AI in contact centres: Success stories — and failures
Add bookmarkListen to this content
Audio conversion provided by OpenAI

In our State of Customer Management report 2025, an overwhelming 83 percent of survey respondents said that artificial intelligence (AI) will play a significant role in their CX transformation efforts over the next 12 to 24 months.
AI has, understandably, emerged as a game-changer in modern contact centre solutions. From enhancing efficiency to reducing operational costs and improving customer satisfaction, the potential benefits of AI in contact centres are vast. However, despite its advantages, not every implementation of AI is a success story — in fact it is estimated that 80 percent fail — twice the rate of failure for information technology projects that do not involve AI.
For CX leaders, understanding the successes and failures of AI in contact centres is essential to making informed decisions. In this article, we’ll explore real-world examples of both triumphs and setbacks, uncover the lessons learned, and offer practical tips to ensure AI enhances your customer journey rather than hinders it.
Source: The Future of Contact Centre Agents
The power and the promise of AI
Two key benefits of AI we hear a lot about are its ability to free up human agents’ time by automating routine tasks, and how it can make certain processes simpler for customers, especially when they need to contact customer support.
A clear example of where this worked was the implementation of an AI-powered chatbot by the UK’s Natwest bank. Currently 80 per cent of the bank’s customers communicate with Natwest digitally, and as a result of introducing ‘Cora’, customer satisfaction has surged by 150 per cent.
The bank’s next step is partnering with Open AI. By reducing the need for human intervention and making its processes even more intuitive, Natwest is now hoping customers will report suspected fraud through Cora rather than over the phone, which is currently their preferred method of reporting.
Meanwhile when Westminster City Council in London needed to modernise its reporting process for residents, it adopted generative AI. The council needed to navigate the complexity of public services, including housing, homelessness, licensing and planning, while meeting high customer expectations set by private sector services such as Netflix and Uber.
After consulting with residents, it found that the reporting process could be simplified if they were able to submit a photo rather than fill out forms. An AI tool can then analyse and identify issues within seconds, rather than the residents having to fill out forms.
Westminster Council’s approach has been successful because it did thorough research, conducting discovery sessions with hundreds of local residents and businesses to discover what their pain points were. It then adopted a ‘cyber by design’ approach, partnering with vendors experienced in delivering secure, unbiased public sector solutions.
When AI falls short
The technology isn’t perfect however. Some typical problems include:
Lack of training and testing
One of the most high profile examples of this was in 2024 when McDonald’s stopped using it at drive-thrus after it kept getting people’s orders wrong. Its voice-activated system frequently misunderstood customer orders due to background noise, different people’s accents and complex orders. These are all things that should be expected at a busy fast-food restaurant.
Ensuring the system’s data included different accents, background noises and complex order scenarios, along with rigorous and frequent testing, could have improved its ability to handle these interactions.
Failing to regularly update data
In 2022, following the death of their grandmother, a customer was mistakenly told by Air Canada’s chatbot that he could apply for a partial refund of his airfare due to bereavement. However, Air Canada's actual policy states that this needs to be approved before travel, not retroactively, and his request for a refund was denied. A tribunal subsequently found Air Canada liable for the misinformation provided by its chatbot and ordered it to compensate the customer for the fare difference, along with interest and court fees. Air Canada not only ended up out of pocket, but also suffered reputational damage.
The root cause of the problem was the chatbot's dissemination of incorrect information, which conflicted with Air Canada's actual bereavement fare policy. This suggests that the chatbot had not been given the most up-to-date information regarding the airline’s policies.
Using AI for AI’s sake
Dan Allen, the Deputy Director of Landlord Support at the National Residential Landlords Association (NRLA), says, “There’s always a danger when deploying technology – especially nascent technology like this – without first clearly defining the problems they aim to solve. “With the prevalence of ChatGPT and the like now, there is a real risk of grey IT popping up all over the place with these models so we’re trying to stay ahead of that through education – we're communicating the value that AI can deliver.
“We’re primarily utilising it to improve internal productivity,” Allen says. “Just by maintaining that sole line of sight for now, we can exercise greater control and mitigate any risks to the business – those errant conversations or unintended customer interactions that can spring up when these programmes are placed directly in front of customers. The technology isn’t ready for that yet.”
The future of AI in contact centres
Investing in AI has become a vital part of business strategy, with executives increasingly viewing it as a means to optimise agents' workloads. According to our market study on the Future of Contact Centre Agents, 85 percent of thought leaders in the CCW Europe community plan to invest in agent-facing AI over the next 12 months. Key areas of focus are conversation analysis and intelligent call routing, quality assurance platforms, and productivity-enhancing technology for agents.
In terms of how customers may use AI-powered agents in the future, author and consultant Steven Van Belleghem envisages that these will essentially become personal assistants, calling them “the ultimate friction-removing tool.” He notes on his blog that, “let’s say you want to organise a road trip in Australia. Today, you must do research about interesting places and people to visit, then go to a website or app to book a flight, then to another one to book your different hotels, then to another one to book a car, a restaurant, a museum… Soon, you will ask your agent and they will manage that entire journey for you in a highly personalised manner, requiring zero-click interaction from you.”
We may not be there yet, but there is a growing number of voices claiming that AI has already advanced to a point where it is indistinguishable from humans in customer service roles. Its capabilities are only going to increase as time goes by.