Benefits of AI in Test Automation

Photo by Kindel Media on

There are several interesting web app automation scenarios that we can improve using AI:

  • Reduce the execution time: Nowadays you have the feature target function already even without an AI test automation project, but with AI you can add this feature without having cucumber in place or even the need to tag the scenarios or features. The AI should be able to identify the features related to the change automatically.
  • Converted manual test cases to automation: you can use Natural Language Processing (NLP) to automatically translate manual test cases into automated test cases. I have seen this done with cucumber not AI yet, but totally possible as AI models work on datasets.
  • Creating different data combinations by training the AI to identify the possible combinations based on a dataset is possible. This would increase the data coverage and bring more confidence to the automation project.
  • Visual validations: Many tools perform this functionality already. I personally tried one tool ages ago called Percy, but you can also try some other popular tools like Applitools and Telerik
  • Test execution stability or self-healing automation: AI can automatically locate web elements when the primary locators fail. You can see this feature in some cutting-edge automation tools like Mabl and Xray and Functionize. Self-healing employs data analytics to identify objects in a script even after they have changed. When your script fails due to being unable to find the object it expected, the self-healing mechanism provides a fuller understanding and analysis of options. Rather than shutting down the process, it examines objects holistically, evaluates the attributes and properties of all available objects, and uses a weighted scoring system to select the one most similar to the one previously used.

Becoming a Domain Model Expert

Creating a model for your test automation requires a domain expert, therefore is critical to have a test automation specialist that also knows the business so the AI can bring the desired innovation. With such extensive use cases, AI systems will need different parameters from domain experts.

Machine Learning Algorithms In Layman's Terms, Part 1 | by Audrey  Lorberfeld | Towards Data Science

Be careful to not run more automated tests than you actually need it. A stage of supervision when the AI is learning the patterns is definitely needed it.


AI for Testing: Beyond Functional Automation webinar

Hello guys, I joined a webinar some months ago (15/07/2020) about AI for Testing: Beyond Functional Automation by Tariq King which was really interesting ! I know how it’s hard to keep up with all the online events now, so I always try to keep the recording of the ones that I couldn’t join and are interesting to listen to when I have time.

So thought about sharing with you as well in case you missed. You will learn about reinforcing learning by giving scores to the right actions and about training bots to recognize good and bad designs with examples. This allows the framework to be more robust when searching for a particular query or asserting the scenarios:


Here it is the link to the recording:

Thanks Tariq King !