Your Kind Of Automation – Using AI

We all know we can’t automate something that we haven’t run manually.

It could be a set of APIs that we need to call in sequence, copying IDs from response to request bodies. It could be filling forms and seeing if we got the results we expected. Only after we’ve seen that what we expect works, can we talk about automation.

We can automate in different ways, and when it comes to AI in software testing, our AI helpers can really give us a boost.

This is especially true if models are familiar with apps like ours. Our Bigger Better Bookstore^TM is really unique, but there’s a rumor about other bookstores in the world (I fail to recall their names, though). But it still sells books. Users can look at book details, sort them by genre, and search for them, like any other (rumored) online store.

And if our AI assistants have been trained on these apps, they “understand” online shopping. This means they know about common user behaviors and, crucially, they also know about testing them. The real power of using AI in software testing, is using its ability to apply this context.

So, if we like what we see, we can ask our favorite LLM to generate Playwright tests:

LLMs never say no, so you can expect to get a script you can run. Does it really work? It depends, and maybe it will not, on the first try. Perhaps it’s not using the right locators, or the system reacts differently than the AI expected. But maybe you’re lucky and it runs out of the box.

Even if you’re not lucky, you still have a good-enough script to work from. It can still be a major time-saver.

AI in software testing – Your way!

Let’s say the test works, or you made it work. That’s great, but it doesn’t look like your other tests. Why? Because your existing tests weren’t written from scratch. You’ve built a magnificent infrastructure, and your newly generated tests don’t use any of it—not the right libraries, base classes, or reusable modules.

You know what the T in GPT stands for? Transformation. Just feed it the right information and it will perform the transformation easily.

For example, if you’re using the Page Object Model (and you probably should), why not ask:

Well-known patterns have an excellent chance of successful translation. But this is just one kind of automation where AI in software testing shines. Of course, you can point the model of other patterns in your code, for your flavor.

Get to the point!

What if we’re not interested in a complete test? Automation isn’t just about automated tests, right? We use it to make our testing more efficient. Sometimes, we just need to avoid performing tedious manual tasks.

Imagine that every time you need to test, you have to go through the “create new user and log in” process. Sure, it can be done manually, but why waste time? Let’s ask for help.

Now I’ve got a script, not a test, to help me get to where I need to be for the real testing. You can use AI to generate differnet scripts for any small, repetitive setup tasks.

By the way, both types of generation—tests and scripts—are marvelous for handing over to a developer and asking them to fix that bug you just found.

So, if you want to get to the point, you can really make use of our AI transformers. they can save you time on your testing.

Your turn! What’s the first manual process you’re going to automate with AI? Share your ideas in the comments below—let’s see what we can transform together.