Transcription
Today we're talking about how to find high lever AI use cases for your business. Everyone wants to use AI to grow revenue or decrease costs. But many business leaders have no idea where to start. A lot of companies are focusing almost strictly on particular tools. They either want to use Chat GPT for everything or they're trying to find ways of using automation toolkits. But that's precisely the wrong way to go about it.
The companies that win with AI don't start with tools. They start by identifying problems in their business. Companies that do this right figure out exactly where AI can drive more revenue, lower cost, or better customer experience before they ever write a line of code. Hi, my name is Joshua Edner. I'm a 20-year expert in AI, machine learning, and data science, including at companies like Apple and Bank of America. In this video, I'll give you a clear step-by-step framework to find the best AI use cases in your business, the ones that will actually drive measurable results.
Okay, first number one, you want to start with your biggest levers. Here's what I mean. You're not going to start with tools. You're going to start with your highest leverage problems. So look at your business and try and figure out where your biggest problems are. The things that if you fixed them would drive the biggest increases in revenue, the biggest decreases in costs and so on. So for example, let's say you have a churn problem. This is very common in a lot of businesses. High churn is often one of the biggest problems with growing revenue in a business. If that's the case for your business, you'd want to flag that. Alternatively, the problem might be customer acquisition or conversion. You might have a customer experience problem. By the way, customer experience and customer satisfaction are often upstream of a lot of those other problems. So, if you have really bad customer satisfaction, you're probably going to end up seeing it in your churn, in your sales conversions, and so on. So you want to go through your business and figure out what the biggest problems are.
But identifying problems in your business is really only the first step. And that brings us to the next step. Figuring out which AI techniques can potentially solve those problems. So you want to go through the high lever problems that you have and then try and figure out what AI technique might be able to solve that. Now I will say this does take some expertise in AI. I've talked about this elsewhere, but if you really want to build a great AI strategy, you need to have people on your team that understand AI, machine learning, data. You're going to need some executives with a little bit of expertise in those areas that will help you understand the types of AI tools and techniques that are going to be available. But a couple of examples, if you're trying to fix churn, in many cases, you're going to potentially build a churn model that can predict churn. You can also use those AI and machine learning and analytical techniques to analyze your churn data and figure out what the drivers are. Something similar can be said for conversions. If you have a customer satisfaction problem, there's going to be a different set of tools and techniques that you might use. So, you might use AI for text analytics and sentiment analysis to try and figure out what the drivers are of your low customer satisfaction. You can also use a variety of let's say decision trees and other things. If you build those systems properly, you can figure out what the problems are that are driving low customer satisfaction and in turn you can go fix those problems. So what you're trying to do here is you're trying to take the problems that you identified and you're trying to figure out which specific AI tools and techniques can help you start to understand those problems and potentially fix them.
By the way, if you really want to use AI in your business, you're going to need to figure out if you're ready for AI. That being said, go take my AI readiness assessment. You can find the link in the description. Click on the link. It's only a few questions. It'll take you less than 5 minutes and it'll tell you if your business is ready for AI, and if not, where your gaps are. Click on the link and go take the AI readiness assessment right now.
After brainstorming possible ways that you can use AI to fix your business problems, you're going to need to do the next step, and that's analyzing the difficulty and probability of success of those different AI solutions. So, here you need to start thinking about the resources that you have, the data, the people, the talent, and the specific techniques and tools that you're going to need to solve your problems, and try and figure out, okay, can we actually solve this? How difficult will it be? Do we have the resources? Certain types of AI solutions need certain types of data. So, you're building a predictive model, you're going to need quite a bit of customer data in many cases. If you're doing things like text analytics and sentiment analysis, you're going to need things maybe like survey data or sales transcripts, customer service transcripts, depending on the exact use case. And if you don't have that data in hand, you're going to need to get that first. it'll make it more difficult to actually use that particular AI solution. Alternatively, different types of AI solutions might need some specialized expertise. So, for example, if you're trying to increase your customer lifetime value, you can potentially build a recommendation system. But recommendation systems are very difficult to build and you're going to need some talent either internally or you'll need to be able to hire somebody or hire a consultant that can help you with that. So, remember what we've done so far. We've first identified the major problems in the business. Then we've brainstormed potential solutions to those problems. But here we're trying to estimate the difficulty solving those problems with those particular solutions with the resources that you currently have. Your estimates, by the way, don't need to be perfect, but your estimates will get better if you have better machine learning and AI talent inside of your business. and if your executives are better educated on different AI tools, the strengths and weaknesses, the difficulties with building things and so on. So, I do recommend one of the ways of improving your estimates is simply getting better education for your employees and your leaders.
After you've estimated the difficulty, you'll be ready for the next step. Here, you want to plot your potential use cases in a two-dimensional matrix. One axis is going to be the impact. So how much relative impact let's say on a scale of 1 to 10 that solving that particular problem would bring for the business and on the other axis you're going to plot the difficulty with solving that problem with the particular AI techniques that you've brainstormed. Then you can divide that up into a 2x two matrix. So for example things that are high impact but low difficulty those are going to be things that you'd probably want to attack first. things that are low impact and high difficulty you'd probably want to avoid altogether. Doing it this way will give you a lot more clarity about high impact, high certainty use cases that you can solve right now versus things that you just kind of want to avoid. And in turn, this will make your AI investments much more disciplined and will help guarantee an ROI.
Finally, after you've identified potential projects that are going to be high impact and relatively low difficulty, you can run a pilot. Essentially, you want to build a prototype as a sort of experiment. Here, what you're trying to do is you're trying to validate the difficulty level that you estimated earlier in the process. Sometimes those difficulty estimations are a little bit off. You might think that a problem is relatively easy to solve and you get into it and you find out that it's much harder than you originally estimated. That's a big problem if you overinvest. So what you're trying to do is you're just trying to build a small prototype as an experiment that helps you gather more information. If you get into it and you find out that the project seems to be much harder than you initially anticipated, you can kill the prototype without too much loss. Alternatively, if you start building the prototype and it proves that the system works, then you can start building it up and scaling it out.
And once you've found the right use cases and you've proven them out a little bit, the next question is how to turn those into a lasting advantage. Because finding good use cases is just the start. The best companies are the ones that build AI systems in a sort of loop in a way such that they continue to learn, improve, and compound over time. This is something that I call the AI data flywheel. In the next video, I'll show you exactly how to build one so that every new customer, every new data set, and every new AI use case makes your business more and more successful. You can check it out here. The AI data flywheel.