James Bourne - 6 September 2017
“There’s all sorts of useful algorithms – the right algorithms are impacting everything that we do,” explains Dr. John Bates, CEO of TestPlant. “It’s about the software learning and refining… it’s still a bit of a black art, but if you can package these things up right and make them consumable, it’s very, very interesting.”
The topic of conversation, naturally, is artificial intelligence (AI). What else could it be? Everyone’s talking about it, every company is trying to get into it – and it’s no coincidence that deep learning and machine learning are at the very top of the peak of inflated expectations in Gartner’s latest hype cycle.
TestPlant, a company with roots in London but also offering an American presence, deals in testing automation software. Yet with the promises of AI beginning to come into fruition, the potential is there to change things.
It’s clear from the outset as well that Bates holds the view of AI giving current employees more room to be creative, rather than giving them a route to the nearest job centre. “Often, when we get involved with customers, they’ve only automated between 30% and 50% of the execution side, and there’s just some things they think is ‘unautomatable’, or untestable, and they’ve still got armies of human beings testing things,” he says. “Human beings are great, but you must be able to make a human being much more productive.
“I don’t subscribe to the ‘everybody’s job is going to be replaced by robots and algorithms’ [view], but I do believe that we have to become much more productive and the human becomes the coordinator of sets of algorithms,” adds Bates. “Automation is currently not all-encompassing – it’s just one aspect of testing in the execution, so it needs to broaden – and there’s still lots of manual pieces in there and we need to make those human beings more productive.”
This aspect of automation not covering all has been the focus of a recent webinar series produced by the company. Yet Bates has seen this effect in motion before in his career. While his predominant focus has been around the IoT with Progress, Software AG, and Plat.One, later sold to SAP, his first startup, at Apama, focused on real-time analytics, with trading making the ideal testing ground.
“When these exchanges went electronic, you saw another massive step change which was the arrival of algorithmic and high frequency trading,” he explains. “Suddenly, simple order entry in brokers was no longer done by humans. You had humans becoming much more productive, but you still had them in there being the expert, the account manager…they would be the coordinator of these algorithms and then artificial intelligence came in on sort of the high frequency trading side of things.”
Hence the opportunity such exchanges provide the perhaps staid testing industry. “If you can give [a human] tools whereby they can be involved in the automation process, they can do so much more,” he says. “They can validate the areas where there are bugs and problems, or they can provide human insight into them, and they can point algorithms and zero in on the right areas for bug hunting and tracking and focusing.
“Imagine how much better our applications are going to be if you’ve got humans focused on sending armies of algorithms out to zero in on particular places and make recommendations about improving the human experience, removing bugs, performance, and so on.”
AI, however, is just one part of an ‘audacious’ wider strategy Bates aims to implement at TestPlant. One of his key focuses is on the gap between business and technology and, ultimately, changing the focus of testing overall. “If you could predict the Net Promoter Score of an application by effectively testing it, you could predict how much it will delight its customers, or the effect it will have on churn,” says Bates. “That’s the ultimate because you’re bridging the gap between technology and business, you’re making the tester much more business relevant, and they want to be, and you’re giving the business visibility into what’s going on which has previously been completely opaque.
“That’s where I think you really start to see testing as a profit centre, not just a compliance function.”