Ways to improve customer services with AI

Along with literally millions of other people this month, I’ve been tinkering around with ChatGPT, the artificial intelligence language model. It’s exciting and scary in equal measures. I’ve been using it in the past couple of weeks to help in my day to day work as a service designer and thought I’d jot down a few ways I can see it helping businesses to improve services.

Knowledge Management for frontline staff

Most customer service advisors use knowledge management tools so they know what the policies and processes are while they’re on a call. These kind of systems help businesses to ensure their guidance is consistently applied. They’re often like internal wikis where various parts of an organisation will place their information for telephony staff to refer to. The challenge with these systems is they can often be quite poorly written and ill maintained because the teams can be funded in a piecemeal way and lack their own consistent funding. Pages can lie dormant for years without being improved, so the content design, layout and general visual language can be very difficult for frontline staff to understand in the heat of a call with a customer. This leads to customers being given incorrect information, inconsistencies in how cases are dealt with and lengthy calls when the advisor can’t find what they need. These fairly common problems can be addressed by hiring content designers and writers to improve the content so it’s easier to find and read. Making improvements to this kind of content can have a huge benefit on the customer experience because the advisor simply gives more accurate information, which saves them having to call multiple times. But as valuable as content design, plain and clean writing and structured information is, it can often be beyond the budget of these kinds of shared services. But with GPT, you can take a complex, dense piece of text and re-write it into plain English that any telephony advisor can understand instantly at a fraction of the cost of hiring someone.

Call listening and categorisation

When customers call up contact centres, they’ll usually need to explain the problem they have to the advisor, so the advisor can help. This is an inspirational source of insight, because it gives you a clear understanding of what’s going wrong with your service, such that people have to resort to calling you. It’s a true signal of the source of customer misery. If you record your calls and can get hold of transcripts, you can feed thousands into GPT and ask it to categorise the reasons people are calling. You could even do this once a day to create a dashboard of health for your service, allowing you to spot a rise or fall in the different reasons people are calling.

Chatbots

This is an obvious one, but chatbots have not been living up to their promise for a long time. Customers often complain that they’re being forced into stilted and unhelpful conversation and will simply avoid chatting to a chatbot. The common pattern is for customers to speak with a chatbot first and then be passed on to a real person via chat to continue the conversation. This has really had very mixed success over the last few years because of the clumsiness and highly structured nature of the conversation. Most conversational bots are little more than conditional trees. But spend just a few minutes with ChatGPT and it’s like speaking with a human. It’s very natural and knowledgable. Once up and running at scale, ChatGPT will be able to have natural chat conversations and customers may learn to trust them again. There’s clearly a margin for error with these kinds of automated chat features, but that’s also the case for humans.

Report writing

For some frontline advisors that have to write a report, such as financial advisors, the writing forms a large part of their workload. These could be generated in a fraction of the time based on a few key facts to a specified standard to allow the advisor to spend more time with more customers.

Clearer terms and conditions

Many customers struggle to understand what it is that they’re signing up for because terms and conditions are written in a way that is meant to be very specific. So it often features language most customers don’t understand. It’s only a matter of time before every browser has a plugin that explains this kind of complex text.

In summary

This technology is clearly going to change the job market forever. But service companies have the opportunity to use it to create new value for their customers by paying attention to parts of their service that they may have neglected for a while.