LONDON -- Telco AI Summit Europe -- Paul O'Brien knows a thing or two about artificial intelligence (AI to the aficionados) in the telecom sector. As the director of AI, cybersecurity and operations for BT Labs, he has a leading research role for the UK's biggest service provider. So when he provides his assessment of how AI can help BT as it wrestles with fiber deployment, job cuts and a depressed share price, the operator's head honchos should be listening.

The good news is that AI's impact goes far beyond the "chatbots" BT and other service providers are now using for customer services. As BT tries to bring full-fiber networks to 4 million UK properties by March 2021, AI has been critically important in boosting the pace of deployment to about 23,000 premises per week, says O'Brien. "That is a three-fold improvement on last year and that is partly by deploying tools based on AI," he told attendees during a keynote presentation here today. "It is not visible, but it is having a bigger effect on customer lives than just through chatbots."

It is making a difference at a more mundane level, too, as BT embeds AI tools in the systems and processes used throughout the organization. The schedules of the operator's 25,000 field engineers are determined partly by AI engines, says O'Brien. BT is using the technology to figure out what resources it needs and how customers could be affected. Routine tasks are being automated.

For all the latest news about automation, check out our dedicated Mobile content channel here on Light Reading.

One example of its impact is in cybersecurity. BT currently maintains 15 security operations centers globally and witnesses around 4,000 attacks every day. It has been developing AI tools to support analysts with speedier threat detection. Unlike a human, those tools can sift through millions of events and rapidly flag up common patterns. "There are 600,000 network events a second," says O'Brien. "That is the scale of the data set."

The bad news, for an operator that looks so overloaded with employees, is that AI does not hold out much opportunity for job cuts, according to O'Brien (staff, naturally, may take a more positive view here). Citing industry research, he says less than 5% of occupations could be fully automated. For the foreseeable future, AI is primarily about augmenting rather than replacing jobs. "It is more Iron Man than Terminator," says O'Brien.

Figure 1: Take Your Pick AI still looks more like Iron Man than The Terminator, to BT's Paul O'Brien.

While staff would undoubtedly prefer Robert Downey Jr. flying to the rescue than Arnold Schwarzenegger on a killing spree, company bosses are under pressure to make BT look more efficient as investors pay growing attention to metrics such as revenues per employee. Last year, BT made just $280,000 for every worker, less than peers including Deutsche Telekom, Orange, Telecom Italia, Telefónica and Vodafone. It is nearly halfway through a plan to cut 13,000 back-office and managerial jobs, and yet its overall workforce has shrunk by just 1,300 employees in the last two years, to 105,026 at the end of September, because of recruitment in other areas. Thousands of engineers have been hired for network rollout.

More sophisticated AI systems in the future might offer more scope for cost savings through job cuts, of course. Indeed, to critics, the AI that shows up in today's telecom networks is really pattern recognition, rather than AI in the strictest sense. O'Brien alludes to the difference. "Deep neural networks are powerful too, but more traditional AI-like statistical techniques are still applicable," he says.

The immediate issue BT and other telcos face is ensuring they are "AI-ready," he says. That means having the right data assets, putting standards and ethics frameworks in place, acquiring the necessary skills and being able to adapt quickly to AI demands. "When we talk about deploying 5G there is a long lead time," says O'Brien. "This is software and it has a quicker time to market."

Related posts:

— Iain Morris, International Editor, Light Reading