Marco Casalena walks through three steps to start with enterprise AI: pick a concrete yes/no prediction like churn, confirm you have labeled historical data to train on, then operationalize by showing predicted probabilities so users can act. He works through a churn example with definition choices and reporting.
Getting your business started with AI can seem complicated, but really, it's as simple as asking the right questions. I'm Marco Casalena, and this is Ask Salesforce. Hey everyone. Today we're answering the question, how can I get started with enterprise AI? I get this question from customers a lot. And really, there are three easy steps. Step one, decide what you'd like to predict. Step two, ensure that you have enough data data to make that prediction. And step three, turn that prediction into action. There are some predictions we see businesses making all the time, like which customers are most likely to churn or who's going to pay their bill late. Let's try applying these steps to predict churn. Step one is to formulate the question. Now, certainly the question is yes or no. Is this customer going to churn or not? But how exactly do you define churn? If you're a subscription business, that might be easy. The customer just unsubscribes and that constitutes churn. But if you're not, it might mean they haven't bought a product from you in the last six months. Remember that this question needs to be a concrete question. Yes or no? There can be no wiggle room. Step two, ensure that you have enough historical data for your AI to learn from. So you should be able to build a report that looks at your customers historically. And next to each customer, you should see a yes or no. If you can't report on it, you can't predict it. And step three is to use your prediction. Now, this is a yes and no prediction, and so the results will come back in the form of a probability. For example, this customer may be predicted to have a 15% chance of churning, whereas this other one has a 30% chance. How are you going to use that prediction? Do you want to put that prediction number right on the contact record? You want to give your users a list of customers prioritized by likelihood to churn any way you cut it. You need to turn your predictions into action. So there you have it. Three simple steps to getting started with enterprise AI. What can you do next? You can head over to Trailhead and start learning more about AI, or you can browse our YouTube channel for more specific demos. Also, make sure you subscribe to our channel and check out the other videos in this weekly series. Thanks for watching and see you next time.
