We often talk about how conversational AI can scale-up your customer services. It’s possible to serve far more customers than you could with just a human team of agents. Most companies may have hundreds or thousands of conversations per month, and it doesn’t matter when they occur, as automated conversations are available 24/7.
Potentially, you could have a limitless amount of conversations, all at the same time, but potential and actual are not the same. We need proof to see how far the number of simultaneous interactions can be pushed.
So, what about 4500 concurrent conversations? That’s just a normal moment for Cigna Express Scripts, at any time of day. They’re a global health company, the 22nd biggest company in the US, and they’ve expanded their automated conversations in a big way to cater to the demand.
“We’re an online pharmacy as well as a pharmacy benefit manager, and we’ll make prescriptions safer and affordable for members. We’ll handle millions and millions of prescriptions via mail order, so rather than have to go to your pharmacy, Express Scripts would potentially take your prescription in and then mail it to you. And so it’s at your doorstep when you need it. And many Americans are on maintenance medications. And so that’s a very large convenience.” said Michael Di Troia, Head of Conversational AI, Cigna/Express Scripts.
Cigna Express Scripts’ Michael Di Troia, Head of Conversational AI, and Rachael Dzyak, Head of Voice, recently appeared in a VUX World webinar to tell Kane Simms more.
Don’t build if you don’t need to
So where did they start? Roughly five years ago the team attempted to build their own NLP and ML models, however they soon realised that they could get to market much quicker by exploring already available products.
The best place to start when automating conversations is with a frequent user need that doesn’t require human analysis to resolve. For Cigna Express Scripts, they pinpointed email order status enquiries as a suitable candidate and automated those first. Those emails were costly and time-consuming to service with staff, so they were a perfect starting point.
Then they focused on their voice channel. By analysing the intents behind calls (the needs customers were calling with) they obtained the data from call recordings to create a roadmap for the services that customers most needed. Then they created automated routing that understood the user’s need, then passed the customer onto the appropriate department. From here, they began adding fulfilment journeys, such as prescription ordering and amendments.
Using Conversational AI to optimise
The voice channel they had was an NLP-based IVR. They had a system already in place that had voice recognition and was directing calls for various use cases, but it was legacy and ineffective. They wanted to improve the customer experience.
“What we’ve been able to do with conversational AI is take it to the next level of complex flows, and have the person interact with a bot, to answer the questions, to help them through the processes.” said Michael Di Troia.
Here’s how they did it
So what did they learn along the way?
To make improvements, you need to understand the customer’s needs – that’s why analysing the customer intents in calls helped them improve customer service and streamline interactions.
Don’t build your own if you don’t need to – building an NLP and ML model would have taken significant time and resources. The technology is proven and it’s cheaper and quicker to bring it in rather than build.
Collaboration is part of the process – in order to create the best result possible, they collaborated with business leads to finetune their conversational flows.
Start small and scale up – by focusing routing inquiries first, they managed to create a robust system that did one thing well, and then once that had proven itself they started to add additional channels and functionality.
Of those, it’s important to highlight the extent to which they’ve scaled things up. It’s astonishing. Now they’re handling 4500 concurrent conversations.
We’ve always believed that was possible, because it’s the potential benefit of conversational AI – to have excellent automated customer service at scale – but it’s rare to see companies achieving that with so many customers being served simultaneously!
Cigna Express Scripts have followed best practices. They started small by focusing on a proven customer need, they ensured the wider team was listened to in the design process, and they saw the potential iceberg in the water – building their own tech stack when they didn’t need to – and deftly sailed around it.
This is a great example of how to roll out conversational automation successfully, proving you can provide great customer service with conversational AI at an epic scale.