Don Jackson
Agentic AI is artificial intelligence that has agency. What do I mean by agency is it has the ability, so whether you're talking about testing or not testing. It has ability to make decisions and take actions with minimal human intervention. That's what agentic AI is. Well, apply that to testing, what does that mean? Think about this. If I were to script something, if I was to automate, let's say I'm working to test airline mobile app. What do I have to do? Well, the first test case is booking a flight. Let's be honest, right, we all understand the concept of book a flight. How would I do that? Well, I would first log in and I would tell it how to navigate to that page. I'd say, click the login button, type, click, the login ID, type in the value, type in the value of the password, click the log in button. And I tell it all those steps to do. And then I'd start booking the flight. Well, how do I book a flight? Well, I gotta find the from and the to location and I need to type in the airport name and the two airport name and then I click find flights and I tell it all those steps. In an agentic AI, what you do is you give it a step that says book a flight from San Francisco to New York in business class, prefer an aisle seat, second preference is a window seat. If there are no flights available, that have one of those seats, I don't wanna sit in the middle. I don't wanna sit the middle. Come back with an error message. And all of the steps that it takes to do that of finding the objects on the screen, typing the values, think about clicking the seat map to find is it close to the window or is it closer to the aisle? How difficult that would be to script. It handles all that for you. And so by going agentic, you don't actually have a script. During at runtime, it takes that natural language prompt, looks at what's going on in your application and then makes a decision about what it needs to do to achieve that objective of booking a flight. All of those steps that you would have had to script out, that's probably like at least 50 steps. That you would to script out and deal with all those objects and all those locators, all that goes away. Your actual artifact that you track to and trace to is your test stuff. There's no script. There's no framework needed. I don't have to know Selenium or Playwright or Cypress or Espresso or Appium or whatever. I don't have to all that. What I know is the business process. And because it's making decisions at runtime based on what the application looks like and how to get through it and understand semantics and context, when cause we all know it's a when not if, the application changes, it just, it doesn't break. There's no self-healing involved. It just doesn't break because it's making that decision based on how the application's behaving, just like as a manual tester. If we change our application, my manual tester doesn't brake, he just or she makes, just goes the different path. And you could start doing things like find, you can start doing exploratory testing with the Agentic AI. Find different ways to get to the shopping cart. There's like 10 different paths. We actually talked with one customer that was a beta customer and they're like, we did some of this. And we said, Hey, here's the 12 paths that you can get to this part of the application. And they said, we only knew about 9. The other thing is, because it's working at the user experience level, is things that were just untestable before, that you just couldn't automate before, are now automatable. We have a financial services company with another one of our beta customers. I don't know if we have permission to use their name, so I'm not gonna use their names, but think about like a stock trading app. And you get the graph, and then you can put overlays on that graph of, okay, so if it's red going down, green going up. And now you can actually write about assertion that says, if the point in the graph before was higher than the current point, you should see red. You should see green if it's not. And oh, by the way, down in the chart where you've got the numbers, make sure the chart matches the numbers that are down in table. Things that the way that you automated it before, let's be honest. You did static data. In order to test that kind of stuff, you do static data. Same thing like think about like a big e-commerce company that we are showing this to, they have a picture of their products and then a description, but they do marketing campaigns where the description changes. Sale on laptops, so every single description says now sale on laptops. Well, they couldn't test those marketing campaigns because it changed the static data that was in there. They'd have to dynamically do that. And oh, by the way, I can write a validation and assertion that says, Does the text match the picture? If the text said HP laptop with a 17 inch screen and a 10 key, it could look at the picture and see, does it have the HP logo on it? Does it have a 10-key and the keyboard? That kind of stuff that you just, you couldn't automate that before. You just, I know what I used to do. I used do the cheap way out because could I automate it? Maybe really hard, might take me a week to figure out how to do all this. I just grab a snapshot and then put a warning message in my log that says, hey, whoever's reviewing this, just look at the picture and make sure it's right. When all that can be automated. And super excited if you can't tell.