32 | A Roadmap to AI Business Impact - Defining and executing successful AI projects that drive business results, with Khalifa Manaa CEO of Shapr

0:00Hello, and welcome to Leveraging AI. This is Isar Metis, your host. And in today's show, we are going to explore one of the most critical questions that business leaders are asking. themselves today, which is how to pick the right AI project and how to successfully execute on them. And we're going to do this with the help of Khalifa Mana, who is the CEO of Shaper, a company that helps businesses from different industries in different sizes select and implement AI solutions to improve their businesses and improve business results. At the end of the episode, I'm going to share news from this week.

0:37There are a lot of really important news this week from various aspects of the AI world. This week's show is brought to you by Multiply AI. Multiply is spelled M-U-L-T-P-L-A-I. So just like multiply, but instead of Y in the end, AI in the end. And if you go to multiply.ai, you will be able to find any kind of education you want about AI for your business from this podcast through courses and in-company consulting and training and so on. So go and check it out, M-U-L-T-I-P-L-A-I dot A-I forward slash services if you want to see everything that they have available. But now let's dive straight into how to pick the right AI projects and how to successfully implement them to achieve better business results.

1:30in the next few years AI technology will change our world dramatically whether you are a business executive trying to catapult your business forward or just somebody who refuses to be left behind and want to advance your career this is the show for you I'm your host Isar Maitis a serial entrepreneur and an AI enthusiast you'll hear invaluable practical tips from innovative business leaders AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

2:11Hello and welcome to Leveraging AI. This is Isar Maitis, your host, and I've got a really exciting show for you today. Everyone now, literally every business leader out there right now, or at least everybody that has been hiding a rock for the last eight months, is trying to run AI project in their business right now. Why? Because everybody feels it's going to give them a benefit and an edge over the competition. If you're going to do it faster, then you can get an edge before your competition does the same stuff. And it's true. The problem is very few people actually know what the hell that means to run an AI project for their business.

2:47And even less people know how to actually do it right. Our guest today, Khalifa Mana, is one of those people that actually knows how to run this project. Why? Because that's what he has been doing in his company, Shaper. He's helping businesses define and launch successful AI projects and products. And it's not my surprise that he's doing it because Khalifa is a serial entrepreneur himself, his main expertise is product launches, like tech product launches. So he, for years, has been in the business of defining projects correctly, launching them successfully, and applying different processes and systems and mindset in order to make that successful.

3:34All he's doing today is he built a business that does that in launching AI projects. And he calls it running impact-first AI projects. So that's what we're going to talk about today. We're going to talk about how to define, launch, and run successful impact-first AI projects, which, as I mentioned, is probably one of the most talked about boardroom and C-suite topics in the world today. And hence, I'm really excited and humbled to have Khalifa as a guest of the show. Khalifa, welcome to Leveraging AI. Awesome. Thank you, Saar. Super happy to be here. I love the pod. I listen to almost every episode.

4:17I've skipped a few. I'll go back and listen to them. And like you said, it's a very interesting topic because after all this hype around ChatGPT, everyone's thinking, well, okay, how do we add this AI to our business? And like you said, there's plenty of ways of doing this. Some of them lead to successful results. Some don't. So that's what we're going to be talking about today. I love that. And you and I had several conversations already, and I know you have a very well-defined process and well-defined strategy. But let's really define first, what is an impact-first AI project? Yeah. Okay. Let's start with that.

5:00This is something that is close to my heart. when we started Shaper, we actually had, we defined four principles because I tend to believe that for you to build the right kind of company, the company where you want to be and you can be proud of, you have to define what that means, what are the core values. The first of those four impacts is impact first. What this means is that we say motor projects where we simply don't see an end result that will create an impact. What I mean with impact is basically significantly improved customer experience or growth in revenue or cost savings. In most cases, we want it to be a number.

5:43And so when we, I'll walk you through the process in a bit, but one key part of this process is after doing ideation, what's possible. We turn this into these idea cards and we set a number of potential unlock value to all. this could be something like a process becomes significantly faster or it could be that we aim to increase revenues by 20 percent it could be that we aim to win market share in a specific segment or something like this but when impact first by the way it can be also some a positive impact for human beings or the earth i'm very happy we're exploring something with wf which could be because I'm a kind of a sucker for nature, so I'm very happy about that.

6:31So in that case, the impact is slightly different. But even there, we set a number to it. So something that we can measure and something that ideally will help the company to end up where they want to be in five to 10 years. So there's a meaningful metric that we can measure. We set the goal before, and then we can see how much impact were we actually able to get. Awesome. So basically, it's just good strategy and good project management. And instead of saying, oh, there's this cool AI tech, let's find a way to implement it. You do it the other way around. You're saying, I assume the first question you ask is, okay, we're going back to what you just said.

