Webinar: How to run an AI-powered Innovation Sprint

All right, let's get going. So thanks everyone for joining today's webinar on how to run an AI powered innovation sprint. Hopefully you found the QR code on the right. If you want to participate, add any questions or comments, you can do it through the Slido channel channel or on the YouTube channel in the comments. So while we were waiting, we had a few folks already answering a few questions.

Yeah, we just opened it up, so people are getting in. Yeah. First question, what are you hoping to get out of today?

So we'll give it a bit more time so that everybody can answer a few questions. In the meantime, potentially, what I can do is I'll quickly give an introduction to who's here. In the call from our end, you have me, Vincent, I'm partner at Board of Innovation. And then we have Leonard, who's a senior consultant and AI lead at Board of Innovation and very happy to run you through the webinar on how to run an innovation and AI powered innovation Sprint. So when we take a quick look at a few of the poll results, we see ideas and inspiration, a different perspective. It's very nice. I think, given that there's thousands of people that are already subscribed. The answers are, are coming in as we speak. So that's very nice.

Good.

And we'll, we'll definitely have a look at all these and get back to you on, on this once. Yeah. Later in this, this webinar. So what's, what do we have for you today? On one hand, we'll start with AI for corporate innovation, then we'll dive more specifically in AI powered innovation sprints. Then the process in the tool stack that you need to do that and then the final steps to run your own AI sprint. And then at the end we'll share a few links to some interesting sources that you can use to do it for yourself and we'll definitely open it up for a Q and A that, as I said, you can go through the YouTube channel or through the slido questionnaires. So let's go into the creative AI for corporate innovation topic. So I think it's clear that AI is making very big waves. We've seen it, yes, in the yesterday's results of Nvidia, but as well overall, in certain studies that have been released, it's been said that it's very likely that it'll be a multiple, that the impact of artificial intelligence will be much higher than potentially even what the Internet did to us. So the productivity increase that we're looking at is massive. And if we take a zoom into what this would mean for innovation and creativity, you can see a lot of examples like Ikea and several startups doing work using AI. Yet the big moves are, for example, where you see Microsoft's GitHub Copilot, where more or close to 50% of its code already being used or developed by generative AI, you see that the productivity and the speed at which things can be developed is much higher. And also in, for example, video gaming and other areas, you see a lot of potential. If we click through on the topic of innovation, and actually maybe more specifically on the front end of innovation, you can actually see that there's. We're definitely looking critically at this as well, is how powerful is AI? Is it hallucinating? Is it actually giving you the results that we're looking for? Can we make decisions based on what AI? AI gives us answers. And there's plenty of studies being released these days. And for example, there's one that's really interesting on the usage of large language models for ID generation and innovation. And for example, that Cornell University study that was released earlier this year shows that the average quality of ideas generated by GPT4 is higher than those by humans. When you rate it on purchase intent. And that's a classical question that we get, is like, are the ideas actually going to drive purchase intent? Are they actually the quality that we're looking for? There's even studies showing that they do. And on top of that, you can even see the impact on synthetic Personas and allowing us to do virtual testing. So when we're doing or experimenting with actually using virtual users to give feedback on concepts or to do qualitative interviews, the question is, again, is it a fact or is it just a perception? And there's a study from the Brigham Young University saying that language models can be seen as an effective proxy for specific human subpopulations in research. So the answers that synthetic users give us are very close or similar to what humans, large sets of human population, would give us. And that's absolutely interesting. But then the question is, what do I need to do with this? If I'm in a large organization running innovation, what is the role that I need to do? And where do I need to guide our company or my team into? How do I need to put them into the right direction? I think it's important to note that as an executive, you have the choice to define where you are in that maturity model. We've defined five different levels or maturity levels, where you start from all the Way human only innovation all the way to fully autonomous. And to give you some examples, if I would take the topic of empathy and understanding consumers, human only would be me doing an interview with a consumer and better understanding what their opinion is. If you look at the complete other side of the spectrum, you're saying fully autonomously generated service. So meaning that the interview script is written by AI, the interview is run by AI, adds questions based on the responses of, of the, of the respondent, and in fact the interviewee is even a synthetic user. And the outcome and recommendations would lead directly to ideas rather than necessarily humans having to curate the insights that come out of it. You can see multiple levels in between and what we're talking about here today is definitely in that middle field, like level two, level three, where it's AI powered. So we're using AI to the maximum amount we can to augment and speed up a lot of the work that we do in innovation sprints. And with that, it's important to note as well that there's multiple ways on how you can implement generative AI in your innovation processes, all the way from more experimental testing out what it means towards scaling it and implementing it across the whole organization. And if we're looking at this scheme, those six different quadrants, this webinar is really focused on that middle field of doing a specific sprint, focused on a specific project. But given that we advise a lot of organizations on multiple fronts, very happy to in follow up webinars maybe go into the strategic side of how to apply AI on a more strategic level or very high level run keynotes that explain the benefit and the value of AI. Before I go into the details of the use case of an AI powered innovation sprint, I'll go two seconds through Board of Innovation's introduction. For those that don't know us yet, we're Board of Innovation. We're global innovation firm that imagines tomorrow's products, services and businesses and creates them today. And we have a team across more than 20 countries to make the impact at scale that our customers require us to do. And we join forces with the world's most ambitious businesses and help them to maximize the value of AI and to launch those new products in the fastest and best way possible. So let's click through on the AI powered Innovation Sprint. What the best and easiest way to describe an AI powered innovation sprint is with a very simple quote which says like a sprint without AI is like running a race without shoes. There's so much technicalities, there's so much tools out There that really allow us to do things that are fundamentally different from a regular sprint, which really gives you a competitive advantage that you haven't seen before. The results of what you can achieve with generative AI are so much bigger than what you have without it. We're going to give you a glimpse of not only the process that we'll go through, but also the exact tools we use and what you need as a team to get going. If I dive a little deeper into this, I can highlight a few elements on what is that design Sprint, the original design Sprint, the five day format in which you go all the way from idea to validation of an idea and why we actually decided to actually augment it with AI tools. I think at first it's important to note that in general, blocking a set of five days specific group of people into one room is quite a big commitment, especially if you want to make sure that that results into to impactful solutions and that in the end it can really be implemented. And the issues we see down the line is that in the early stages of such a sprint, you often lack some quality on concepts. So the viability, the visual design of those concepts are very early stage, which makes it hard to evaluate one against the other. And that actually also leads to selection bias. A lot of it is done by humans, be it within the room, be it within the strategic decision makers. There's a lot of selection bias, whereas that's something that you can definitely take away using AI. And there's a lot of overthinking. Once you've made a choice, what idea is the best, you select the one that you want to go forward to and you rethink every feature, every product or service feature that you, you plan to test out with consumers in the end. And then finally you end up with although is it clickable, is it a visual prototype, it's still lower fidelity. The fidelity of, of concepts is so much higher using generative AI. And on top of that, maybe even more importantly, there's no quant backing. A lot of it is still based on qualitative feedback, just a bunch of consumers, a subset of consumers sharing their feedback, which is very valuable for innovation. But if you want to make a global decision on moving forward with an idea, a product or a solution, you do need quantitative data to really back that up. And that can really be enabled using generative AI. So while we are launching, while we're waiting, I'd love to get more answers on the question that we've raised for you on which aspects of a Design sprint do you think will be most impacted by AI? I see that the variety of ideas is definitely one of the elements that some of the people on Slido have have added. And the prototype creation based on user preferences. Absolutely. Interesting to see.

What is it for you, princess?

Sorry?

What is it for you?

