BT exec stokes fear of AI impact on jobs by likening staff to horses

Every morning, BT saddles up its beasts of burden and sets them to work splicing fiber or answering phone calls. They trot long-faced around the UK telco's call centers, braying nervously as colleagues are carted off to the local knacker's yard. Even the horsiest-looking telecom-sector workers are hard to mistake for real equine mammals. Unless you are Harmeen Mehta.

"I don't know how horses felt when the car was invented, but they didn't complain that they were put out of a job; they didn't go on strike," BT's chief digital and innovation officer told Raconteur in apparent response to the charge that artificial intelligence (AI) may lead to widespread unemployment.

Mehta is, of course, quite right to argue that horses were uncomplaining about automation, bar perhaps the odd whinny, just as farm oxen voiced few intelligible grievances when the plough straps came off and the machines came in. You didn't equate oxen with militant trade unionism or see them waving placards and desecrating the land in protest. Nobody ever called a horse a Luddite.

There was a sensible point buried deep within her remarks – that fear of negative consequences can hinder technological progress – but her interview was quickly picked up by the UK's Guardian newspaper for all the wrong reasons, as far as BT would be concerned. The UK increasingly seems like a country where the ruling classes are badly out of touch with the ordinary man or woman. Now we have a senior telco executive likening employees to animals.

Hemorrhaging staff

But it would have been troubling enough had Mehta left out the horse analogy and merely said people should stop their complaints. BT cut nearly 7,000 jobs, including contractors, between July and September this year, a fact little noticed by the mainstream press. The drop represents about 5% of its entire workforce. CEO Philip Jansen, due to be replaced at the start of next year, reckons BT can eliminate 55,000 jobs altogether by 2030. AI will claim about 10,000 of those, he previously told journalists.

Even before AI takes root, big telcos are hemorrhaging staff. Deutsche Telekom, Europe's biggest, cut another 2,500 jobs in the recent third quarter and warned that nearly 7% of jobs at its T-Mobile US subsidiary – which employed exactly 66,226 people in September, according to its report – would be laid off in the next three quarters. Telecom Italia has cut about 2,000, or 4% of the total, since the start of the year. Vodafone, which had an average of 98,103 full-time employees in its last fiscal year (ending in March), aims to cut 11,000.

Headcount at major telcos

(Source: Light Reading, companies)
(Editor's note: BT and Deutsche Telekom data on US staff are for full-time employees.)

It is much, much worse in the US, where staff can be dismissed at a day's notice and social safety nets are threadbare (it's a progressive country, apparently). AT&T has cut more than 10,000 jobs this year, 6% of its total. Verizon has dispensed with 6,600, also 6%. A planned reduction of 7% at T-Mobile US would equate to roughly 5,000 jobs, including staff who aren't full time, and follow the 4,000 cuts that happened last year. Charlie Ergen's Dish Network, which discovered terrestrial mobile networks about 20 years too late, says hefty layoffs are coming.

Operators can't blame all of this on high inflation and other immediate factors because they've been at it for years. AT&T, astonishingly, has shrunk by more than 115,000 employees in less than five years. France's Orange has shed 14,000 jobs over the same period, about 9% of the total. Spain's Telefónica employs 16,500 fewer people than it did at the end of 2018, when it had about 120,000 on its books. Asset sales explain some of this. But many operators could report inflation-busting growth in revenues per employee if that metric were shown.

There is acknowledgment among executives, especially those on the technical side, of the impact technology has already had. A case in point is the network operations center (NOC). At one time these were dimly lit beehives, redolent of coffee, where people sat behind giant screens and tapped the occasional key, looking as if they had just stepped off the set of a drama series about the NSA. By late 2019, Scott Petty, now Vodafone's chief technology officer, was eyeing a "zero-touch NOC." Staff numbers had already fallen from about 1,500 to the "low-single-digit hundreds," he revealed.

For a sense of how low an operator might go on headcount, look at Japan's Rakuten Mobile. Earlier this year, it was operating a nationwide network with an operations team of just 200 people, according to Rabih Dabboussi, the chief revenue officer of Rakuten Symphony (a subsidiary that sells telco products). An equivalent "legacy" operator would normally have between 6,000 and 7,000, he reckoned.

The cloud could take a further toll on telco staff. Operators like BT and Deutsche Telekom are holding out with efforts to build their own clouds for telco workloads. Others are inviting hyperscalers or technology specialists to do it for them. In that scenario, parts of an operator's network could move into a highly automated public cloud facility. If government rules or service requirements make that tricky, hyperscalers are now offering to bring their hardware and software into a telco's own facilities and effectively take charge. Operators would no longer need their own staff to manage and update those resources. A company like AWS could theoretically do this for dozens of telcos.

Trusting the machines

Automation's apologists love harking back to previous episodes that show fears were misplaced and that technology ultimately gave rise to new opportunities and jobs. But AI – if it really fits that description – is not the same. Anything that approximates human ingenuity is quite obviously a direct threat to your livelihood, whatever it might be, just as a younger and more skilled professional might one day replace you.

The current crop of "generative AI" is arguably not intelligent at all – just a very sophisticated pattern-recognition system that speed checks what humans have already produced and then churnalizes it like a lazy, unthinking reporter (we've all been there). Its bizarre hallucinations and cock-ups prove it does not even have a childlike understanding of what it is doing. Genuine AI would tick that box, and it is purely science fiction.

Yet people are struggling to tell the difference. Buying into the AI hype, organizations are clambering over themselves to use technologies produced and ultimately controlled by a small number of Internet billionaires. And while these might not be intelligent, they are extremely complex and potentially dangerous systems that hardly anyone understands. This complexity, moreover, is increasingly hidden from view with phenomena such as "low code, no code" programming. Engineers can use Microsoft's GitHub Copilot to write software. Intent-based networking means specifying an outcome and letting bots do the hard part.

But systems are not becoming less complex – quite the opposite – and networks are growing. As the telco workforce shrinks and the original innovators retire or die, the upshot is there are fewer people around who know how those systems really work. This might be acceptable if there were a benign, trustworthy and infallible AI to look after them. The likes of ChatGPT, Llama and Anthropic hardly qualify.

Parts of Australia were plunged into telecom darkness this week after a network crash at Optus affected more than 10 million people. For millennials who take smartphones for granted but wouldn't know a basestation from a biscuit tin, it must have been a terrifying glimpse of what their parents endured last century, when conversations were mainly in person and meetings had to be arranged in advance. Lasting 14 hours, it would certainly have carried an economic cost given society's dependence on mobile. Engineers eventually fixed the problem. In future, it may be down to the bots.