How AI-Assisted Test Automation Can Transform the UI: An Interview with Gil Sever


In this interview, Gil Sever, the cofounder and CEO of Applitools, explains the importance of automation in modern testing, why you need to be customer-obsessed, and how your UI can determine the success of your applications.

Josiah Renaudin: Welcome back to another TechWell interview. Today I’m joined by Gil Sever, the cofounder and CEO of Applitools. We’ll be discussing the increasing scope of test automation in modern software teams.

Gil, thanks so much for speaking with us. First, could you tell us a bit about your experience in the industry?

Gil Sever: Thanks, Josiah. Good to be with you and the TechWell community once again. As a serial entrepreneur, I have always been personally motivated to solve technical problem through technology and software. Having held leadership positions at other technology startups and NASDAQ companies, our adventure here at Applitools is one of the most exciting and interesting I have ever experienced.

Why? First, it’s the clarity of the problem—we have historically had no way to automate UI visual testing like we do unit and functional testing. UX is breaking everywhere, all the time, at huge global companies like Google, Microsoft, Amazon, UPS, and Twitter. No one is ignorant of the scope of the challenge, but few believe automated visual testing is even possible.

Second, it’s the simplicity of the solution—a single line command line that adheres to your existing functional testing code to automate UI visual testing. Having shared this “aha” moment with thousands of engineers is gratifying for sure. You feel good every day about solving a huge problem in such a simple and elegant way. Engineers truly love it!

My experience in the industry has rarely been so uplifting and positive. We are really enjoying and appreciating the passionate developer community building up around the visual management aspects of applications.

Josiah Renaudin: Could you explain how the scope of automated testing is moving all the way to the “front door” of digital transformation initiatives?

Gil Sever: Actually, I think the way you pose the question suggests a lot. How can we avoid automating testing at the front door as quickly as possible? The UI is more important to your brand in today’s world than ever before. eCommerce will double again in the next three to four years to over $350 billion annually. Customers are spending an average of ten hours a day on screens of all shapes, sizes, operating systems, and responsive designs. Pick your favorite stat here, but the reality is we have a potentially debilitating combination of trends hitting test automation engineers as a result. On one hand, digital transformation is here and creating hundreds of application UX variations daily. On the other hand, CI/CD, including Microservices, are extending agile to a whole new level of release pace. This is putting enormous pressure on testing to manage the UX aspect without sacrificing quality.

Josiah Renaudin: Thinking of customer needs has always been critical, but why is being customer-obsessed more important than ever? And are there any possible drawbacks by consuming yourself with what you think the customer needs?

Gil Sever: It’s a great question, but my point of view is about what we mean by “customer obsession.” In the world of automated visual testing, it’s about making 100 percent certain that the application you build is visually perfect. There will always be debate about what customers want or need in terms of features and functionality, but once you’ve made the commitment as a dev team to build them—you should absolutely be obsessed with making sure the customer experiences them exactly as intended. There is no possible drawback to that obsession in my opinion. And, analysts from Forrester, for example, have proven that a better customer experience directly correlates to with higher revenue growth for the business.

Josiah Renaudin: It’s increasingly difficult to roll out updates across all devices that look and respond the exact way you want them to. Is it even possible to do so without a hefty amount of test automation?

Gil Sever: Absolutely not. Not only are UX variations continuing to proliferate, but release cycles get getting shorter. For example, even if you manage to run a 60 percent coverage manual UI test, regression will start to occur almost immediately. This does not even account for the error-prone nature of manual QA on your UX in the first place. Artificial intelligence technology that can mimic the human eye and brain has opened the door to automated visual testing at scale, and behind that first door is a whole house full of applications and use cases that will be addressed with Visual-AI. This is exactly why we believe a new category is emerging called application visual management.

Josiah Renaudin: How can application visual management paint you a better picture of what your UI will look like and how it will respond once it gets into real users’ hands?

