3.2 How can I introduce AI into my organization? - Video Tutorials & Practice Problems
Video duration:
8m
Play a video:
<v ->All right.</v> So how do you introduce AI into your current process? So let's start by looking at maybe an example of a current process that you might have where human judgment is really what's making the process run. And that human analysis, human decisions, human actions, this is the problem that you're trying to solve and you'd like to automate it, you'd like to use AI, instead of completely relying on human analysis and human actions. And so, what would you do here? So your current process starts with data, some type of data. And so that data is then analyzed by humans. And it could be that the way that they analyze it, what they're basically doing is they're gonna create some kind of PowerPoint deck or some other type of slide presentation. And that's where the analysis really leads where they're looking at a situation and they're saying, "We've analyzed this situation. "Here are the insights that we've come up with." And that deck, that slide deck then turns into a set of suggestions for improvements. And so it depends on what the problem is, what those suggestions might be. But once you've got those suggestions, what do you do with them? Well, now what you do is you have human beings, maybe other human beings than the ones who did the analysis, and you have those human beings take action. Now, when they take action, what happens? Well, what happens is it likely changes the data. So now, whatever the thing is that they worked on, now, you're going to use human beings to analyze the new data. Now that they've followed some of the improvement suggestions, and they've taken those actions, now you're going to analyze it again. Now you're gonna come up with more improvements. Now you're going to have humans take more actions, and you go round and round and round. And so this first step that uses humans to do the analysis, this is actually how your processes probably work today. And so this is your starting point. So what's the second step in the process? Well, the next step would be to start to introduce some AI into your process. And maybe we'll take machine learning as an example and do that as an experiment. So start to experiment with putting AI into the process. So, now, instead of the data just being available for human beings to analyze, now, you're gonna need some type of process to allow the computer to see the data. How would your machine learning system actually get access to the data? Now this is something a data scientist can help you with, but you don't always need a data scientist. Sometimes you can have an IT person who can actually bring the data and put it into a form that your data scientist is able to then use as part of this machine learning analysis. And so the very first step is something that they call data ingestion. So ingestion basically means that it takes data in whatever the form is that you have now. So it could be that it's already computerized data, so it might have to move it into a new form, or it could be that it's actually data that's on paper, and you might have to have someone key whack it, or maybe you're going to have some way of scanning it, or some way of digitizing it. And so whatever way that that data exists now, you have to use some kind of ingestion process that basically prepares it for that machine learning analysis. And depending on what that data scientist thinks that they ought to do, you might have different types of models that you might create. But the idea behind all of them is that they're gonna create some type of dashboard. They're gonna create some way in which they're representing that data and showing some insights. And so, in addition to that slide deck that the human analyst used to create that showed the insights, now you might have some insights that are coming out of your machine learning process into some kind of data dashboard. Well, the other thing that's gonna happen is instead of your human analysts only looking at the data directly, now, they're gonna look at this dashboard, and this dashboard is actually going to give them some insights that they might not have picked up on their own. That's what the benefit is of using machine learning here. And so the other thing that'll happen is the human analysts will actually look at it and say, "You know what, "I know that your clever AI model thinks "that that's a brilliant insight, "but it's actually wrong or it's dumb, "or it's very trivial, "and that's not actually helpful at all." And so there's other things that we really want you to look at. So they're gonna provide feedback for you and your data scientists to then tweak the machine learning analysis so that your data dashboard gets smarter and smarter and smarter. Now, the human analyst is still going to take the insights that they got directly from the data and the insights that they got from the machine learning analysis, and they're gonna make the improvement suggestions. And you're still going to have human action in order to put those improvements into place. And then, again, it will go back change the data again and go through the process all over again. So this might be your second step to your process where you're now started to experiment by introducing some AI into your analysis process. The last step, the one that really can provide even more value for AI in solving the problem is to automate the decisions, to actually have machines take some of the actions. So it's not that necessarily you'll eliminate human actions, but this third step of actually implementing automated actions based on the AI, this is where a lot of the power can kind of come in. So let's give an example. So you remember the example we used from Amazon, where they were doing product recommendations. And so they showed you, if you came to a product page, it said people who looked at this product page went on to buy these other product pages. Well, you can imagine in the beginning, those recommendations were actually put there by human beings, and you can imagine the next step they had is they tried to use some data to analyze, well, what would be better products to put on some of these pages. And human beings looked at that machine learning analysis and they updated the pages. Well, what do you think the final step was? Well, the final step was for human beings to maybe oversee some of that, but for the machine itself to actually put those products on the pages. So it could be that human beings still go back, spot check things, see if maybe there's something that's not working quite right, or see if something seems somehow inappropriate. Even if the data seems like a good thing, maybe it's just not a good idea because it's gonna cause people to kinda not like what you're suggesting and have other things in there that're not just purely data-driven. So there might still be people who are overseeing those product recommendations, but the final place is to implement the machine learning so that the machine is directly making those recommendations, and you don't need to have a human being involved all the time. And this is where a lot of times you'll get the speed you're looking for out of the process and you'll get better effectiveness. And so these three steps of starting with your current process, then starting to experiment by looking at what the AI can do to add to the analysis, even while human beings are still making all the improvement decisions and then taking the actions, and then the final step where you start to trust the AI enough that you're gonna allow it to automate some of these things to make some of the decisions on its own. And so these are really the three steps that you can use to introduce AI into your organization to solve any particular problem that you may be identifying it for.