Supporting Vulnerable Customers: Voicebots and Chatbots

Welcome to our month long blog series on how you can better support vulnerable customers in your contact centre. 

Throughout October, we'll be looking at the different communication channels your contact centre might use, and examining how these channels can be optimised to ensure every caller receives an exemplary experience. 

Bots and the use of AI (Artificial Intelligence) are growing in popularity with low code/no code, and drag and drop functionality now widely available, so this week we're exploring how Voicebots and Chatbots can be used in the contact centre.

Building a useful conversation

It's important to follow the fundamentals of conversational design when building a bot journey, whilst remembering this is a human to bot interaction, so the challenges are much more complex.

The bot needs to understand the language the user gives, and needs to give the user the information they need, in a way the user understands.

Before we consider the conversation and words, it’s essential to think of the user first:

• What are their needs and expectations?
• What are they trying to do?
• How can you make this task as clear, simple and easy as possible?

'User trust' is the key word – and it’s fundamental to the success of bot adoption

Who would have thought that we’d be comfortable having a dialogue with a piece of plastic in our homes a few years ago? Now, we’re normalised to trust the output from our smart devices, but if the interactions were painful, or we didn’t get the outcome we asked for, we wouldn’t keep using them.

When we start to design conversations, we begin with a happy path – i.e. we identify what a successful bot interaction would look like, and then test the sample dialogues – one person plays the user, one plays the bot and is only able to respond with content that has been scripted by the conversational copywriter.

Even for a chatbot, where voice isn’t used – we read the words out loud to ensure that they are clear, relatable and match the question or input.

Testing the dialogues

Think about the needs of different vulnerable groups and run the sample dialogues to see how the journey would need to change and how the conversation and content could be amended to suit their needs.

Wizard of Oz testing is where we can prove this out:

• You should have a diverse group of testers (between 5 and 10)
• Brief them on the conversational context (i.e. what’s the use case)
• Let them initiate the conversation
• Stick to the scripts created and see if they hold up
• Adjust the script accordingly with each test

So how can we make bots reliable?

As in life, things don’t always go to plan, and when creating conversations for bots, the same amount of thought and energy needs to be spent on the repair flows.

If “computer - or bot - says no” this can be really challenging for a user. There are many reasons why this happens, and the conversation needs to change to reflect each one.

One of the most frustrating things – for both voice and text is when the user input is not recognised, or the bot makes an assumption about the input which is wrong. This is called Misrecognition.

You can improve this by adding a variety of additional training phrases, like this example:

A bot has a use case to enable customers to report a missed bin collection. Automating this process should reduce the number of phone calls into waste and recycling teams.

Bot is trained to recognise phrases such as “missed collection”, “bins weren’t collected”, “missed my bins” and such like.

If a user referred to the bin as a ‘dustbin’, for example, and this word wasn’t added to the training phrases, then a user would get an “I’m sorry, I don’t understand” style message – the trust is broken and the ease of use has gone as the user needs more effort to make themselves understood.

There should be a minimum of 20 training phrases for every intent, plus colloquial phrases and common mis-spellings too.

Using confirmations and empathy when a user inputs certain answers is also really important to build trust and confidence :

For example, in the missed bin scenario, if the bot recognised the input it could reply:

“I’m sorry we missed your collection today. Would you like me to…” and then display three options the user can choose from.

A few tips we'd recommend

• Using buttons for clearer navigation simplifies the user journey and minimises frustration as this is a functional journey, without the need for the bot to recognise intent as the available options that the bot can do are presented to the user
• Enrich the bot with links to helpful resources and offer these as part of the journey
• Ensure that any escalation is clear and concise. When the bot has not been able to complete the request, and if a hand-off (escalation) to another channel is needed, the data and transcripts pass to the agent so that a user doesn’t need to repeat themselves, and the agent can adjust their tone to reflect that this is an escalated journey.
• Don’t overload the user with fast, immediate responses, and be mindful of sounds that can be used to accompany messages
• Treat your bot as a member of your team - they need to be monitored, measured and regularly trained to ensure they perform the role they were created for

In conclusion

Throughout our Supporting Vulnerable Customers series we’ll be sharing a whole heap of tips and information about a range of different touch points, through a plethora of channels.

Now, you can take any one of these individually and do your best to optimise it, but none of them are a one-size-fits-all solution. The best answer is a mix of several different solutions working in concert.

When advising on how to get the best from these technologies, the strand that always unites our approach is to start from the viewpoint of your customer.

Their needs, their wants, are the paramount consideration and will naturally guide you with regard to which solutions you then employ, how you deploy them, and what content will fill them.

From the music you choose on your phone system or in your videos, the voices that represent your organisation, how you script every chatbot response or recording and in which languages you operate, all depends on who the customer is.

How can we help you provide outstanding CX for every customer?

Trusted by over 350 of the UK’s biggest brands; our award-winning strategic approach seeks to streamline communication, creating a seamless and positive experience that is at once creative, clear and on-brand. Our work has a direct, tangible impact on contact centre performance, customer experience and brand reputation.