping shape the surfboard. Please welcome Erik, Chief AI officer at delegate. Erik and powers businesses to harness AI driven insights for transformative growth and innovation. He's the architect behind the delegates. 3 Day our workshop, which we're going to talk about a hands on Sprint that takes companies from. We should do AI too. Here's a product ready solution. So many companies are struggling with this at the moment and he's not guiding enterprises through the maze of generative AI. He's championing responsible AI practices across the Nordic tech scene. Fun fact, before leading fun fact before leading the the the this charge in AI, he's been working in the LM space for years in this language space. So with that, Erik, welcome to the show.

00:01:32 Erik David Johnson
Thank you so much.

00:01:35 Mark Smith
I'm excited to have you on because going through your brief and bio and what I do in my world is there's a lot of similarities. And so I'm excited to to share notes and understand your journey in the space and what you're seeing. You know, out in the market with customers. Before we go there, tell us about this workshop, this three day workshop that you run, and how did it? How did you come about creating it?

00:02:02 Erik David Johnson
Sure. Uh. So I'm doing it right now with two uh customers actually, but the. It really just came about because it was an apparent need in the market. Uh, so a lot of companies were struggling to go from the very first intuition of we should probably also do something with AI, but what exactly should that be? And then? To actually getting started in in practice. And uh, what is needed is sort of a road map and a use case catalog. So that's really the end goal of these three ideation, AI ideation workshop days is that we talk to the business and map out all the use cases then we. Try to evaluate them in different ways, or get into some more details, but but just the. Main shape of it is then uh. We try to qualify them and uh, we also present them with a candidate architecture and our suggestions for which of the use cases should they start with. And that sounds quite simple and it's not. So I have done this for a long time now and I've had a lot of learnings and one of the major learnings were that. You really need. To respect. That the participants know a lot more about their business than you do. You're sort of an outsider, and it's it's actually more realistic to educate them in the major trends and especially what's possible within AI. And that's machine learning and generative AI. Also understood this last language models and that's much more viable than you being an expert in their business at at their level, but really it's it's sort of. A joining of the two. So we need to learn about their business and they need to learn about AI. So the ideation workshops usually start with and it's often me. I love to educate and and be the the teacher in the class room. I I start by teaching them everything I can about AI and. I've been doing a lot of that inside and outside of ideation workshops. Also just doing master classes and yeah, I just did up the numbers. I looked at the numbers recently and I could see the last 18 months. I've talked to a little more than 5000 people a little more than 120. See boards organization companies. Also the government, the government institutions and so on. And when they learn about AI, then they're ready to have this. This commonality in terms of language. So we're talking about the same things and we mean the same things when we the say these words. UM, and then we start talking about, well, if that's what's possible with AI, what's possible for us and looking at their where they want to go and a really important point is here that we don't go data 1st and we don't go technology. 1st So data first is the the common mistake that you say we have this data and that means that this these are the things. That are possible. Which is a terrible way to go about it. That's sort. Of. The the the data that you coincidentally have, that's what's going to. Shape your future. Why not sort of take charge of your future and say this is where we. Want to go? And that means that all our AI ambitions actually amount to us having to invest in data and invest in the data strategy and and so on. Uh, and the larger companies? Are these the easiest for them to grasp this but but it it's it's a a problem for everyone, especially because sometimes a is seen as something where you can sort of do a small investment and get a lot back very, very fast and. That's. Contradictory to this idea of investing in data to. Really. Have a strong baseline for you of future innovation potential and the other part of it that makes it difficult besides sort of going against the the the data first is is is the the technology first impulse where it just tell us what tech we should use and then we'll figure it out. Uh. A really good example of this is that people are really confused about whether they should use. Just use chat, KBT or maybe just this version of copilot. Or should I invest in starting to use data bricks and Azure AI services when it's needed? When? And really we don't talk about that until we're completely done, sort of looking at. Other barriers, like what are the organizational barriers, what are the cultural barriers, what are the barriers in terms? That the the users are not willing to adopt the the solutions because there were about some future scenario that hasn't been addressed. Will I be replaced by AI or will I my job become uninteresting because of AI and so on? Once everything is addressed, then we. As AI experts look at, well, you should. Probably start with copilot for this use case, it makes sure that that everyone is on board. It has a very low entry barrier and it will show everyone. Ohh AI's are so dangerous and we we will sort of convert the culture of the company towards AI and then start investing in data. And then along the the road, some of this data may be relevant for more enterprise. Level solution which might be in Asia I services and so on and that's how the whole journey works with a start in the AI ideation workshops.
 