Gil Sever: We define application visual management (AVM) as an extension, or completion, if you will, of application delivery management (ADM) and application performance monitoring (APM). The automated development and management of the application has largely excluded UI until now.

We started with automated visual testing, given the clear path to ROI and the easy point solution integration with the existing ecosystem—especially Selenium, Appium, and rendering services like Sauce Labs, Perfecto, BrowserStack, CBT, etc. In minutes, you added automated visual testing to save dozens, even hundreds of hours, for individual engineers and allow dev teams to release faster in the process. Most importantly—visual bugs were squashed before release. This is huge. Now we know that users will experience a new or revised application UI as intended—our first major impact of AVM.

From here over the next one to two years, we will experience AVM rolling out across the entire R&D team. Front-end developers shifting left will be able to visually test component-level UI elements well before handing off to development. Manual QA teams will have codeless testing capabilities bestowing new skills and new levels of efficiency on this function with which they can dramatically improve automated UI test coverage. DevOps teams will use Visual-AI to monitor and maintain application UI by catching UX issues the moment they emerge through an unforeseen third party impact like a browser or OS update. And finally, business owners will have a fully visualized way to see their brand in market both today and in the past—understanding what changed, why, when, and by whom.

Josiah Renaudin: How important is artificial intelligence for this style of test automation? And do you feel AI is going to be integral to the future of software testing?

Gil Sever: This style of test automation is tailor-made for AI. The necessity here is to view the UI and see it and conceptualize it the way a person would—paying attention to what matters, ignoring what doesn’t. “Looking” at hundreds of screens with a closed end hit list is a rote process for people and one that we simply do not do well beyond an hour or two. Getting someone or something to “see” and “understand” the way the human eye and brain do in an open ended manner, without fatigue or errors, is really difficult to do without Visual-AI. Many people think that pixel-based approaches could work, but anyone who has attempted this approach at scale will know that it generates far too many false positives and often does more harm than good—in terms of time, money, and release cycles.

Josiah Renaudin: Will this level of automation eliminate the need for manual testers? I know testers don’t want to hear that their jobs are being phased out, but do you still need skilled testers to pull off this level of automation?

Gil Sever: This is a tough question, and an important one to explore more deeply in another interview. Frankly, we reject the notion that automated visual testing will eliminate manual testers. In fact, we see a huge opportunity to train manual QA teams on the use of a “codeless” version of Applitools that will give these teams a new professional competency and path. You have to remember—Applitools sees what the eye sees, conceptualizes it the way the human brain does, but the Visual-AI engine does not pass judgement. Human beings judge what is acceptable versus what needs to ticketed in Jira and fixed. The AI makes this judgement massively scalable and accurate, but ultimately manual testers will find a new path to “AVM certification” as the category progresses in the market and creates a new job opportunities, as opposed to eliminating them.

Josiah Renaudin: What’s one major automation trend, outside of AVM, that has you excited for the future of testing?

Gil Sever: We look at test automation trends being driven by AI, IoT, and blockchain in 2018 and beyond. Personally, I watch all these core technologies routinely, as they are all destined to bring substantial value to test automation in ways that are obvious, but more interestingly, in ways we might not expect. That is always the thrilling part for me.

Gil SeverGil Sever is a serial entrepreneur who is driven to solving technical problems with innovative solutions that didn’t previously exist. Prior to Applitools, he founded Safend (acquired by Wave Systems in 2011) and Storwize (acquired by IBM in 2010), and was the COO at Ectel, a NASDAQ company. Prior to this, Gil served in an elite intelligence unit of the Israeli Defence Forces and held a variety of research, development and management positions. Gil has a B.Sc. degree in Electrical Engineering from the Technion, Israel's Institute of Technology, and an M.Sc. degree in Electrical Engineering from Tel Aviv University.

About the author

Upcoming Events

Jun 02
Sep 22
Oct 13
Apr 27