I think it definitely starts very early in the front end. I think the idea generation is absolutely most impacted by AI, but it's interesting to see the dynamic of the prototyping itself also being a very big one. So the visualization, the quality and fidelity and speed of design is just completely different. Super nice to see so many respondents sharing their answers and would love to get more even throughout the presentation and we'll share it back afterwards on what the final results were. So with that some of the main benefits of AI powered innovation sprints what we see is on one hand, it removes bias. The reason why we highlight this is that instead of making a decision within a small room and having a few folks that did a deep dive on previous research, actually all available data is considered in the selection of ideas, in the creation of ideas and in the synthetic testing of concepts. So all data that is previously generated is reused for ideation, decision making and concept design. So it really moves, removes bias on decision making, on what idea is best, what idea is good or will perform well. Second, it improves the quality because the quality of concept writing and visualization gets really drastically improved. The speed at which you can get to a very high level, high fidelity concept is much, much improved in terms of speed and quality. Then you can move faster because you have so much efficiency gains at every step of the innovation. The result is really that fast go to market. Second, you can look much broader, generating more options and creating springboards become allow you to really go much further. So instead of a traditional sprint where you want to go for identify the most creative ideas, in this case you want to just go for the most quantity of ideas, the highest amount, because the AI generated concept will go much further than sometimes what a human brain would come up with. And all that's the speed at which you can move allows you to iterate much sooner and you can test those concepts much earlier than you would in a regular setting. And finally, the role of innovators is absolutely to think harder. And as executives as well, you'll get results, you get all the data in a heartbeat. And that really forces you to make the right decision. You have everything at hand, so you're forced to make the right decision on the spot. With that I would love to transition into the pillars of what an AI powered innovation sprint is all about. There's three main pillars. It's very simple. It starts with generating insights ideas, then you validate them and you curate them. So it's a circle that goes from concept generation to then insights capturing, curation, which ones make most sense, and then a decision to further generate the next level of our next set of concepts. That circle is iterative, as I said, and can be done powered by AI, so each different element, the generation of ideas, validation of ideas, and curation. But there's even a next level where actually you use AI to run the full circle. So it generates the ideas, it validates them automatically, and then it curates and selects the ones that are the best and it really gives you recommendations on which ones to go forward with. In this case, the AI powered innovation sprint really breaks it down in each chapter, allowing you to really go with the flow and truly understand each of the steps separately. So that brings us to that simple yet powerful process where you start with really data reviewing and data and model training. So you want to use as much information that you have and input it in your own large language model, allowing you to train the data so that during the sprint, everything is set up to get the right results, to do the right evaluation, and to get you going. The second step is the actual sprint. You go through those three elements of first, the generation of an exponential amount of insights into a summarized subset, and second, an exponential amount of ideas that will be generated in the process. Second, you go for validation, validation through synthetic testing and synthetic design. So you create new concepts really fast and of really high quality. And finally you curate the most, the best ideas. You do high fidelity concept design and you actually bring those concepts to the next level so that they're nearly ready for that go to market. Or if you're not satisfied with the results, you can obviously iterate again with that. I think it's most interesting and probably also a big reason why many of you have joined this webinar. We'd love to give you a true deep dive into each of the different steps and the tool stack that is needed to make it happen.