 00:07:55 Mark Smith
How are people finding you? As in, you know, it's a phenomenal amount of presentations and people you've you've you've addressed are people as in our company, scrambling to get their AI strategy sorted out and get into the AI game, so to speak, for their business. Is that how you're? You know you're finding these people.

00:08:17 Erik David Johnson
That's sort of A2 sided question because my immediate answer. Is people doing great at adopting AI? So uh, last year it was all about uh learning about AI and using it for personal efficiency. This year it's all about let's get started building things into processes. Let's make our first RAC architecture. Retrieval augmented generation flows not initiated. That means simply that the last language model can answer questions based on your own knowledge base, which is typically a collection of documents. And then that's sort of it. I think this is sort of a not for most not ambitious enough. Not sort of self reinventing enough and I think it's because. AI has been sort of overhyped for a long time, and now the the the, I wouldn't say violent, but the the really extreme transformation that generative AI presents for everyone is overlooked and underestimated. So I think everyone is doing something, uh, but very few are taking it seriously in a way that probably would be best in their own interest.

00:09:40 Mark Smith
Yeah, interesting. Interesting in the workshops here, are you using a design thinking type process to get the engagement with?

00:09:47 Erik David Johnson
Folks, yes, that is sort of working behind the scenes, but what we found was that the more we sort of put these activities into frameworks. And papers and schedules. And you have to fill out this field and sit in groups and so on. The less creative and thinking outside the box people become, so really you have to master a huge degree of fluency in the room and you are as the consultant, they are consulting the one who's responsible for fitting everything together. One a major learning has also been. That. And I'm very harsh and very direct about this. When I'm talking to companies that that they, they they ask me who should be in the workshops and I I say. If there's anyone from upper management, it should be 1 Max. Two people no more than that. Middle management. I don't want to see a single middle management person. The workshop might offend some people, but it's my responsibility to get results and and what I really need in the workshops are people who are. At ground level, doing the actual work, they are the ones who will see, oh, if I could do this faster, then this would be really beneficial and so on. Because AI value is AI in practice and not sort of these ethereal constructions that are easy to market and hard to implement.

00:11:21 Mark Smith
I like it. I I like the stance that you take on data. It's very interesting and and it is.

00:11:28 Erik David Johnson
Or I have another one of those? I have a pet peeve that I want to. I want to kill the the step ladder the the the step, you know. Did you hear about that that first you were sort.

00:11:28 Mark Smith
Different.

00:11:39 Erik David Johnson
First, you are, uh, digital, so you don't have stuff on paper. It's it's digital somehow. That's the first step. Then that's then, then, then you're ready for for the next step, just like being promoted or completing levels in a computer game, right. The next step is you can do some reporting and once you achieve that, you're ready for the next step. Which is sort of doing some predictive analytics. And then the next step is machine learning and and really complex generative AI and so on. And I hate that model because it will sort of prevent people from being timely in their innovation. Needs uh, because the the the thought process is then ohh AI we shouldn't have. We don't. We're not thinking about AI. We're only at the reporting level and I'm saying to people the moment you should be thinking about AI is the moment that you are moving from paper to digital. Or even before I mean. It's so dependent on data. Also with generative AI, which is overlooked. And. The moment you are starting to collect or not collect data which is be a very conscious decision in in your business, then you should have some AI expert telling you asking you what are your future goals and goals in a three day workshop or whatever and what requirements in terms of a data collection strategy. Should be part of that and and it's. It's made even more complex with the coming EU AI act here in in in Denmark where which is part of the of the EU.

