How to create a consistent brand experience with conversational AI
Author: Antoine de La Gardette
- Service Data and AI
- Date 12 April 2021
Our Head of Conversational AI explores how to shape a consistent brand experience around your chatbot, including the best practices to adopt and common pitfalls to avoid.
By 2024, Insider Intelligence predicts consumer retail spend via chatbots to reach $142 billion worldwide — up from just $2.8 billion in 2019. But with market growth comes higher expectations of the firms which deploy them. Last year, the volume of customer service traffic through conversational AI climbed by as much as 250%. More than half (58%) of these customers say emerging technologies like chatbots and voice assistants are changing their expectations of companies.
When it comes to expectations of the technology itself, common rebuffs expressed by users to chatbots include: “I already knew that”, “I was expecting some personalised information”, and “That’s it?”. Generic information provided by chatbots is often perceived by customers as a desire ‘to appear smart’. Whilst customers leave a chatbot experience feeling far more impressed when they give them relevant, personalised and contextual information which solves their specific problem quickly.
This means prototyping and testing your chatbot’s user experience is just as important as making sure the technology itself works with the content you plug into it. You can therefore think about a chatbot in two parts. Whilst you’ll start with a pre-planned design of the bot and its content, you also need to build in an ‘interaction model’ which continues to learn and personalise each user’s experience.
These two components are hard to balance when creating your chatbot, which is why TPXimpact has put together a series of recommendations. They explore how you can shape a consistent brand experience around your chatbot, highlighting both best practices and common pitfalls to avoid.
Finding your voice
An organisation’s ‘voice’ is unique to them, and depends on an array of factors – such as the industry it sits in, what sorts of consumers it caters for, and what the brand wants to achieve through its messaging.
In the charity sector, for example, a chatbot’s tone of voice needs to reflect trustworthiness, efficiency, and gratefulness. You can also assess the image your organisation wants to project. As well as the use case you want your chatbot to fulfil. Is it informative? Donation-centred? Or sales-driven?
Thorough user testing and audience research can help you uncover the answers to some of these questions. By retrieving feedback from the users themselves, you can begin to understand how your bot’s language can be mindful of each user’s mood. They could be in distress, frustrated, or embarrassed – it completely depends on why they’re using the bot in the first place.
Balancing tone of voice with UX
Brands have long prioritised consistency in their messaging to consumers. This means the maintenance of an organisation’s tone of voice is no less of a priority when deploying emerging technologies like AI and machine learning (ML).
What firms must remember, however, when applying such an age-old marketing strategy to AI and ML, is that focusing too heavily on tone of voice can be detrimental to the user experience. Brands might try to be funny, in keeping with their light-hearted product range, when in actual fact users just want to get a job done on their ever-growing to-do lists.
Part of this includes providing users with an easy exit. Whilst some customers might appreciate the conversational nature of your chatbot, especially if their request is easily dealt with, customers with more complex requests will appreciate the option to escalate their query to a human agent.
Heather Nolis, T-Mobile’s Senior Machine Learning Engineer, tells Forbes it’s as simple as ensuring words or phrases like “help”, “real person”, or “operator” can lead to an opportunity to speak with a human.
Benefits of a playbook
Whilst the vision your project establishes on Day 1 may be clear, it can often be lost if you don’t write it down. Without a record of it, future employees who take over the project will likely disrupt the brand experience you’ve so carefully cultivated.
And as companies scale their chatbots, replicating them across different departments, they can often diversify their conversational AI too much. This can result in your brand message becoming lost or at least somewhat confused. It becomes unclear who the user is talking to, which can, in turn, discourage them from returning to use the function, instead reverting to a finite amount of human agents.
This is why we recommend putting a ‘Style Guide’ together, which is easily accessible to everyone in the organisation. Publications have long used these to maintain a consistent approach to the way they present content, specifying everything from punctuation to spelling, to symbols and numbers.
In a chatbot context, such a guide can act as an incredibly useful point of reference when trying to maintain coherent interactions between artificial intelligence (AI) and humans. That way, the voice you assign to your brand will remain in place as the bot grows, avoiding the jarring effects of inconsistent messaging.
Iterating sooner rather than later
There is an infinite number of outcomes to the conversations your chatbot can have with its users. By continuing to train the language component of your bot, intent recognition becomes more accurate, which in turn breeds consistency and customer satisfaction. Here’s how you can begin prioritising this process:
Track the number of times a customer says each statement
Evaluate a statement’s relevance to the chatbot’s function
How does this statement impact your bot’s overall language model?
Measure the impact of each statement on the user experience
Weigh up the effort to update responses to this statement on the bot’s intent map