Case study: Start small, prototype and action feedback

In CX Network’s recent Expert Insight eBook: Chatbots In CX, Toni Kershaw, customer access manager for Calderdale Council in the UK, explained how 24/7 live chat was the first step toward the body's roll out of virtual assistant VIRA. 

Implementing live chat

Kershaw and her team met some initial skepticism from internal stakeholders about implementing live chat, but after assessing data that mapped customer needs, Kershaw recalls: “We truly believed that it would transform the customer’s experience of interacting with the council.”

The pilot was rolled out within eight weeks, Kershaw said: “We implemented it rather quickly but we also did it fairly quietly. We had heard a few horror stories about how chat functions can fail, because of teams not being prepared for the contact volumes and we definitely didn’t want to start with customers receiving a busy message.”

A monitored inbox was used to manage the chat and on launch Kershaw recalls this was practically a full-time job due to the large influx of tickets. Within the first two months, the council received around 150 pieces of feedback from customers that were used to optimize the website and service.

“It was brilliant because we got real-time feedback, we were able to act on it straight away and then say to the customer we’ve fixed that link for you. In most cases they were quite impressed and grateful. That sort of feedback is invaluable to us,” says Kershaw.

The body even advanced to provide preemptive support to customers, says Kershaw.

“Proactive chat has been a brilliant tool for us,” she notes. “It allows us to say to customers: ‘You seem to have been on this page for a long time, are you struggling with anything?’ We can then assist them before they leave the page and we lose them.”

Jumping from live chat to a virtual assistant

In early 2019, the council took their chat offering to the next level by rolling out a virtual assistant.

First, the team examined live chat transcripts and emails to create a pool of 40 query topics. These topics, which included the likes of ‘Can I change my address?’ or ‘Can I amend a direct debit?’, and their relevant responses were loaded into VIRA, the virtual assistant. Some topics were flagged to automatically escalate to live agents, including the likes of mental health issues or death reports. Variations in language had to be considered to minimize the chances of VIRA asking the user to reword their question.

Within the first few months VIRA was running at a 70 per cent success rate with chats in and out of office hours. Unanswered cases would be escalated to human agents.

Test and learn approach

Seven weeks after rollout, the team continued to monitor VIRA and improved its responses to streamline experiences for users. Frequent meetings were held to review and test interactions, with more information added to boost success. Users were initially asking lengthy and complicated questions, so the bot was adapted to request that customers keep questions short and simple.

Early tests have indicated that VIRA has the capability to handle up to 80 per cent of council tax queries – one of the most complicated areas of council services.

VIRA has been handling waste enquiries since December 2019 and is successfully answering almost 80 per cent without escalation to a human agent. This has been a great way of helping to manage seasonal demand, in particular through the recent flooding crisis. A Christmas recycling campaign was used to launch this initiative where customers could ‘Ask VIRA’ what they could recycle over the festive period, which proved to be very popular.

Looking ahead, the team will continue to examine VIRA and will refer to usage stats and customer feedback questionnaires to guide their steps on how best to educate VIRA and give it new functionalities – like the ability to cope with longer questions.

Lesson

To reach a point where human agents are freed of certain repetitive tasks, a steady and continuous test and learn approach is best. Live chat can be a useful learning arena that will facilitate a smooth entry into advanced chatbot capabilities like virtual assistants.