Yes, let's dive into the meat. So we have activities, we have tools, and we have of course, the people, the team that should make this happen. Let's maybe start with the first one, the activity. So to put some more meat to the bones, what we mean by training, everything that comes before is actually you want to gather all your research together, you want to gather your trends Your market entire, your user insights, your competitive intelligence, you want to put all of that together to basically be able to tap into that during the three day sprint. Great tools for that are also mentioned, but we'll dive more into tools later. As I said then the Sprint itself is actually kind of like very iterative loop of generating, validating and curating. And that will take place a couple of times throughout the three days. But let's say day one is mostly generation where you basically immerse with your insights that are already there. You basically built or trained your own large language model and you want to tap into that and you want to fully understand the market, you want to nail the problem and you want to generate jobs to be done for every opportunity area that you might cover. And then using Genai tools you'll actually quickly get to one hundreds of ideas with the right prompts. So more about that later as well. So you have this short list, you have this long list and then on day two we'll actually get to more shortlist, type of number of IDs to cover and those you'll get with synthetic users also more about that later. But basically you want to just keep something that is manageable and then you want to basically build on this. You want to listen to synthetic users and you want to get their feedback to refine and also to prototype, prototype, make visual representations of what your idea is about to then on day three to be able to put it in front of real users and get real consumer feedback and iterate with them as well. You'll also do concept writing, great tools for that as well or Jasper AI and durable. You want to do concept writing, refinement, you want to work on the name, the description, the claims, the benefits, especially in cpg. Also the reasons to believe you want to do a cross check with your competitive positioning, where do you end up? And yeah, actually in two days you already get to such a high fidelity prototype that this is something you can proudly take to your real consumers to then basically iterate or go to market after already. You want to capture all your learnings from the sprints, of course, the process, but also from what you learned with regards to your concept. And you want to work on recommendations and next steps. Now diving into that day one and also getting a bit more into tooling. First you want to basically study the markets and you want to do that with tools like GPT4. Nothing new here, but it will allow you to quickly get a view of the market, especially if it's a new market using Pestle analysis, SWOT analysis and to get a basically an understanding of what market you're playing in. You also want to augment this with synthetic users that can help you to really nail the problem. They will shortly actually allow you to run custom scripts. It's still in beta, but we're trying it out and it's actually quite nice. So you can cover the full. Yeah, your full demographic and within that you focus on different users and you see how they would react to the problem. To make sure that you nail the problem, but also to potential solutions. What you also want to do before diving into ideation. So building that large language model with everything I've said before, but also in B2B there's actually a massive pile of data like annual reports, like earning calls, like Q and A sessions with stakeholders, all of that is readily available. You can basically get that from your suppliers, from your customers and you can feed that to your large language model. What we recommend is for example GPT 3.5. We've actually found it in fine tuning it with your own data that it would be, and this has actually been confirmed by OpenAI earlier this week that it would be more precise and less risk of hallucination if you use 3.5 and not GPT4 to train it with your own company data. You can also use Claude2 which can handle up to 100k tokens. So that's also a massive input that you can give it. Or Chatbase is one of the other tools that we've been experimenting with. And then you basically want to find patterns in your data, of course. So you have this bot, you have this huge pile. But we use for example Displayr, which is a great tool for needs based segmentation using K means clustering or we use Accio, which is a machine learning tool by the way. All of these tools are on our website and we'll share that link as well. So that's coming. What else, especially if you're in cpt then this tools to do social discovery. So what are people blogging about? What are they posting about? What you see here is como AI. And also tastewise, and Tastewise is actually reading what people post specifically about food and bev. And so in this example it for example points out that Vietnamese coffee has been trending lately on social media. So if you're a coffee company, good inputs to work with, of course. And then you basically start whenever you feel confident that you, you understand the market. You start with generating IDs. And here as well, we're actually still turning to GPT4 and we noticed that a lot of ideation tools in the market are working with GPT 4 or 3.5 as their source. So we're trying a lot of different prompts on GPT4 and that will allow us, or that will allow you as well to generate quickly 100, 200, 300, 400 ideas and really get to that long list. And whereas before there was always this selection bias and I've seen a question on that in the chat, we'll get to that as well. But before there was always this selection bias of people