7:14Where do you want the company to be in three to five or 10 years? What can get you there? What bottlenecks exist today? and then you look for the right AI implementation to go there? Is this roughly the process? Yeah, yeah, it is. And since last November, a lot of companies are curious into how do we attach this AI somewhere. And it's usually they've seen someone else do something with ChatGPT and now let's do the same thing, which in some cases could actually make sense, but in most cases, maybe not. So our process starts exactly like you said, we start always from top down. So we want to start with, typically our process is five days.

8:01It's roughly based on, in my previous life, I used to do design sprints, which is a Google Ventures five-day innovation format. Amazing. By the way, as a side note, this is a bit off topic, but we're looking into doing ADI design sprints in product cases where we have a challenge in mind that we want to solve. Basically doing ideation with chat GPT or GPT-4, testing ideas with AI and so on, just making it significantly faster and less biased because usually people, we fall in love with ideas. Okay, so our five-day process usually always start with we gather around the core team for the sprint and then it's led by the leadership of the company.

8:48And it varies. In some cases, this could be the CEO. Sometimes it's a transformational officer. Sometimes it's a digital officer. But someone from high and top kind of communicates to us first clearly, where do we want to take this company? What do we want to, in an ideal world, what would happen in three years or five years? And this kind of gives us an idea of where do we want to go? And then we start to ask questions of what's preventing from that happening. And we tend to do it on the first day of this process. We have this session. So we have the core team. We have the leadership. It's very important to have kind of leadership.

9:32Ideally, a champion can then drive things forward and have them all in the same room in the beginning. So I want to pause you for one second. Who else is on the team? I agree with you 100%. you always want like a C-suite buy-in because then there's somebody who can put his stamp of approval on everything that was said and represent the vision of the company. But who else sits in that room? Yeah. Usually we want to have someone from products. Ideally someone looking at the offering level, not a product manager. Someone from customer service, sales, IT, marketing, operations. Basically the key segments that we might look into.

10:11Okay. There's some variation because in some cases, companies have already that this is the area where we want to look at, where we want to find challenges or growth. Then, of course, the team could be smaller and slightly different. But in an ideal case, what we really want to do is look at it holistic. Let's just look where do we want to take this company and then look at how could we make that happen. So these are usually the people we want to have in the team. activities that we usually do are in this first session they're design thinking tools they're interviews with other people making notes of how might we and this all just is fuel for ideation later on this usually takes half a day and then another half a day to do documentation to give you an idea of what this could mean in a banking customer this ended up being that they They want to kind of transform their, in a way, their brand image to be, they've been, they've had a very traditional image.

11:19They want to freshen it up. They were planning to do a kind of a rebranding, like a facelift for their brand. And they want to be seen as more digital ready. And then they had in their wholesale or business banking, they wanted to increase their market share there. So this was an outcome of where would they want to be in five years. They want to be seen as a digital modern bank and they want to increase their market share. So this is awesome because this leads to a lot of follow-up questions because this doesn't sound like an AI project, right? This sounds like a traditional business strategy decisions of this is the decision we want to take the company.

12:02how do you turn that into an AI process? What are the follow-up questions a company needs to ask itself? Because this is a perfect example, because it doesn't sound like an AI kind of thing. And yet, I know you've done this project. So let's continue from here. What are the follow-up questions you need to ask yourself as a company in order to define, A, whether this is an AI project, and B, if it is what the project looks like. Yeah, that's, by the way, a really critical point. A big chunk of this process is not AI. Of course. And I haven't found a way to get to a good outcome by skipping this, like without doing this part.

12:46It's just the other road seemed to lead to a solution first thing. And then we fall in love with the solution and the technology. We forget about why we're doing this. Usually no impact. So a key part of making sure that that happens is going through this route. By the way, I agree with you 100%. I do the same thing. So I, in addition to the podcast and the courses and everything, I do consulting for businesses about how to implement AI. And it's the same exact thing. The beginning is traditional business consulting. Let me understand the business. Let me understand where you're struggling. Let me understand your bottlenecks.

13:21Let me understand what you're trying to go. Let me understand what resources you have, what resources you're missing. What are the gaps? So I agree with you 100%. And I think that's probably the biggest problem companies have today when they come to implement AI is exactly this, where they want to jump into the tech side because it's exciting and it's cool. And like you said, they see other people doing it. And yes, you can solve a small problem very quickly, meaning there's ways to harvest low-hanging fruits. but if you want to change the trajectory of a business like strategically and leverage AI to do that it's I would say 75 % traditional business strategy and then using the other 25 % to apply AI on how to get there and how to close the gaps but this is exactly where we are right now in the process you're describing what do I need to ask myself in order to actually go the next step Yeah.

