AI in Test Automation: Is this what we’ve been waiting for?

It feels like we’ve been here before. Many times. The current hype around AI is causing waves across the globe, with almost every business and technology under the spotlight to understand its impact, with many arguing that we are early in the hype cycle and that expectations are inflated.

In the world of test automation, it all feels very familiar. Since the early days of test automation, predictions around the elimination of human testers have been made, with little progress. Yes, features in tools and technologies have allowed test engineers to become more efficient, providing assistance to accelerate common tasks. Some tools have arguably taken it too far, with the ever-increasing number of “no code automation” products masking the fact that you will always need to write lower-level code in some cases and that the “drag and drop” GUI approach to automaton coding is often “clunky”, inefficient and inelegant, aside from often driving a bigger wedge between dev and test.

In the world of AI, visual automation tools have been around for many years utilising image recognition and OCR technologies. “AI” in test automation is therefore nothing new. What is new is that the majority of test automation tools now promote they have AI elements to them.

But what does this actually mean? So far, I see another set of features that in some cases make day-to-day test automation even more efficient. I say in some cases, as the huge amount of hype around AI is leading to an explosion in features, many of which in time will be seen to be counter-productive.

Those currently leading the pack include automation code assistants/accelerators (in common with wider development), LLM-based test data generation (although I would argue that this will quickly lose benefits versus more traditional methods for anything beyond the simplest of data models), and “self-healing” automation based around ML algorithms to re-identify objects at run time as UI changes are introduced. In addition, visual automation tools continue to develop which, when combined with other automation tools to drive user interfaces, can be very effective in enabling combined functional and visual testing.

In other words, more features that accelerate activities and make human test engineers more efficient rather than replace them. The age-old argument that nothing replaces the value and importance of proper test analysis remains, with the role of the tester to ensure business requirements are being sufficiently met and to understand and define scenario-based tests being as important as ever. Can AI bring together the key stakeholders of an organisation to collectively understand its risk appetite for new technology products, changes and releases for example? If and when it can, I’m heading off to a remote desert island with my best mate Wilson.