just picking their own favorites and it didn't matter if you went up to 100 or 500 ideas, people would always turn to their own preferred ideas. But now we're actually convinced that quantity will lead to better quality. And why is that? Because again, synthetic users are great to do that first selection and it might not be there at the point where it calls out like These are your 10% most promising ideas. But at least it's already at the point. And all research or past projects have shown that putting this next to real user data, then we found that synthetic users are actually consistent in identifying the worst performing ideas. So they can already remove that bottom layer of worst performing ideas, making that long list shorter and shorter. And I honestly believe it's only going to get better. And yeah, it's really going to help you to focus faster on the ideas that do count. And one other thing that is important here is actually GPT4 has been proven to be more, more efficient and generate even better quality ideas when you fuel it back with your best performing ideas. So the ones that were best performing in the synthetic user analysis, you feed it back and you, you actually get even better quality ideas by, by feeding those back to your prompts. And there are a couple of tools like here we have Storms AI for example, a couple of tools that really tap into that space of like you double tap on an idea and it will generate 10 more just by click of a button. And then at some point you're done generating ideas. You're also done refining them with synthetic user feedback. And you want to work towards that third day where you actually invite real users to the room and then you need something more tangible so you start prototyping. And yeah, this is also like easier than ever before. But we also want to guide you a bit with prompts. So if you, if you don't feel confident making your prototypes, but also not confident making your own prompts, there's websites like promptbase,.com where you can get a lot of inspiration and they also tell you the prompt, sometimes it's paid but for a small fee. They will also tell you the prompt that goes into it that can get you to make similar high fidelity visualizations. Some other ideas are actually Midjourney's describe function where you give it a good prototype and it will actually guess what prompt goes into that. And then you just have to iterate on the existing prompt and you can make your own prototypes in the same style of that specific visual. Where I touched Taste wise briefly a couple of slides ago in the Insights, they are also now working on Taste GPT. So these insights are being connected and are actually leading to their own visuals. So imagine you're this mozzarella panini company and you, you basically want to find out what's next for mozzarella panini. It already connected it with garlic bread that is apparently like trending on social media and it comes up with this new concept that, that. Yeah and a visual already that makes it all very tangible to put in front of customers. And then the third thing that we're also exploring already actively with clients is since we have that powerful loop of data going into a prompt and then making a new visual, you can actually get taste test results of how your product actually tastes, how people actually would describe it and you can feed that back to your prompt to then create better visuals for the package of your product. And then for the first time ever, you would actually optimize your products for retention, not for acquisition. And then on day three, all is left is actually refining your your marketing content and putting your product or visual or prototype in front of real users. And whereas before we were playing on that third and fourth level of AI actually talking to humans and to synthetic users, now we want to go more high fidelity at least for now. And we want humans to have conversations with humans and to learn like how do they really like what we came up with on day four. And meanwhile I've also opened the poll to get your own tools and it's a word cloud. So this is telling me copilot orchestrate. Yeah, I will need to check the. This phase doesn't exist. It's a great tool indeed. I will need to check the raw data to really see what's here. Microsoft is also mentioned a couple of times. Miro also great, great, great AI features. That's true. And notion as well. So synthetic uses as well. And I know the, the team is present here today so they will like that check your script.com PDF AI great tools, great tools. Thank you so much for this. But of course, tools are just tools and yeah, you need more, you need an amazing team to make that happen. So also what we suggest for you to succeed in this sprint is a tool that is consistent out of some familiar faces like a project leads, of course, a facilitator. Especially now with the AI, teams are making different speeds, so facilitation becomes more than more important than ever. And of course your R and D people, your packaging people, you want to get keep the pulse of the market and the pulse of what's possible throughout your sprints. You want marketeers, of course, to be present and you want visual designers that now will transform into visual prompt designers because the prompt is more important than ever. There's also AI data analysts that will help you to design your experiments. So yeah, those will actually in a tandem run most of the AI testing. And then of course it should also be someone overviewing this in the room, legal and AI ethics person that keeps an eye on everything that's going on. And what you want to do throughout the Sprint as well is to establish a prompt library with prompts that specifically work for your team, for your industry, for your company and those who want to keep and be sure that everyone in the room can actually access them and use them onwards.

