The prevalence of chatbots to assist with customer experience (CX) tasks is growing and set to increase according to research platform Statista, which predicts the market will rise to $454.8 million in revenue by 2027, up from $40.9 million in 2018.

From automated calling services to online complaints, bots have a lot to offer in enterprise – and for big brands such as EasyJet they really came into their own during the pandemic .

However, as a recent panel session at DTX Europe discussed, for the bots to interact with customers effectively, they need to be able to gather and utilise the necessary data required to make CX more personal.

“It’s not one size fits all,” said panel chair Neil Chitre, a senior consultant at Davies Consulting.

“It’s really important that you do your research to understand your customer journey, the type of customer you have and your demographic,” he added and stressed that it was equally important to know when chatbots may be better left out the equation altogether.

As Hugh Curran, director of product marketing at CX-based software firm Freshworks pointed out: “If you have a 78 year old woman who wants to get in touch with her health insurance company to check if she has a test at the hospital, it’s a no brainer that a chatbot isn’t going to be the correct form of communication.”

To bot, or not to bot


One thriving use case is NatWest’s chatbot Cora, which has a presence across a number of the bank’s digital platforms, according to senior product manager Bill Agar, although, he stressed: “it’s not everywhere”.

Cora was developed by the banking giant in 2017. It engages in over 200,000 conversations per week and 20% of these are then handed over to an agent “by design”.

“We have a number of rules,” Agar explained. “If the customer wants to talk to a human, then we hand you over. If Cora gets confused three times, we’ll hand you over. If you say ‘no, that wasn’t helpful’ twice, we’ll hand you over.”

He added that one question the designers and programmer always ask is: “Can Cora actually answer the question?”

Agar advises enterprises looking to deploy chat bots to ensure that they take the time to analyse the conversations and data between their customers and chatbots to see where the questions are coming from to work out where the bots might best be used.

The product manager acknowledged that there were some scenarios where agent intervention was inevitable: money can be an emotive topic and some customers will want to talk directly to agent about their affairs.

However, he added Cora covers a broad enough range of topics for customers to want to engage with it.

Arm yourself with customer data


According to Cheryl Allebrand, a senior digital consultant at IT consultancy CGI UK, personalisation is key to gaining repeat business.

“If we think back to when we weren’t doing everything online – say a coffee shop – one of the things that keeps you going back to the same place is someone who sees you, recognises you, knows what you want and can get it to you quickly.”

Personalisation to many companies is ‘did I put insert a name here?’ but Allebrand argued that businesses need to do more.

“Enterprises need to understand their customers better – which is something that they’re not capitalising on currently. For instance, we should allow people to let us know what channels they would prefer to be served on and what languages they would prefer,” she said.

According to Allebrand one challenge firms face is the ability to access CX data easily – as it tends to get stored in siloes.

“All of the companies have all of the data… but what’s important and are those things connected?” she asked.

To address this issue, Agar described how NatWest breaks data down into customer type.

He explained: “We know this customer is a credit card only customer, so when it comes to a payment, we don’t need to ask, ‘was this on your debit or credit card?’ so we skip that question.

“It’s about using the data you have to segment your customers in a way that you can skip questions -since every question you give is an opportunity for confusion or for someone to drop out”, he warned.

Agar revealed that the bank has also started tapping into its customer’s transactional data to create a transaction decline route using API integrations.

When asked: “Why has my transaction been declined?” Cora can select the transaction and find the reason code. The code then prompts the bot to provide a list of suggested teams that the customer can speak to.

“It’s so easy and it’s a lot quicker than waiting in a queue to talk to someone or waiting for an agent to turn up with that information,” said Agar.

Education, education, education


Despite efforts to better the bots, Chitre said that customers may be “put off” after having bad experiences in the past – “so how do you get those customers to give the technology a chance?”

According to Agar, it is about persistence. “It’s how aggressive are you going to be, right? If someone says, I want to talk to an agent straight away, we will try an educate them and say, okay, look, give me a try, or, I can help you get to the right agent.”

He added that approximately 6% of Cora’s conversations are people who want to talk a human. One in 10 of those people will persist with Cora, however the rest are set on speaking to an agent.

Curran concluded that successful implementation of chatbots boiled down to listening and learning about what the customer wants. “We need to educate ourselves first and then improve the processes,” he said. “Then the next time people come back to use the chatbot, they’ll say,’ this is different, this is better than it was’, and they’ll give it a go.”