14:20Awesome. So in our process, the next step is usually looking from inside out. So the third step is bottom up, where we then go into practicality of what needs to happen. Inside out means that we're looking at, now that we define what we want in the big picture to happen, now we go into looking from the customer's point of view. How does it look like today? This can mean sometimes we just do testing, whatever is interacting with this company. Sometimes we do customer interviews. In some cases, we might even do shadowing. So basically go see how their customers are interacting with this business.

15:02Some companies do this in-house and they do this periodically. And then we can leverage the knowledge they already have. In this banking case, for example, what we discovered really quickly was basically the pain points that enterprise customers, when they want to open a bank account, firstly, it takes really long time, up to 100, 120 days. So three, four months. Oh, wow. From the moment that I want to open a bank account till you actually get it. And then when we spoke with a few customers who recently got their accounts open and joined, they all had these stories of how initially they felt like they gave a lot of information and documentation.

15:45And then two weeks later, they were asked about the same stuff again. Then a few weeks later, again, then they had no visibility of what's going on. Then finally, three or four months later, they get an email that, hey, you have a bank account. So the experience is really lacking compared to what the expectations are. Because we asked that tool, what do you expect a bank account to be? And it is exactly like it is in consumer world. Digital front end, I give information, two minutes later, I have access to a bank account that I can start doing my business. So that's the expectation. The reality was very far from it.

16:23At this point, we know very clearly that, okay, for them to be able to, or one clear way how they can win market share is to solve this bottleneck. But then, of course, we have to figure out why is this happening? Why is this taking 120 days? Awesome. So let me summarize so far. Step one is gather the right people in the room, which is C-suite people plus e-decision people or key operations people from the different departments that are going to be involved. Step number two, define where you want to go and what are the gaps that you're trying to close or the direction that you want to take. Step number three is now go and look what's the perception or what's the reality check of the situation of that company by asking clients.

17:18So now you have the idea that the company has about itself. Now you also have the reality check coming from the real world or, okay, this is what we see with regards to those topics that the company is talking about. What's the next step? So then we look at bottom-up. Now we've identified, okay, what's the customer experience upgrade? We know where we want to go. And now we start looking into what's for us to win that market share, for us to get to wherever we want to go, certain things need to happen. What's preventing that? And this is where we ask questions from kind of people in operations.

17:58In this case, what's happening in those 120 days? what's going on. Walk us through the process very clearly so that we can identify root causes. We tend to map challenges. We tend to look at, for example, one exercise we often do is what would the 13-star customer experience look like? I think you're familiar. I made many listeners. This is an old Airbnb story, which I really love, and it actually worked really well in real life. I'll try to very shortly summarize it. It starts by what would be a five-star experience? Like one to five stars. What's a five-star experience? We map it out. And then we think about, okay, what's the three-star experience?

18:39What's the one-star experience? Then we know what really sucks and what they think is. Then we're like, what would a seven-star customer experience look like? In Airbnb's case, it could be that, okay, the seven-star is when you enter the apartment, they come with the champagne bottle. They have a personalized sign. And what would 10 star experience, what would the 13 star experience? And then it gets into the territory of Elon Musk's private jet to pick you up. You end up in Mars and you have a party there and so on. But this kind of gives us an inspiration. What could be possible if everything was possible?

19:17And that kind of is, again, fuel for where do we want to get? I want to touch on that because I love that point and I think it's critical. In this point of the conversation. One of the amazing things that I enabled us to do today is to really think outside the box. And what I mean by that is we are as humans and especially humans within a structured environment like a business. we think of things through existing lenses that we were trained to think through for decades. Why? Because for decades, we were trained to look for process efficiencies. How do I do this process that is the given better, faster, more effective?

20:13The reality today that AI enables us in some cases to skip the process and get the outcome. And get the outcome. I'll give a simple, stupid example. Let's say you want to do an SEO project. You need to go and research clients' websites and see which keywords they're ranking for. And then you need to do the analysis of how hard it is to rank for these keywords. And then you need to figure out topics you want to write about. And then you need to define the categories. and then you need to define the articles and then you need to actually write the articles and then you do need to put them on the right places on the website and you need to do the backlinking.

20:54Now you can put this all into a system and it will do all of that. And you get 160 articles for the specific keywords of your competition with the right ones that you can actually compete for, et cetera, et cetera. And it's holy crap, like something that used to take a year can now take you five days. Now, yes, you need to have the knowledge on how to do that. But if you have the knowledge, you can get the outcome and not improve gradually different steps of the process. And there's multiple examples like that. Meaning, you have to think without any limitations, especially when it comes to knowledge and software.

21:41Meaning, yes, if you have supply chain management, there's no way I can land 10xX more resources on your factory tomorrow. Cannot do that. But when it comes to knowledge, marketing, decisions, optimizations, software, you can do things not 20 % better, but 10x. And so you have to stop thinking in the frameworks you had before. Think of the 15-star, let's go to fucking Mars solution. Because you probably can do it. And it's not easy. Exactly. We're trained to do something else. Yes. And that's the kind of part of the challenge is, like you said, it's so mind-blowing. We don't think it's possible.