With that, we'll help you out in a few simple steps on how to run that AI powered innovation Sprint for you. And the summary is really simple. You have four steps. First of all, you select the case and the AI Sprint team. So as you've seen, not a massive team, a small team, but enhanced with an AI prompt library and the AI data scientists that are required to analyze and to build the right prompts to get to your answers. Then second, you build that AI tool stack and prompt library, as I mentioned, everything to be really prepared to do your homework, let's say. And then you collect all relevant data and train those models so you make sure that everything's set up so that in a short amount of time you can run through different iterations. Then you run that AI powered Sprint and document those learnings because that's super important, so that you can then do later on, potentially in that next level, a more strategic exercise on, okay, what is the potential of such a methodology? Are there more ways on how we can do this at different parts of the process, not just in one sprint, but actually throughout the entire R and D process? For example, maybe one question that came in through the chat as well was like, why do we do it in three days and not just in one day. I think it's a very good comment. I think it all depends on how thorough you want to do it, how well prepared you want to do it. I think it's very possible to take insights, come up with ideas and visualize them straight away in one cycle, in one day. Something that was very difficult in the past, but very possible now with AI tools. But if you really want to make thorough recommendations on, for example, whether or not to further scale an idea or solution, you do need a few days to do a few iterations, to do a few strategic checks internally in your company to make the decision whether or not to move forward with them. So four simple steps. They're also very well described on the website and that's maybe a good segue into that next part. If you want to know more, feel free to reach out to us. On that website you find all the detailed steps and links to all the tools, but you can also reach out to us if you want to have help or you want to run your first sprint with the expertise of some of our facilitators and AI experts. Beyond that, you can definitely get access to all the tools that we mentioned here today, but also many, many others. We have that curated list of our top AI tools to run that AI powered innovation sprint. You can find the URL here at the bottom, but also, yeah, straight away at our website. And beyond that, I can definitely encourage everyone to also tune in for the upcoming webinar which is fully focused on autonomous innovation. So it's a more forward looking future perspective on where we believe innovation might be heading, where it's not just where human curation might become less important and that I can really run entirely by itself. With that, we would love to thank everyone for participating. It's nice to see thousands of people really tuning in today and would love to take a few questions and Leonard and I will go through the different questions and pick the ones that come most often, for example. So thanks everyone for tuning in and we'll go through the Q and A straight away. Maybe Leonard, you can kick it off with answering a question.

