Author:Katie Waldeck

The essential preparation for chatbots

We’ve evolved with our clients over the years; as the face of the contact centre has changed, so have we.  What was once a call centre is now a multi-channel contact centre, dealing with enquiries via the customer channel of choice: phone, webchat, email and whitemail. Chatbots are a natural next step.

As service communications specialists, we spend our days talking to clients and experiencing their customer service functions, whether that’s across large and often multiple contact centres – or smaller teams.  And you can’t deny that artificial intelligence is a hot topic and tells a tempting story: it’ll save you money on overheads and repeat contact, it’ll help you get better information/data on your customers, it’ll be quick and easy for your customers etc.  That said, a contact centre or customer service function always needs to strike the balance between operational efficiency and good customer experience.

Implementing chatbots

AI is still in its infancy within contact centres, certainly in the UK. Some companies are already embracing innovation and implementing this new capability but many others are only just starting to evaluate how they could integrate it.

When speaking to clients or listening to specialists talking about the world of AI, virtual assistants or chatbots, the focus tends to be on the functionality and efficiencies that these tools can deliver. There’s no denying the benefits of AI, but I still question, ‘at what cost to the brand and customer experience?’

As with any new approach to customer service e.g. webchat, it’s still vital for companies to be clear from the outset on the customer experience of a bot and how you want to ‘sound’.    As we work with more and more clients on exploring what chatbots could mean for them and how we can prepare, we talk about two essential ingredients:

Ingredient 1: Tone of Voice

Speaking to your customers in a clear, consistent tone of voice is no different when you’re thinking about AI.  There are constant developments in Natural Language understanding (NLU) – the way machines can process and interpret customers’ interactions and then react accordingly.  But, as with any touchpoint, you still need to make sure that as much as the bot is reacting and responding to customer requirements, there’s a reflection of your true brand promise through its ‘voice’ – text, speech, images and video.

It’s important to look at the brand persona and language they use to make sure they still replicate a human, warm and empathetic tone with customers, as well as any other distinguishing features that set a company apart from its competitors. 

Our core competency is how companies talk to their customers through service channels.  As part of our process, we look not only at what the ‘brand’ or the organisation’s persona is, but we’re also focussed on the customers or end users. Our key aim is to make sure that whatever language we use a) resonates with them and b) delivers the right outcomes for both company and customer.

Ingredient 2: Knowledge Management

One question that came up at the recent ‘AI in Customer Engagement’ forum was:

‘Will there be any circumstances where AI will deliver the opposite i.e. negative outcomes?’

The answer was ‘Yes, if it’s not implemented properly’.  For any bot to be effective, it must understand your customers’ needs as well as what help or outcomes are available to them.

This question led me to a realisation of something that’s true of today; Whether it’s training agents directly or giving them the tools to provide the best customer experience possible, you need to have a good way for customer-facing staff (or interfaces) to easily have all the right information ready for the customer. You need well-structured and accessible knowledge management.

As a customer, we all face the metaphorical wall of ‘No’ or ‘I’m afraid you can’t’ or ‘Sorry, I can’t help you’. Sometimes it’s to do with agent capability and sometimes it’s to do with what information is available across an organisation.

We often find, in the work that we do, that there’s a lack of centralised information or data that clearly outlines:
a) what the company can do to help a customer 
b) what a customer can do themselves  
This applies both to providing the best assistance to a specific enquiry as well as proactively offering ideas or information to offer service above and beyond.

A bot or virtual assistant is only as good as the knowledge and tools that you give it, so the best way to prepare for AI is to consolidate and craft your base of process, knowledge and customer insight. As organisations use knowledge in different ways, there are different systems and approaches to how you can organise your data.  Investing in this sort of taxonomy from the outset will mean you’ve mapped out the complexities of your organisation into something AI-digestible.

Our knowledge management projects empower your business to give customers the answers they need, quickly and clearly, leading to higher satisfaction scores. Read more about our results and Knowledge Management approach, or get in touch for a chat about how we can help you get AI-ready.

High Stakes Customer Communications

It goes without saying that extra care is needed when you’re communicating with large groups of customers about sensitive issues. For example, the stakes are always high when letting them know about their bills, price rises, renewals, changes to terms and conditions or even regulatory messages etc.

The irony is that if you get it right nobody notices. But get it wrong and the damage is instant with large numbers of cancellations, customer complaints not to mention a costly increase in contacts to handle.

But what does extra care mean in practice? We work on high stakes communications from bills to renewals letters. Our approach is to design — taking into account both the needs of the business and consumer — and test our ideas with customers as we go. The result is that design decisions are continually validated. And if there’s a problem, it’s spotted early.

6 tips for getting it right.

1. Success criteria — Decide what the success criteria are in terms of outcomes (e.g. contact volume, renewal rates, CSat etc.). And contributory factors (readability, whether what’s most important is easy to find etc.). These should be based on what you’ve seen previously for something similar.

2. Decide what you need to say — Make a list of everything that you want to say. It’s worth breaking this down into must and nice to haves, just in case compromises need to be made further down the line.

3. Find out what customers want — This could be a qualitative (e.g. interviews) or quantitative (e.g. survey questions with defined answers) study with your existing customers. Or, just put yourself in their shoes and ask others what they think.

4. Design (how to say it) —  Sometimes your needs and those of customers are the same. When they’re not you’ll need to strike a balance. Customer focussed organisations compromise on their own needs. When it comes to creating the actual communication think graphic design, language, structure (what’s most important first) and written tone of voice. Some of these may be covered by your corporate identity manual and you’ll need to work within those guidelines. Our Cheshire CAT (CCAT) service writing rules are another useful guide.

5. Test — Try out what you’re thinking on small groups of consumers as you go. This approach is sometimes called ‘Fail Fast’. You’ll quickly pick up whether there’s anything major wrong and be able to correct it before too much time has been spent. When it comes to testing ‘Think, Feel and Do’ is a useful rule of thumb. Try and understand the logical thought processes of customers (Think), their emotions (Feel) and what the likely outcomes will be (Do). Groups of 6 – 12 people that represent your customer demographic are a good start. Going around the design and test loop 3 – 6 times is not unusual.

6. Validation — Once you have a final design proposal, a validation test is critical. The aim of this is to predict — with a greater degree of accuracy — what all your customers will do when they receive your communication. You can also validate contributory factors, such as readability and whether what’s most important is easy to find etc. Two types of validation to consider are:

• Real world — this may be harder to set up but will give the most accurate results. You’ll need to send the proposed communication, or a link, to a subset of your customers and measure the actual outcomes. You can also do a post contact survey to validate other factors.

• Simulated — this is easier to do but is less accurate. Sometimes it’s not possible to run a real world test for reasons of secrecy for example. With simulated tests you’re aiming to mimic real world conditions in so far as you can. Outcomes can be measured by asking what customers are likely to do. However, saying something and doing something are not the same. So, there’s a greater degree of uncertainty because of this.

For both types of tests the group sizes will need to be large enough to make your findings statistically significant. The acid test is whether what you find meets the success criteria you’ve set at the first stage.

How far you go with design and testing is going to depend on how much is at risk. But when just a 1% reduction in renewals or increase in contact could cost millions, the benefits of being able to understand the consequences far outweigh the cost.