22:32and that's part of what we need to do is break those barriers and using we use design thinking methods just to force people to think differently we force them out of the box yeah because like you said a lot of things these days we're just science fiction can be reality so let's go back to our process. Now you're trying to figure out what's the shooting for the stars solution. What's next? Yeah. Just to conclude that point, in this example case, again, the local bank here, what we figured out after going into the root cause was we have fairly strict regulation on know your customer process, especially for wholesale banking, has to do with money laundering and this.

23:28Banks want to be safe. So to do this, they need to go through a lot of non-standard documentation related to the company setup, who are the owners. They need to figure out that they have a list of 50 to 60 things that they need to find. For example, who are the owners who own more than 7 % equity of this company? Who has the power of attorney? all these kind of things. And because these documents are, they're not standard. It's just been a problem that they have not been able to solve. And because it's because of these constraints, they've stopped thinking about it. It is reality. It is what it is.

24:04In their kind of case, what they have done is basic OCR to read documents, but still there's someone either actually doing this, going through page by page and finding the items, writing it down, or then scrolling through PDF trying to find stuff. Way to make that process faster, hire more people to do this. But that's been it. So that was the root cause. Usually what happens after this, we start two tracks. One, we do a discovery on data and technology. So basically we need to understand what's the tech architecture, what data sources we have, how structured is the data. That's its own track with its own people.

24:48Usually IT, any data analysts they have, group up with a few of our engineers to just go through the details and big picture view of what are we dealing with. And then simultaneously, what happens is with the core sprinting is ideation. Now that we have this problem, how could we solve this? Again, here, we use design thinking tools to force people to think creatively. We usually start with kind of what's possible. What are the capabilities of AI these days on a high level? Just we understand what's possible. And we aim in a process where we start with kind of like a white hat where there's no restrictions.

25:34And when we discuss, we have a rule of you can't say yes, but. but you can say yes and, so you can build on top of things because it's really, like you said, we don't believe that these things are possible. So that's why we want them to really dream and think big. And then we put on a black hat. We look at, okay, can we actually do these things or not? And this process, in my experience, the most important thing is to have a diverse group. So you want techies. So you want to have a machine engineer there. You want to have data scientists there. You want to have someone from their IT. You also want to have someone from customer support who has no technical background.

26:16And you want someone from marketing. You want a very diverse group of people to come up with potential ideas. How could we solve this? In this case, the solution was actually fairly straightforward. What we need to do is read through these documents and to identify the right parts of information from there. What we ended up doing is fine-tuning a large language model. We used an open source model. So back then, OpenAI didn't allow fine-tuning yet. They didn't have their enterprise plans. So we ended up using a LAMA tool, fine-tuned with training material, of course, what the bank has a lot, to be able to find these 50, 60 piece of information of each company.

26:58In most cases, with the documents we ask, we're actually able to perform the KYC instantly. Of course, then we have flags, we have risk scores when we need more information. As we did the first implementation, we left that process manual for now. But now we're looking into using LLMs to communicate with customers and automate that part as well. So that's in this case. So I want to touch on that on the last two things you said, because I think they're brilliant and I think they're critical. One is what's the practical aspect of shooting for the stars? And the practical aspect, you said have two steps.

27:46Steps number one is really defining what shooting for the stars is from a outcome perspective. Okay, we want to do the process faster. That's fine. What does that mean? Okay, it means asking for specific documents. And again, I'm not talking about this case, but this could be generalized for anything else. But in this case, it's okay. We need to ask for specific documents. We need to put them in some kind of a database. We need somebody to read all these documents, fishing for specific pieces of information, following a checklist to verify that we have all of those, and if not, raise specific flags.

28:26Okay, that's what it means from a practical perspective. Can we do this with AI? And in this case, the answer is absolutely yes. And I think you touched on another point, and especially in a bank. there. And I think even today when OpenAI, even now with their enterprise level that they're saying they're not going to train on your data, I don't know. But for a bank or anybody that has more strict guidelines on data management, whether it's healthcare, banking, financial institutions, legal, etc. Defense, you probably want to run your own sandbox on-premise server that you know where all the data is and you don't have to put it on third-party servers.

29:14And so that's another practical aspect of the solution. But there is a practical implementation that allows you to build that shooting for the star solution. So these are the steps. And then you're saying, okay, now you can do the follow-up. Okay, so you can go beyond you were at the 13 stars kind of thing. Okay, can we go to 14? And the answer is, yeah, we do a step two of the project. We'll figure it out and we keep on going. Amazing. We're about to continue from there in another example, I think. Yeah, I was about to say that usually this idea of a session depends on how wide we're looking at this case.