Yeah, my pleasure. Thanks Sina as well for curating because there's a lot of activity in the chat, so we haven't been reading everything just yet, but one of the questions that popped up is like, how is there not a selection bias in the AI large language models themselves. And I would say yes, it's definitely a catch too. AI is trained on historical data, so there's Always like a bias still present, but like selection bias. What we mean is specifically that a human decides what idea, or one human would decide what idea would proceed. And what I see, idea should be killed. And that for sure will be significantly less if you give the lead to AI, which also doesn't mean that it's a black and white story, of course. Like of course it should always be creation and this should always be someone looking at the results. And if you really believe in an idea that it should proceed, you can always put it in. But you put AI more in the lead of doing that. Filtering, especially filtering out the worst ideas. Yeah.

There was another question on pure AI. Help us to kill an idea faster if it won't fly, or can we simulate the adoption through synthetic testing? And Leonard already highlighted quite a few elements of this, but the answer is absolutely. You can actually train using Personas and historical interviewing data. You can train a synthetic user, a synthetic Persona, to actually answer whether or not it has a certain buying intent, whether or not it likes certain features. So you can definitely really filter ideas using synthetic testing. So. So the answer is absolutely yes. And at the moment we do recommend to also do human testing. So we definitely have actual humans giving feedback through social media, but also directly, but. But the data also shows that.

It.

Is giving very accurate answers, both from a quantitative and qualitative standpoint.

Yeah. Another great question, by the way. Why one week for an ideation validation and prototyping loop? Why can it not be done in one day? And actually it's true, it's still speeding up. So for now we recommend three days and to really take the time. But let's talk again in six months. For now, we believe that if you want to do it properly, that three days are still needed.

There's one question as well on normally in a regular Sprint, people get their hands dirty and how does it change while using AI? Do you teach people how to use AI during a workshop, for example? A very valid question, something that we thought through quite a lot as well. We get a lot of requests on can we make sure that you're building a capability, not a dependency on the pure tech side? One element is yeah, for sure. I think it's all about documenting learnings, building your own prompt library, building your own AI, our toolkit and playbook, so that people can really repeat the process. Second, yeah, it's obviously about informing and educating people on the use of AI, the possibilities, but also the downside. So there's definitely an educational element to it. That being said, there is we do have a strong belief that the role of an innovator and anyone that is involved in innovation is transitioning from creation. So coming up with ideas, for example, scripting an interview and doing an interview yourself towards curation or editing things. So the role of humans in this setting is different. So you create a clear brief, designs are done in a heartbeat and your role is to select the concept that is the best. Your role is not to come up with the idea. So it's just a very big shift. You're still very involved Often actually we do more than in the past in person because it's really good to experience and see what it's like to use AI. But the role of humans is definitely shifted towards more editing versus creating.

