Successful AI-driven testing in action at Moët Hennessy Louis Vuitton

Artificial intelligence and machine learning are undeniably key enablers in the tech industry. However, there aren’t many enterprises that have adopted them effectively. A recent Forrester report states that ”all executives need to make strategic decisions about how and where to leverage these technologies, but few leaders have experience with them, so misconceptions abound, causing poor outcomes, wasted resources, and resistance to future initiatives.”

Despite huge demand for AI technology, only a select set of innovators have implemented it. With the successful deployment of AI being scarce and its demand being high, it is critical to learn from organizations who are ahead of the game. Johann Gaggero, Head of Omnichannel QA for the perfume and cosmetics division at Moët Hennessy Louis Vuitton (LVMH), recently demonstrated their AI solutions that have come to fruition, and crucially, are delivering positive results.

Gerta Sheganaku, Product Engagement Manager for AI and machine learning solutions at Tricentis, discussed with Johann their respective approaches to AI-driven testing – LVMH injecting it for smarter and faster test results analysis, and Tricentis using it to drive test automation.

 

The tipping point for using AI in testing: demand and data

Before we delve into the strategies behind LVMH’s AI rollout, let’s consider when it’s the right time to make AI a strategic part of your testing practice. The tipping point for Johann was knowing that he had an abundance of impressive data in their warehouse, paired with a significant amount of interest from colleagues to leverage this data to bring various valuable use cases for AI to life. The data output of their 10K automated tests was sitting dormant, and the time came to recruit a new trainee to realize their AI aspirations.

Gerta observes that alongside increased data availability, the cost of storing data has significantly decreased over the past few years. It’s no surprise then that we have also recently incorporated AI across our Tricentis product line, but what’s interesting is that there are distinct parallels between the approaches at LVMH and Tricentis.

Start where it makes sense and where you can scale

Johann and his team at LVMH selected a very pragmatic use case in which they could apply AI to their test process. By focusing on the attainable area of developing MLP (Machine Learning Python) scripts based on screenshots, they have achieved an impressive 50% reduction in the costs of their test automation.

At Tricentis, Gerta stresses that it’s imperative to consider what really matters. AI uses probabilities, which is something both LVMH and Tricentis have leveraged to determine how to proceed, ensuring that AI is applied to ‘mission-critical and scalable’ testing.”

Free the tester to focus on what they love

Gerta and Johann agree that AI should function to enhance the tester’s experience, debunking the misconception that AI will replace us. LVMH’s implementation exhibits that having an AI element running does indeed free the tester:

”Before adding this machine learning model, it took my QA engineer half day to analyze the logs and distinguish false positives from real bugs, and now it takes her no more than two to three hours.” – Johann Gaggero, Head of Omnichannel QA at LVMH, PCIS

How does this machine learning at LVMH work exactly? At every single test, a screenshot is taken, and these are split based on whether they’re related to a passed or failed test. Based on this data set, every time automated tests run, Python machine learning will judge whether they have passed or failed and rate the accuracy of its recommendations. Smart, huh?

At Tricentis we have liberated the tester by adding deep learning to Tricentis Tosca, which can understand and drive visual interfaces the same way humans do. This Vision AI feature is trained on 9M controls, 12.5M examples, and is continuously learning. AI can also now be leveraged in Tricentis LiveCompare for SAP, predicting the impact of updates and highlighting which tests to run for each release.

If it’s broke, Self-Healing AI fixes it

When your application changes, Tricentis Self-Healing AI adapts to the change, so even if the technology layer alters underneath, your tests can still run. Vision AI in Tricentis Tosca has Self-Healing AI properties to adjust to any changes, meaning that testers can spend less time fixing breakages, and more time being innovative. This is one of the primary reasons Johann is exploring how to integrate Tricentis Vision AI into their process at LVMH:

”If indeed we want to go to DevOps, we need to trust our automated tests, right? We need them to be robust. Having auto-healing automated tests on platforms such as mine, ecommerce platforms that are changing every day, is just not a ‘nice to have’ anymore, it’s just a ‘have.’” – Johann Gaggero, Head of Omnichannel QA at LVMH, PCIS

Next-generation automation

Making the jump to AI to doesn’t have to be daunting – if you carefully implement it in an area that has the potential to make a huge impact, it could be extremely rewarding. Johann saw it has a fresh challenge, recognizing that AI already permeates our daily lives and presents an opportunity for groundbreaking change: “AI is already here, all around us. The question is do we want to catch the train or not?”. Watch the Virtual Summit interview with Johann.

Find out more about how Johann conducts his testing practices with flair, and his colorful career in luxury brands, in the new Tricentis podcast, Transformation in 10.