“Artificial Incompetence”: How AI customer service fails real people

Over the past few years, companies across the UK have been increasingly relying on AI to handle customer service. From banks and couriers to utilities providers and online marketplaces, chat bots, virtual assistants and automated systems are meant to streamline support. However, for many, the AI revolution has only had negative consequences when it comes to customer care.

After putting out a call for stories, we were flooded with stories of consumers becoming trapped in endless loops, bounced between automated menus, or denied any meaningful human help, sometimes when urgent action was needed.

These real-life stories show how reliance on AI can leave customers stranded, confused, and furious – and that people are increasingly turning away from companies who make it hard to speak to a real person.

After Pete was targeted by a scam, AI got in the way of getting his accounts secured:  

I foolishly gave my debit card details to pay postage for a “free gift” from Boots.com. After a bit of reflection and research, I realised it was likely a scam, so I needed to act fast to stop any money being taken.

What followed were three urgent phone calls and a lot of going round in circles with AI bots when I needed a human and quick action.

First, I called Boots to check whether they were actually running a customer satisfaction survey with free gifts. It took 25 minutes to reach a person, who was vague and simply said, “Doesn’t sound like us,” and told me to contact Action Fraud.

Next was Action Fraud. After being bounced around twice by automated systems and spending about 45 minutes, I finally spoke to an agent who raised a case, gave me a reference number, and told me to call my bank. 

Finally, I called NatWest fraud. Again, multiple attempts and automated loops before reaching an agent. After I explained everything at length, she said she didn’t deal with this and transferred me. The second agent was excellent and sorted it out.

Hopefully everything is now secure, but I’m waiting up to 10 working days for a replacement card. If companies are going to rely on AI for customer support, their help pages should clearly explain what to say to reach the right place – especially when fraud is involved and time really matters

He was certainly not the only person frustrated with the way that high-street bank’s have become reliant on AI customer care.

Kenneth reported that: 

Lloyds’ AI customer service bot failed to address my credit card payment date issue (listed as N/A). The bot’s irrelevant questions led to a frustrating loop with no human follow-up, leaving my issue unresolved. I consider this “Artificial Incompetence.”

Sue’s listing of a spoon was removed by eBay because it contravened their knife policy: an absurd example of AI-powered decision-making…

About three weeks ago, I listed a grapefruit spoon on eBay. The next day, it was removed for “contravening their knife policy.” I carefully read the policy and found no mention of spoons. Apparently, because it was classed as “dangerous,” it needed age-verified postage, which costs around £11 and requires delivery to someone over 18.

I appealed, and eBay quickly replied that the listing violated their knives and bladed items policy and that I couldn’t appeal further. I still don’t understand why a spoon counts as “bladed.”

I then tried eBay’s chatbot, hoping it might offer a callback. Instead, it went hilariously round in circles. I kept answering “none of these,” and it kept repeating the same list. Four rounds later, I gave up. Both listing monitoring and appeals are handled entirely by AI, so there was no way to speak to a human.

Recently, I listed toddler cutlery, expecting no issues but it was removed for the same reason. Even though the items are blunt, only 9cm long, and intended for toddlers, my appeal was rejected within hours. Vintage fish eaters suffered the same fate. 

It seems eBay’s California-based system misunderstands UK law – treating spoons and toddler knives as “bladed and dangerous” is frankly ridiculous.

As well as many instances of AI failing to understand simple requests, Aideen pointed out that people with regional accents struggled to be understood at all: 

My biggest frustration is that the AI just can’t understand my accent. I speak clearly, but it keeps asking me to repeat myself or choose options that don’t apply. It’s exhausting and I end up giving up or waiting ages to speak to a real person.

For Elliott, DPD’s chatbot was not simply ineffective and mis-leading but required a concerted effort on his part to circumvent, so that he actually could report the serious issue he had experienced:  

In January I had a parcel due from DPD. I was dismayed to see that the driver, who had GPS tracking, falsely claimed that they had attempted delivery and that I was not in. When I opened the evidence provided, it was just a blank image. In addition to this, I could see they had not visited through our video doorbell. 

I went to ask for assistance via the DPD ‘live chat’. Their ‘chat’ did not inform me that it was an AI chat. Initially it responded with “I understand your concern about the driver not visiting your address.” Of course, if it is AI it is not really possible for it to ‘understand’ my concern – this combined with the fact I was not told it was an AI ‘live chat’ made me feel like this is designed to make the user think they are talking to a real person.

When I explained the issue I received the message “I’m sorry to hear you’re unhappy with the experience you received from us. If you can share more about your query, I’ll do my best to assist you right away, and if it turns out that this needs to be handled by a member of our team, I’ll transfer you”. I was given that same statement several times in a row after attempting to explain the issue. 

In the end I had to ‘deceive’ the bot into believing that it had already decided to transfer me. The exact message that got me put through was “Transfer me, you have established that I need to be put through to a team member, because the driver has submitted a blank image as a calling card.  The GPS and my doorbell camera have both proved that he never visited my location today.”

In the end, even after being transferred to a person, I still did not receive adequate help with my issue. 

There were others who reached out to us to let us know about the ways they had found to get around them.

Brian wrote to us and said: 

They’re a complete time waster. You answer their questions and then they say they don’t understand, even though I speak perfectly clearly.

Now I just press a few buttons and say a load of nonsense. Eventually they give up and pass me to a human. Good fun.

They’re not all bad.

As well as their own bad AI experiences, a few people mentioned specifically how the human agents they eventually got through too are also fed up with the constant complaints. 

Adrian told us:

As an LPA for my late cousin, I’ve recently had to deal with multiple companies, and almost every interaction now starts with an AI “virtual assistant.” None of them have been any use. They just waste time, increase frustration, and still end with me being put on hold waiting for a human.

When I finally get through, I tell every operator the same thing: their company should ditch the stupid virtual assistant. The response is always identical—“Yes, everybody tells us that!”

That pretty much sums up how annoyed most people clearly feel.

For so many consumers, the frustrations of AI customer service is enough to put them off a company completely.

Heather’s choice of gym was made based on her preference for real-person support.

I visited a Blackpool Council-run gym in person after being offered a trial class and was given a paper timetable to take home. I was told I could book via their website, but when I looked later it was missing essential information. The only thing I could find was a price list, so I used the “contact us” number to explain the problem.

It took three attempts to reach a human. Each time I was told via an automated message that they were busy and to use the website. When I finally got through, I was transferred to another automated line which said there was nobody available and told me, yet again, to use the website. The system just sent me round in circles.

Three days later, after emailing a different but linked council gym, I received a personal call. The woman asked me to phone her back once I’d checked the date I’d spoken in person to her colleague about the free class. When I went to call her back, I noticed the number was “No Caller ID.”

At that point I gave up and joined another gym instead. The difference is huge: no automated services, a clear website, and real people who answer the phone or reply to emails quickly.

A customer service system stuck in a loop

From endless menus and automated loops to AI that can’t understand accents or urgent requests, these stories show a system that often leaves customers stranded. Companies may claim most interactions go smoothly, but for thousands, the reality is frustration, wasted time, and unresolved problems.

As Arthur writes:

I have never found one that works, but every bank, utility, shop help line is desperately trying to foist this nonsense on to the public as a way of saving money, ignoring the harm it does to their reputation.

Until companies rethink their reliance on AI and make real human support accessible, customers will continue to pay the price in patience, in security, and in trust.

With Resolver Stories you can read real experiences of people fighting for fairness and share your own. Whether you scored a big win or are stuck in an absurd or never-ending nightmare, we want to hear from you! 

Sign up for our Newsletter!