Yeah, perfect. Someone else asked, do you have a case study you can tell about? We definitely have. So there's one on our website. If you go to boardofinnovation.com creative AI, then you'll definitely find it over there. One of the sprints that we recently did. There's also more on slido itself. So the top photo question there. How do you address confidentiality concerns within ChatGPT? How do you make sure nothing proprietary is uploaded and used for trading? Good question. And it's still evolving. It's still like even the matter of who's owning the ideas, it's still evolving. So we're closely monitoring that. But we believe that there's no need actually to give away like for example, your brand name. Leave it out of the prompts. There's no need for that. You've seen for example, one of the prompts that we did before the mid journey one with the McDough McDonald's hamburger. You will also see if you looked at the prompt above, we never mentioned McDonald's. We basically added that later. So there's no need to really get confidential in those prompts. So for now I think that that's the best and safest workaround to go. Hope that answered the question.

Maybe one other question that came up was are your clients comfortable with giving away the decision power of selecting ideas by a machine? And I think the simple answer is not yet. So I think you can really compare it to, for example, guiding us in the right direction. Several years ago, everybody would like to pick up a map and decide for themselves. These days we all trust Google Maps to give us the right direction because it simply knows better. And that's the exact same situation that we're in right now. We're in a position where we still trust human Decision making better because we don't know the depth and the quality and the freedom that the algorithms have. But once we understand better, there is definitely a route where decision making should be fully trusted by machines or AI. And to give you an example, how we often do it is we do parallel sprinting. So where we do have the traditional sprints, or actually longer several month tracks, and then we replicate that same process in one to three weeks. So typically a process of six months, we do that in three weeks and then we compare the results and, and then the decision making, the final decision making is still up to the company, but, but at least you have, let's say, neutral data that you compare what AI did with what humans did. So a very powerful way that we're really monitoring closely to, to really build up the case study ourselves.

Someone has been asking what about corporate B2B clients? Since this was maybe very focused on, on, on, on Food and Bev specifically, I think the, the impact can be great there as well, and if not greater because the starting point is just like it's a slower moving industry, if you then adapt to those tools, you'll, you'll quickly get a massive competitive advantage. And I'm not sure if this question was raised before or after I shared, for example, how you can pull up annual reports and earning calls and all of these in like a large language model, but that's definitely a way to go. So that's definitely something that should be explored further to basically try to mimic your corporate B2B clients. Because even if you're B2B, you're in the end selling or buying, selling from a human, selling to a human that has a certain goal, role description, pains, gains, jobs to be done. So to that extent, it's not that much different.

Yeah, and on that front I can build on it. We've done a lot of work using AI in a B2B context, also in a pharmaceutical context. And one of the benefits even of using, for example, synthetic users or using it in that context is that often you don't have an easy access to your customers. It's a super small subset of people. Or you're simply not allowed to because there's barriers in between contacting people directly. You can actually train a model to synthetically represent that individual and therefore you can actually get answers and perspectives on certain topics without every time having to go to them directly. Obviously you still need to do it in the end, but it's especially an innovation process, super powerful to get continuous iterative feedback and that can be done synthetically. So definitely plenty of opportunities to do it. Plus concept generation and ideation. Everything that front end is powerful to use AI. And then finally as well, the large amount of data that are captured in the supply chain and other parts of the process can be reused to for example, simulate whether or not a product would go smoothly through the current supply chain or whether it would not. So absolutely plenty of opportunities there.

Cool. Let me pick another question someone's asking, like what AI and non AI skills are we specifically hiring for? Well, non AI, you should talk to our recruiters. But AI, I would say curiosity, as in the space is so rapidly evolving. Remember that you're only six months or 12 months ahead of behind the way you want to see it from the top experts in the field. So, so it's so it's so rapidly evolving and the tools that we presented today might change again in a couple of months. So be curious as a corporate innovator.

Awesome. Unless if we have any, any big questions that we've missed, I think we, we could potentially wrap it up here and then we'll yeah, we're super happy to to answer any follow up questions directly through Leonard and myself, but mainly would recommend to make sure to go through the website. The Creative AI page collects everything in one place and then through that you can go through the AI Sprint page, the AI tool stack, the reports that we've written, and definitely plenty more that will come soon, so don't hesitate to reach out. And thanks a lot for tuning in and looking forward to see you in the upcoming webinar on autonomous innovation.