So you think you want a chatbot? 10 questions to ask before you begin
At Fjord London we recently wrapped a chatbot project with a client and, increasingly, we’re seeing a trend where clients are exploring chatbots and/or conversational UI. On reflection, as a newcomer to chatbot and artificial intelligence design, there are questions I wish I had asked at the start of the project. Each question is important to consider, and many of them have implications on the rest.
Some of the questions may seem obvious—but all are crucial to answer in order to get alignment across your organization—and not least to help the agency that’s helping you build it.
- What is the chatbot for – what types of tasks will your chatbot solve?
Does your bot answer questions by providing customer support? Help and advice? Searching for answers or information? Or does it perform actions or transactions: Is it reserving a hotel room for your vacation to Valencia? Is it ordering you a pizza? This is an obvious question, but is hugely important as you begin to navigate the chatbot scene. There are a countless number of ways to construct your chatbot. For example, many purchase or reservation journeys involve first searching for available options, and then confirming a reservation. How can you move seamlessly between these modes?
- Who is your audience?
Is your chatbot going to be customer facing? Employee facing? Or will it have a broader audience, like the Quartz news app? How frequently will they interact with it – every day, or only when they have a problem?
- What do you want your chatbot to look like?
Will it have a human face, like many virtual assistants? Or will it take the form of a text message conversation between your organization and a user?
- What’s your chatbot’s personality?
Is it a pal you want to get a drink with, or a stuffy, yet trusted advisor? Does your chatbot’s personality change based on who it’s talking to; as in, is it a Living Service that learns from every single conversation it engages with and replies in an individual manner, specific to the user, and adapting to personality and preferences? Does it want to be my friend, my servant, or am I subservient to it?
- How will you establish a relationship?
Establishing a relationship with the user is crucial to the long-term success of your chatbot. How will you introduce your bot? How will you explain if you have a hybrid human/bot experience? Will you be open and honest, or deceitful about the responses that came from the bot? Will you take the opportunity to instil a unique and distinctive personality in the bot, to differentiate it from your call centre staff? Will it push messages at you, send you offer codes, annoy you?
- Where will it live?
Where your chatbot lives has many design implications – is it embedded in your app? Do you engage through existing social media platforms?
- What’s the platform you want to build on?
This really is a question of whether you choose open source, proprietary, or a virtual assistant. Microsoft and Facebook are now active in this area, as well as other chat services in different regions. Amazon Echo is making a strong start but the installed user base is still small compared to for example Apple’s Siri. Siri has an advantage due to its third party integration offers, which are currently key use cases, and will most likely expand over time. Are you going to integrate your service into multiple platforms? Take a look at the newest bots on botlist.
- How do you want it to learn?
How can you define the measures for success in answering individual questions? How does the bot collect feedback to decide whether the decisions it made were correct? For example, understanding how some of the different technologies (Deepmind vs Watson) evolve can help provide a framework for you to answer this.
- How do you want to activate it?
Will users activate your bot by using their voice, text message, or something else? Some people prefer one or the other, some circumstances make voice more convenient. Text has the added advantage of the feel of an on-going transaction, conversation, and experience. Voice is more engaging, but more likely to suffer due to recognition errors, and could be a barrier to entry.
- Do you really want a chatbot?
The final, and potentially most important question to ask. What are the objectives that you are trying to achieve? Which KPIs you, as a business, trying to improve? Is the goal to reduce average minutes spent on each transaction, so that on average you can reduce manual customer service staff and hence reduce costs? Do you have a customer engagement measurement that you could use to measure the impact of a transition to bot support within an existing chat experience? Or do you want to reimagine your customers’ experience, using the chatbot as a helpful way to establish a relationship with them? In summary, chatbots can be awesome and they can be awful – we recommend that you don’t solve every problem with a chatbot solution. If your chatbot is solving a real problem for your users, then go for it. In Part 2 of the Chatbot series, you’ll hear more on successful ways to design your chatbot experience.
This article has also been featured on BrandChannel
Thank you to Leon Bovett for contributions, edits, and inspiration.