Anyone who has worked in telecoms technology will have seen new tech come and go. Some have been more hyped than others, and generally the greater the hype (and associated expectations) the greater the initial disappointment when the technology is rolled out. Is generative AI going to be any different given the amount of investment and momentum behind it?
In August 2023, Gartner published its hype cycle for emerging technologies. Guess what was the top of the peak of inflated expectations? That’s right – generative AI. And anyone familiar with hype cycles will know what comes next – the trough of disillusionment. Looking at the hype cycle, it’s forecast that generative AI will reach the ‘plateau of productivity’ in just 2-5 years.
That’s a fairly short timescale for an emerging technology to go from over-hyped to become mainstream. However, the investment that is being made in generative AI is so vast that we need to ask if the trough will be much shallower than normal and will last for a shorter period.
However, there will be a period when some businesses try to apply generative AI to gain tangible business results and ask: ‘Is that it?’ In order to navigate through the initial trough of disillusionment it’s important that CSPs place their AI bets on use cases that can take a pragmatic approach to implementation and that can also show quick wins because there is a significant enough problem that needs to be fixed in the first place.
Where better to start than customer care?
Being a customer care agent is tough. Generally, people don’t call customer care to praise the work that a CSP is doing. They call when they have a problem that needs to be fixed, or to complain. While CSPs have been trying to reduce the number of calls to agents (humans, not chatbots) the reality is that when a customer feels the need to complain to their CSP, most of them call customer care.
Each year the UK regulator, Ofcom, carries out a survey of customer satisfaction levels with UK CSPs. While the results are for the UK, we can make the assumption that these would be similar for the most advanced and competitive telecoms markets.
The Ofcom survey from May 2023 showed that for mobile customers, 77% called customer care when they had a need to complain, and only 17% used web chat. A further 12% of mobile customers said they had cause to make a complaint to their CSP, and of this number, 71% went on to get in touch with their CSP.
In an ideal world, all these calls would be dealt with quickly and to the customers’ satisfaction, but the reality is that after an average call waiting time of 2m 23 seconds, 57% of customers did not have their problems resolved on the first call – and 10% of calls to mobile CSPs were abandoned with people hanging up before even speaking to an agent. It’s fair to say that any help customer care teams can get to improve these numbers would be very welcome.
According to McKinsey, Generative AI can improve productivity in customer operations by 15%-40%. This is all very impressive but it’s not going to happen overnight. Generative AI engines are developing all the time and the more training they have, the more accurate answers they will give to questions.
Generative AI can be used to help customer care by giving them the tools to help do their jobs to improve first-call resolution rates, reduce call handling time, and deal with more calls which reduces call waiting times.
However, it’s important that customer care functions in telcos take a pragmatic approach to using generative AI and get through the initial trough of disillusionment by showing real productivity gains. No doubt, in the next two years, there will be generative AI customer care projects started and abandoned because they were too ambitious from the start.
To avoid this and navigate through the trough of disillusionment, AI customer care projects need to have realistic expectations and AI needs to be used as a tool to help customer care agents. This is why using an AI copilot approach to customer care, where the agent is in control and the AI engine is a supporting tool, can help improve customer care performance.
Typically, when a customer is online and they engage with customer care, the first thing they see is a web chat pop-up where they can enter basic details (phone number, initial query area – network, billing, etc). AI can analyse the language and tone used here to determine sentiment (how angry / upset is the customer) and divert to customer care agents who have been trained in dealing with angry customers.
In some cases, the webchat function can provide the answers to fix the customer’s query and there is feedback (e.g. rate your happiness of how this call was handled) to measure the customer’s experience. Then when the customer is transferred to speak to the care agent, generative AI’s ability to summarise and process data (such as interactions in the chatbot and earlier in other channels) can power a co-pilot experience for the care agent, providing both sentiment analysis, summary of previous interactions and possible resolution to the identified issue the customer is likely to have.
However, the agent should be able to override answers and recommendations from the AI engine and not take the answers from AI as the only source of the truth. In the cases where a customer calls an agent and doesn’t use a webchat, the agent uses the co-pilot to help with the call and provide a faster resolution.
There are many uses of generative AI in telecoms – everything from contextually aware marketing to traffic management. However, the consensus is that it will first be used in customer operations, specifically customer care.
As generative AI is new and improving all the time, it’s important that a pragmatic approach is used to set up fast wins and productivity improvements with no major setbacks. This is why an AI copilot approach to customer care, where the care agent is in control, could provide the initial AI use case for customer care.
Head of Digital Marketing