00:13:18 Mark Smith
So let's talk about that. The UAI act because it's going to affect a lot more than the EU. There is the specific clauses in there that are really affect any country that have EU citizens traversing through them that might be using an AI in any form. So I think it's more wide reaching. How are you? Looking at when you're consulting with your customers, the application and the UAI Act in their readiness because from a couple of levels one, there's obviously quite clear about what what's allowed and what's not, but also there's. There's a massive requirement to educate your staff on AI as part of the UAI Act. In other words, you've gotta bring their AI literacy up and have gotta be able to show that you're doing that over time. How are you working that into what you're what you're doing with customers?

00:14:13 Erik David Johnson
Uh, so of course, just being a trusted advisor in any field requires that that they feel they they the customer field they can trust you to, to to sort of keep the interests at heart also not just selling them. Something that would be illegal in in one year. So so you, you I think you're obligated as a as a someone who provides IT solutions to to be part of that understanding so personally I've. I'm very. Involved with the UI act. Uh delegate. Uh. Was one of the first uh primary stakeholders on the official assessment list that later became became the UI Act. So we tested earlier versions of it on our AI projects and gave feedback to help shape it. And I also personally member of the European EU AI alliance. And also, helping Danish politicians sort of catch this in the right way, participating in different UM. It's called follow groups. I'm I'm not sure what the English word is. It's sort of advisory boards or groups with experts sort of helping the government steer the right path through this. And I was took the the EU AI act education that lawyers have to take. To. To be proficient in the UI act. So I have a lot to say about it and I also I want to start by correcting you somewhat if that's alright and that is. Yes, or everyone's affected by the UA act, but it applies in the EU, so if you are a country whose customers are outside the EU, you're not directly affected by it. It it applies to all uses within the EU. So I heard someone say I'm gonna move my shop to China. I'm and then. It's it's too difficult to to sort of deal with that. And I said, well, aren't your users still in Europe? And they said yes, well, that's not gonna help you. It's where it's used that, that it applies, right. Uh, then there's, uh, the the, the, the, the, the uh article concerning uh education and that it's the sort of uh you're obligated to educate anyone who uses and and also works with. Sort of delivering AI solutions and the. There's the official stance that you have to do this. I also sort of educated all of delegate. In in several times, actually just in in rooms of 12. So it took it took some time to get through everyone, but that was an investment. It was not to comply with the UA. I act the way they act is set set up right now is that it? It's very, it's not really matured. So they aren't great at saying what's required exactly. So so really you could get away with. Interpreting it any way you want and say, oh, I thought you meant I just had to sort of send a YouTube video to everyone, and then I was done. And then you can't actually be fined or get in trouble. Then they'll have to say once they get there, stuff together. Well, that's not enough. You have to do something more and then they have to tell you what that is. And then you do it. So it's it's really nothing to worry about at this stage unless you do absolutely nothing. And you have no documentation of doing anything at at all.

00:17:41 Mark Smith
Yeah. Interesting. You mentioned lawyers there, the training that lawyers that. Need to take. Are you seeing that the legal profession start to develop? AI practices. So in other words, part of their, you know, they might have a a practice that does different types of legal work. But are you seeing them now starting to go? You know what we see an opportunity to set up as a dedicated AI legal practice within their businesses?

