To my session. Thanks for being here. My name is Andrea Dobrand. Currently serve as the head of AI and the product owner at G and Co in silico.G and Co in Silico is a tech bio company, based out of California and focusing on the precision medicine, and we are targeted at going public this fall. Before that, I was the chief data scientist at IBM Consulting. I build the AI engineering automation competency for the public market. I've been there for about eight years. And before that, I was working in the foreign industry for about a decade. I worked at Pfizer, Novartis, Boringo Income, and the ESI. Over the years, I've seen so many successful and failed AI project. And, I realized it's rarely the technology, if it fails. It takes the right strategy, people, and the governance as well. So you need all of them. So that's why I'm talking here about this topic.

We're gonna talk about the lesson lesson learned over the last decade that I learned from all this AI project. Alright. So let's start with a paradox. 95% of the journey I project have failed to reach production. And at the same time, 74 of the CEOs are now personally leading the AI positions in their organizations. So this number doubled from a year ago. Think about that. There are more and more senior people in the building making the call. However, the failure rate hasn't moved. That means the AI initiative are getting more and more attention and that we are getting more and more investment and more and more executive ownership. However, you are reaching the same results. So that is a crisis of confidence, and it's not going to be solved by using a better, larger language model or have a a bigger budget.

So what's the reason? Let's think about that. So very often, it's not a resource quest. It's not a talent issue. It's also not technology stack issue. And what is it? It's we don't know how to absorb AI. Organizations don't know how to redesign the workflows around it, how to govern it, and how to make decisions with it. So in one line, the failure isn't about the model quality. It's a learning gap in the leadership. So how do you close that learning gap? It starts with two things. First, what's the business problem? Who are your users? What problem we try to solve? Is it about inefficient business process or is it about inaccurate output that AI generated?

So we always need to think through what are the business problem we try to solve. I've seen this over and over again, and sometimes we got excited about playing around with the latest technology, the video generation, some cool stuff. Or we got excited about only an asset or only a product. We have a proud index. But we always, always have to remind ourselves. We have to understand the business user. Who are the user? What are their needs and goals and the pain points? And how can we take them from current state to the future states with AI capabilities to help them? So that's the first thing we need to think about. The second is we need to have a clear picture of the business impact in your organization.

What kind of a business value it can bring in sixty days or ninety days? You need to have a very clear picture early on. You need to evaluate business impact, efforts that needs across different teams, and what are the execution feasibility and the financial feasibility. You need to think through all of that. So you don't need to get started just having a rough idea and saying that, okay, can AI do something interesting here? That's not enough. You really have to have the business mindset before you get started. For any AI initiative, if you cannot answer that question concretely, maybe it's better to take a pause to think through. Right? It's better than spending six months and cost a million dollars on the wrong problem.

It's a better pause than spending the six months and the $1,000,000 on the wrong problem. So look at the right side, the code here, organizations that win are not the ones that come up with the best model, and they are the ones that got the right problems first and then moved before the window closed. So that's not a technology advantage. That's a decision making advantage, and it starts with asking ourselves these two questions. So I see some questions coming up here. Let me go over the all the content, and we will have a q and a session in the end. Okay. Now now you have the right problem identified. So the next question is how do you deliver? And this is where the most organizations struggle. And it is not because they don't know how to build AI, it's because they never clearly defined whether they should build it or buy it or partner with someone.

They drifted into, okay, we need to build it because all we do is confidential, we have this proprietary information, we don't wanna work with others. Right? Let's build it ourselves. And then 12 later, they invented something new, they think, but they could have aborted it in six weeks. So we have the data from the Deloitte here. Deloitte this year found out that organizations with a fully integrated AI are four times more likely to report revenue growth. Four times. So the word integrated matters a lot here. Everyone has AI projects, but not everyone has AI that actually invalidate me how the organization runs. So the gap between those two things is where the four times lives. So now how do we solve it? We need to make the delivery call early and then make it very honestly. Right? We need to be very honest to with ourselves. What are we capable of?

What is our differentiation? So you should build it if it's your differentiation, if it's your moat, the things that made you different in the market. You keep it to yourself, and then you buy it when it's resolved. For example, the document processing, chatbot, the standard analytics. There are a lot of loose tools out there in the market. You don't have to build it yourself. Right? You just need to buy it, incorporate it, and then move on. So every month you spend rebuilding a commodity is the month you are not spending on actual on the things that actually differentiates you. Now when do we partner? So lots of people struggle with that. Shall we partner?

