Ask ChatGPT whether AI will replace humans in the customer service industry, and it will offer a diplomatic answer, the summary of which is “they will work side by side”.

Humans though, are not so optimistic.

Last year, the chief executive of Indian technology firm Tata Consultancy Services, K Krithivasan, told the Financial Times that AI may soon mean that there is “minimal need” for call centres in Asia.

Meanwhile, AI will autonomously resolve 80% of common customer service issues by 2029, predicts business and technology research firm Gartner.

There is currently a lot of hype around “AI agents”. That is the term given to AI systems that can operate more autonomously and make decisions.

They could turbo-charge current non-AI chatbots, known as “rule-based chatbots”, which can only answer a set list of questions.

My own recent experience with parcel delivery firm Evri’s chatbot illustrates the existing, non-AI state of play.

My parcel had not arrived, and Ezra (the name of the chatbot), offered to “get this resolved straight away”.

It asked for a tracking reference, and after I had typed that in, it told me that my parcel had been delivered.

I could request proof of delivery, and when I did so it showed me a photo of the package… at the wrong front door. And there was no option to advance the conversation after this “evidence” was shown.

In response, Evri tells the BBC it is investing £57m to further improve the service.

“Our intelligent chat facility uses tracking data to suggest the most helpful responses and ensure the customer’s parcel is delivered as soon as possible, if this has not happened as scheduled,” it says.

“Our data confirms the vast majority of people get the answers they need from our chat facility, first time, within seconds. We’re always reviewing feedback to ensure our services are as helpful as possible, and we continue to make enhancements on a rolling basis.”

On the flipside, rival parcel delivery firm DPD had to disable its less rule-bound AI chatbot after it criticised the company and swore at users.

Getting the balance right between being on brand and genuinely helping customers is a tricky one for businesses to grapple with as they migrate to AI.

Some 85% of customer service leaders are exploring, piloting or deploying AI chatbots, according to Gartner. But it also found that only 20% of such projects are fully meeting expectations.

“You can have a much more natural conversation with AI,” says Garner analyst Emily Potosky.

“But the downside is the chatbot could hallucinate, it could give you out-of-date information, or tell you completely the wrong thing. For parcel delivery I would say rules-based agents are great because there are only so many permutations of questions about someone’s package.”

Resources and money are among the key reasons businesses may be considering the move from human to AI customer service. But Ms Potosky points out that it isn’t a given that AI will be cheaper than human agents.

“This is a very expensive technology,” she says.

The first thing that any business wanting to replace humans with AI will have to do is ensure that they have extensive training data.

“There’s this idea that knowledge management becomes less important because generative AI can solve the fact that their knowledge is not particularly well organised, but actually the opposite is the case,” adds Ms Potosky.

“Knowledge management is more important when deploying generative AI.”

Joe Inzerillo, chief digital officer at software giant Salesforce, tells the BBC that call centres provide fertile training grounds for AIs, particularly ones that have been moved to low-cost areas such as the Philippines and India.

This is because a lot of staff training will have been done, which the AI can also learn from.

“You have a huge amount of documentation, and that’s all really great stuff for the AI to have when it is going to take over that first line of defence,” he says.

Salesforce’s AI-powered customer service platform, AgentForce, is currently being used by a range of customers from Formula 1, to insurance firm Prudential, restaurant-booking website Open Table, and social media site Reddit.

Mr Inzerillo says that when Salesforce first put the platform through its paces it learned some valuable lessons about how to make the AI seem more human-like.

“While a human might say ‘sorry to hear that’, the agent just opened a ticket,” says Mr Inzerillo.

So the AI was trained to show more sympathy, especially when a customer has a problem.

Salesforce also found that not allowing the agent to talk about competitors proved problematic.

“This backfired when customers asked legitimate questions about integrating Microsoft Teams with Salesforce,” says Mr Inzerillo. “The agent refused to help because Microsoft appeared on our competitor list.”

The firm subsequently replaced that rigid rule.

Salesforce has ambitious plans for the continuing rollout of its AI agents, and so far it claims that they are a hit with its customers. It also says that the vast majority of customers, 94%, are choosing to interact with AI agents when given the option.

“We’ve seen customer satisfaction rates that are in excess of what people get with humans – then AI can unlock the next level of customer service,” says Mr Inzerillo.

It has also meant that the firm has cut customer service costs by $100m, but he was keen to play down recent headlines that suggest this has led to 4,000 jobs being slashed.

“A very large percentage of those people got redeployed in other areas around customer service.”

Fiona Coleman runs QStory, a firm which is using AI to offer human call centre workers more flexibility in their shift patterns. Its customers include eBay and NatWest.

While she sees the value in AI improving working conditions, she is not sure the technology can ever replace humans entirely.

“There are times where I don’t want to have a digital engagement, and I want to speak to a human,” she says.

“Let’s see what it looks like in five years’ time – whether an AI can do a mortgage application, or talk about a debt problem. Let’s see whether the AI has got empathetic enough.”

The use of AI in customer service could, in fact, already be facing a backlash.

Legislation currently proposed in the US to move off-shore call centres back to America also requires businesses to disclose the use of AI, and transfer a caller to a human if asked to do so.

Meanwhile, Gartner predicted that by 2028 the EU may mandate what is called ‘the right to talk to a human” as part of its consumer protection rules.