So you launched a chatbot to answer all your customer needs, but you’ve checked the data and it’s barely being used and calls to the call centre are the same as before launch. What went wrong? Why is no one using your chatbot?
The problem with chatbots
Chatbots are the simplest Conversation User Interface (CUI) to implement on a digital product, there are many off-the-shelf solutions which will scrape your FAQs and promise to answer all your users’ questions, but sadly, the simplest solutions often suffer from poor implementation.
Common problems we see with chatbots:
Expecting the bot to do too much
Understanding natural language is tricky for computers to achieve and many chatbots fall at the first hurdle – “I’m sorry, I don’t understand” – killing the experience for your customer. If your bot isn’t that smart and requires a specific response, make it clear upfront what the bot’s abilities are, give them quick reply options, and always pass the user on to another form of customer care if it fails.
Pretending to be human
Be transparent with customers that they are using an AI and make clear the limitations on the actions available to them. If users think they are talking to a human, only to realise later that it’s a bot, you will both frustrate them and lose their trust.
Walls of text
No one wants to read long responses (regardless of how witty or on-brand they may be). Keep conversations natural. Short sentences with clear direction for the user with how to respond.
Overwhelming the user
Human conversations have natural pauses between responses, the pauses for breath and the moments to think. If a machine instantly responds too quickly, or with many short messages, users become overwhelmed. Add pauses between messages, mimic the ‘…’ or ’user is typing’ status updates to make users feel comfortable.
Expecting a silver bullet
Some customers simply don’t want to use digital tools and they will always want to speak to a human. CUI’s should complement your wider ecosystem of channels and not attempt to replace your customer care team. Understanding how your customers feel about CUIs and whether they are a good fit for their behaviours will save you a lot of pain in the long run.
Think bigger than the bot
Conversational interfaces take many forms; medical diagnosis apps such as Ada, branded assistants such as Capital One’s Eno and speech integrations such as Interflora’s Amazon Alexa skill are all examples of using conversations to meet user needs.
So… what is a CUI good for?
Communicating with customers in familiar places
Messaging apps (e.g. Slack bots, Facebook Messenger and WeChat) are the dominant form of mobile communication, being able to identify and respond to customers using the platform they choose will enable you to keep customers happy and engaged without them having to seek out your websites.
Getting customers to the right answer
Carefully implemented bots can be very effective in dealing with routine customer support issues (expected delivery dates, checking opening hours, etc.) which do not need expert human attention. Ensuring customer data is carefully integrated in a safe environment is key and if the query is outside of the bots ability, ensure a seamless handover to humans.
Some users are unable to use a physical interface either temporarily hands full or through permanent limitations. Other users simply may prefer using voice rather than a screen. Implementing a CUI using voice or messaging enables more ways for your customer to communicate with you.
Available when the customer is ready
Conversational interfaces do not need staff to maintain in real-time and are available 24 hours a day, enabling customers to self-serve at a time that suits them.
In a post-COVID-19 world, enabling users to interact in a non-touch environment will be important for businesses offering shared devices such as kiosks and PIN pads.
How to fix your chatbot problem
We have created a range of conversational interfaces here at Cohaesus, from internal tools for staff to update timesheets within Slack, to customer care tools for the energy sector which allow customers to safely change their account information.
We are an insight-driven company and our first step would be to analyse your existing approach to understand where it might be falling short. By examining the usage data and by speaking to your customers we can identify why your bot might not be performing.
Rapid prototyping of your conversation using tools such as botsociety, Hubspot ChatBuilder or Voiceflow as well as roleplaying techniques will enable us to validate whether an updated approach fits the needs and expectations of your audience. It’s far quicker to validate a prototype than to develop a full product, saving time and resource.
We recommend developing small conversational UIs which complement your existing services through a hybrid model. Take a subset of the customer enquiries from your staff and enable customers a convenient new way to meet their needs. Experiment with extending your CUI’s range of skills, continuously monitor how users are interacting and optimise your approach.
Get in touch
If you’d like to know more about conversational interfaces and whether they may be a fit for your business, get in touch, we would love to help!