00:18:09 Erik David Johnson
Right. So there are two sides to that. The first is to what end and to what degree can legal help sort of? Get us compliant with the UAI Act and in that regard that that was set to this room of lawyers where I felt like I. Was sort of a spy. Almost with technical knowledge but but because I'm not alone at all to any extent, but what they said was well, unlike GDPR, you can't sort of handle this yourself. As a lawyer, you will need to sort of cooperate with technical people to understand whether you're compliant or not. Unlike GDPR, this has to be sort of cooperation between AI technical people and legal experts, so that's a very. Huge difference from from GDPR, which is also a big thing in in the in the EU. Then there's what does it mean to to be within the legal field with everything that's going on with with AI? And I think it's a huge. Deal whether you have the data to have all the case material. Is it publicly available or do you have case material that others don't have and that can be turned into a business? So if I'm a a legal office and I'm an expert in certain kinds of cases. I'll have a lot of history on that and I can use that to set up an architecture where I can easily draft cases, so I'll be much more efficient and the future will be that the legal officers that have all the data will transform themselves into a service that people can apply for and saying now we can talk to our legal bot. Which has access to all the relevant data, so it can draft cases that are relevant and compliant, and we will also sort of help sort of reading through the last part. And in terms of EU Act, we will take personal responsibility. This has to be someone said. This goes uhm, so so they're they're also being transformed quite intensely.

00:20:10 Mark Smith
Tell me about the UM, the ISO 42001, which is the, you know, the act or the ISO IC regulation. AI management systems. Have you looked at that and? And the implication of those I see companies now starting to have but more attention to it.

00:20:29 Erik David Johnson
Not. Not as much as as the EU. Yeah, act. And that's because it seems like this more technical. Standard is is more relevant to the to the CTO, so I'm trying to grasp uh everything but uh. I think it's so important to say I'm not an I'm not an expert in that, but as soon as the first customer says, well, how's that relate to what we just worked on? Yeah, I'll look more into it. Yeah.

00:21:00 Mark Smith
So interesting talking to Erik, hopefully I get to meet you in person sometime in the near future on the MVP side of things. What do you do?

00:21:09 Erik David Johnson
I don't know where to start, so of course I'm active in the in the in the Asia Community, I went to Sweden recently because they felt they didn't have sort of. A. They have limited experts on on last language models, so I I researched. I did research in language. 15 years before Chatt came about. So I've been working with that for, for. A long time. So that's why I got the the expert status. How how are you an expert in a new technology and that no one just didn't know before at a time? Well, I actually in my research I predicted what we're seeing with with large language models. Like 19 years ago, before deep learning and Transformers were even invented so. I'm proud of. But uh yeah, it's very it's available at my website if anyone wants to read my research, you can be among the first five to ever read it. So so I'm participating in the Asia community just as a large language model expert. And also of course, following very intensely what are the features in the Asia I services domain, how can they be used in different ways and really interested in personally really interested in code interpreter. What that means for for use cases and applications of? And and of course the agentic aspect as well. Then I also have a a column and and then Mark's biggest computer magazine called Computer World. So it's it's a. It's a fixed column, so it's it's the online and printed version where every month there's a new article from me about some topic relating to. Asia I and large language models and copilot and so on. And then of course I'm out talking to a lot of people all the time. A lot of it is actually paid, so it doesn't really support my MVP status but but, but also some some unpaid stuff in my which I do myself in my free time. I'm also a mentor. Uh for AI mentees? And what else? Uh, yeah, helping the government. Uh, sort of do this right. Uh, very instrumental. Well, and the the IT industries new tech Policy Board which gives the recommendations to the government and also taking responsibility of should Denmark have a national large language model. So where it was a big part of of drafting that that that legislation. And also the doing policy briefs right now sort of overcoming the the legislative part of can we get the Danish data and put it into large language models? Sometimes the, sometimes the politician sort. Of. Text me and I'm like, how did you get my number? And they say, can you help me draft something or explain something right. That's so cool. Yeah. I'm very lucky.

00:24:01 Mark Smith
I love it. Thank you so much for. Coming on the show. Sure. A pleasure. Hey, thanks for listening. I'm your host business application MVP, Mark Smith, otherwise known as the NZ365Guy. If you like the show and want to be a supporter, check out buy me a coffee.com/nz365guy. Thanks again and see you next time.