Lots of organizations resist most on this because it feels like at the meeting, you cannot do it by yourself, but it's not. It's really to about leveraging all the resources that you can to achieve your goal with the fastest paths. It's a faster path to the integration, and the organizations that scale fastest are the ones honest about themselves and honest about their gaps. So to build or to buy or to partner. The wrong call is not making the call. You really have to think through early on. Okay. Now, the next steps. So once after you've done all of that right, what can still kill the projects? It's the governance. We don't know how to manage it. And the it's not about the model. It's not about the the GVD or Copilot. It's not about the data. It's not about the talent.

Even the leaders say it is themselves. They don't know how to manage that. They don't know how to govern that. Right? So sometimes the process they build to manage AI is the thing that's killing it. We have a lots of organizations that they focus on training their employees on how to use a different kind of tool or learn all those new tools that come up in the market that can help them. But we also need to have enough training on leadership, on the judgment, on AI judgment. So sometimes the bottleneck isn't really the people using AI, it's the people who make the decision on AI. So what does a minimal viable governance look like? We need to do three things here. The first is we need to figure out who's responsible for what, who are the owners, and how do we evaluate?

We need the metrics. We need the responsibilities. We need the accountability. We need to have a very clear understanding of that. Then after that, we need to think through what AI project to do and what's what are the measurable value. Are you improved efficiency? Are you improved accuracy? If you cannot see the impact in sixty, ninety days, Maybe you are not building the right thing. Lastly, exit condition. After knowing that, okay, maybe I'm not doing the right thing, What should I do? Should we kill the project, or should I continue? If you can answer all these three questions before you start, you're already ahead of the most. Alright. So, lastly, let's just summarize. So what are the things that actually separated smart bets from costly missteps? The first, we need to start with the pain points. Right?

Identify the user, understand their their, business problem, needs, goals, and the pain points, and then focus on the pain point to think through what's what's possible with the latest technology, and how can we help them take from current current stage to the future stage, and what exactly how we can help them.

And the second, we need to all be honest with ourselves. What capabilities we have? What are the capability out there for us to choose from? Can be existing tool? Can be consulting firms that we can partner with. And the third, we need to clear ownership and the responsibilities, and then we can establish risk based metrics, right, and have a clear understanding of the business impact, six days, ninety days, where we are gonna where we are gonna be and how what we are gonna achieve and when to exit.

And then lastly is we need to keep developing our leaders, how to evaluate the AI, how to make a decision around it. So now I have a question, one thing to take back. Right? Find find the person who owns your most important AI initiative and ask them this question. If this model does not give you the recommendation that you agree with, what would you do? If they can answer that clearly, that's good sign. And if they cannot, that might be your bottleneck. Right? And it has nothing to do with model. K. That's all I wanted to share here. Now we are open for questions. Okay. I see a question about, the build versus buy versus a partner decision. Sounds simple, but who actually should make that call? So I will say it's a it's a joint call between the technical leader and the business owner.

So the problem in most of organizations is to they have a separate teams. These two people are separate. And even from the early on, after they've identified the AI initiative they want to prioritize, these two people these two group of people still working in silos. They are not working together to make a decision. So the technical leader knows what's feasible, and the business leader knows what's urgent, right, knows their pinpoint. They are the the other end users. So the technical team is to serve the business owner. So they should all be in the room together to make the call. So this is related to the governance question. Right? How to name the owner? Who should own it? And then who should be responsible? Who's accountable? Who should be informed? Hope that answer your question. Okay. I see there's another question come to me.

How do you convince leadership to pause an AI initiative where there's already pressure to show the results? I think this is a very difficult, decision, and I've seen it over and over again with the some of the startups that you the leaders, the CEOs already approved $1,500,000 and, work on AI initiative for more than a year, but then they don't see results. What do you do? I would say, you know, you you don't frame it as a pause. You're framing it as a protecting the investment. So the question the owner and the even more senior leader need to ask themselves or, still, what does the success look like in ninety days, right, if we keep going? If nobody can answer that, then continuing the project might not be the wiser decision. It's it's very expensive. In the six months from now, you might be in a even much harder conversation about why nothing happened. Right? You don't see a business value. Yeah.

Hope that answer your question. Alright. Okay. If there's no more question, then we can conclude the session. And thank you all for joining this session. If this topic resonates with you, I'd love to continue the conversation. And thank you for joining. Enjoy the rest of the conference. Thank you.