Chatbots have a very mixed perception in the public’s consciousness. If you said the word chatbot to a person on the street, chances are they may think of a virtual personal assistant, like Siri or Alexa.
Another member of the public may think of the sensationalist bot stories we read in the papers, such as bots becoming inadvertently racist, or creating their own bot language that human kind can’t understand. And on the complete other end of the scale, they may even envisage blockbusters starring AI that are set on global domination, or highly publicised thoughts of some public figures that thinking-machines could be ‘weapons of terror’.
It’s safe to say that the way people perceive virtual assistants and chatbots is very confused. On the one hand, they are a silly bit of technology that falls over, on the other, they are ultra-intelligent and about to embark on world-domination. In truth, neither image is reflective of, what is essentially, an enterprise level technology that is being used to aid a variety of businesses in a number of purposes, from customer service to internal HR.
For businesses, chatbots have the potential to be an incredibly financially-efficient solution. Take a use case that is commonly explored today: customer service. At their best, a chatbot can solve a customer’s query on its own, reducing the human resource required by a brand and satisfying the customer’s need – which we know is the best way to have customers coming back for more.
However, the mysticism and confusion around chatbots means that, too often, businesses that are building and deploying AI and chatbot solutions focus on the wrong thing – trying to make their chatbots good at chat. This is not to say a natural conversation flow isn’t important to a customer experience – it is. However, brands should remember their bots are not going to be used in the same way Siri and Alexa are – they are there to fulfil a business function, not answer questions on the weather in Barcelona tomorrow.
Developing a chatbot requires a lot of resource and time, so in preparation to go live, brands should be focusing on what is most important to users: getting a useful answer rather than the dreaded “I do not understand”.
Let’s go back to the example of a customer service bot. A study by Ubisend earlier this year revealed that when communicating with chatbots, 68% of consumers say that reaching the desired outcome is the most important facet of their interaction.
>See also: Chatbots: catering to the instant shopper
Contrastingly, only 33% say that they’re looking for a personalised experience when interacting with a chatbot. This proves that while having a chatbot that uses high-quality language is nice, consumers ultimately just have a question and need an answer.
So, how can companies develop chatbots that are answer-driven?
Don’t just stick to the script
Traditionally, chatbots have relied on scripted chat, which is essentially a predefined response to a predefined user question. The issue with the scripted approach is that it limits not only the data the chatbot can pull from to get the customers answer, but also if a customer asks simple question which does not fit the script, the customer gets the dreaded “I don’t understand response”. Too often, the result is repeated miscommunication and customer frustration rather than an accurate and helpful response.
An unscripted chatbot means the AI is able to respond better when a response sits outside of the narrow scripts. With less limitations, and the ability to pull information from the businesses’ available data, unscripted chatbots are far more likely to understand the question, and provide a helpful answer.
Automated training of chatbots
Rather than humans having to manually input all the questions and answers they can think of to service every possible communication, automated training dramatically improves chatbot capabilities.
By plugging into an API that automatically trains the bots, brands also avoid the issue of question variants – as hundreds of variants of a question can be programmed for in minutes. The same task would take several hundred man hours, so businesses are saving money, as well as increasing the likelihood of the customer getting the answer they need first time.
The use of suggestions
In the case that the AI has low confidence in answering a question, offering multiple options from a broad set of knowledge is another option to help the customer along the way to their answer. This is an alternative to defaulting to “I don’t know” when the chatbot is unsure, as many chatbots do for fear of generating the wrong response.
According to Ubisend, almost 70% of chatbot interactions are based around simple service enquiries – such as finding out about opening times, or asking for more information about a product – meaning the broad knowledge will help the end user. The result is that the chatbot will have a much higher engagement and better results.
Chatbot use and development is a lot further along than the public sphere would suggest, but there is still a long way to go. In the future, businesses will undoubtedly use chatbots that are both human-like and useful, but the reality is that we are not there yet. In the meantime, businesses should treat chatbots like any other solution, and design them to achieve their business purpose.
By ensuring that your chatbot is focused on the viable means of actually getting the consumer their solution – rather than how polite it can be or how human-like it is – brands will be able to successfully satisfy a consumer’s needs.
Sourced by Adam Harold, managing director at Humley
The Women in IT Awards is the technology world’s most prominent and influential diversity program. On 22 March 2018, the event will come to the US for the first time, taking place in one of the world’s most prominent business cities: New York. Nominations are now open for the Women in IT USA Awards 2018. Click here to nominate