‘Stop hiring humans': Customer service under threat as robots take hold

Telegraph

17 hours ago

Workers commuting on the London Underground have been confronted with a terrifying message in recent weeks.
Scrolling digital adverts, displayed on the Tube's escalators, have urged businesses to 'stop hiring humans' and use artificial intelligence (AI) robots instead.
The spooky messages are the work of Artisan, an artificial intelligence (AI) start-up, which launched the guerrilla-marketing campaign to promote its AI software.
But the ads have hit a nerve with London's commuter class, tapping into deep-seated fears that fake human workers are coming for their jobs.
Hundreds of businesses across the country are now deploying AI workers instead of people. But instead of a brave new world of efficient robotic workers, there are mounting fears that the increased use of AI and chatbots will simply make things worse for customers and workers.
Daniel O'Sullivan, a customer service analyst at Gartner, says consumers have justifiably had concerns about the rise of AI because 'chatbots have historically sucked'.
Lisa Webb, a consumer law expert at Which?, says: 'Not all chatbots are built equally. While some can be helpful, others can send customers round in circles and make it difficult for them to get their issues resolved.'
Replacing drudge-work
For managers looking to cut costs, the pitch to replace human workers with AI software is compelling.
On Artisan's website, it pitches 'your future colleagues' – Aria, Ava and Aaron – and promises customers they can get results 'without increasing headcount'.
For now, Artisan is targeting the drudge-work of business sales, helping companies to automate outbound cold emails and the initial conversations with potential clients, work normally done by a very junior sales worker.
The start-up is not yet dealing with consumer-facing customer service roles, which Jaspar Carmichael-Jack, Artisan's 23-year-old founder, says raise 'more issues' and risks.
In the context of a business transaction, there may be little to lose with using AI to automate and personalise thousands of cold call pitches, job adverts or PR emails which may never get opened.
But when it comes to interactions with consumers, 'the error rate is too high across the board', Carmichael-Jack says. 'That is why people have this anti-AI sentiment.'
That has not stopped hundreds of businesses using AI worker and experimenting with AI customer service, whether the public wants it or not.
Dozens of technology businesses have promised AI helpers that can smooth over customer service functions.
These include AI bots from start-ups like the UK's PolyAI, which bills its technology as the 'most lifelike' voice agents, to tech giants such as Salesforce and its 'Agentforce' bots.
High-profile businesses such as Klarna, the buy now, pay later provider, have already raced to replace jobs once taken by humans with AI bots.
A question of quality
But so far, these fake humans have produced mixed results. In some cases, companies have completely reversed course.
At Klarna, Sebastian Siemiatkowski, the chief executive, went all in on AI chatbots such as ChatGPT, seeking to swap human-led customer service for AI, dramatically cutting jobs.
However, in an interview earlier this year, the Swedish company's boss admitted this AI zeal had not worked out.
'What you end up having is lower quality,' he told Bloomberg, adding that 'investing in the quality of the human support is the way of the future for us'.
So far, the public is not sold on the idea that AI agents are going to lead to an improvement in customer service.
A survey from Gartner, published last year, found that 64pc of people would prefer it if companies did not use AI in customer service interactions at all, and 60pc feared it would make it harder to speak to a human – putting up an AI middleman.
And there are plenty of examples of early attempts at AI-powered customer service getting it wrong.
In 2024, a DPD chatbot swore when prompted by users and told customers that 'DPD is useless' and 'don't bother calling them'.
In April this year, an AI support bot called Sam for the code editor app Cursor went rogue.
After users found a bug within Cursor's service that booted them out when they tried to log in from multiple machines, Sam told customers that this was part of a new policy. Users were not aware that Sam was a bot, leading some to threaten to cancel their subscriptions.
And last year, Air Canada was forced to honour a refund after its website's chatbot invented a policy when interacting with a customer.
Generative AI-powered chatbots, which are trained to speak in plain English, suffer from a problem known as 'hallucination', whereby the AI will sometimes simply make up information if it does not know the answer.
These bugs in customer service bots risk driving consumers away. A survey from customer service firm Acquire Intelligence found that 70pc of consumers would take their business elsewhere if they were let down by a bot.
Such errors and risks mean that some companies that were among the first movers to try out AI agents as a replacement for human workers are already winding back.
In a survey published this month, Gartner found that half of companies that were planning to replace their customer service staff with AI were considering abandoning the plans.
'AI agents'
However, O'Sullivan predicted that rapidly advancing technology and changes in customer expectations mean the shift to AI workers is unlikely to stop completely.
'Perceptions here are changing very quickly,' he says. 'Even in the space of one year, we have people becoming more accustomed to using AI.'
He added that when it comes to customer service woes, people want their problem solved, and the 'means through which they solve the issue is not necessarily the most important thing to them'. If bots get more effective than a human, consumers could quickly decide they prefer them to speaking to a real person.
If the tech industry is to be believed, this shift is just around the corner.
Increasingly advanced 'AI agents' – big tech's latest buzzword – are supposed to be able to take on ever more complex tasks from human workers.
While customer support bots once could only provide basic question and answer functionality, agents will be able to draw information from across a business and function more autonomously.
'Customer frustration with traditional chatbots has typically stemmed from the tools' limited capabilities,' says Heidi O'Leary, a partner at Deloitte Digital.
'Agentic AI goes a step further, allowing these assistants to take actions on a customer's behalf – for example, initiating a return or refund without human intervention.'
For now, she says the most successful uses of AI in customer service have kept humans in the loop, using AI as a tool to boost the performance of human staff, for instance, by quickly drafting emails or notes.
Artisan's Carmichael-Jack says AI bots are 'not currently as effective as a human in a lot of use cases'. But this is rapidly changing, with tech companies building more accurate AI bots and attempting to instil 'reasoning' in them.
He expects perceptions of AI employees to 'shift over the next couple of years because we know it is going to get it right'.
In the near future, he expects that people would 'rather be put through to AI than to an offshore call centre. People will want to speak to an AI agent and you will be that annoyed if you are put through to a human'.