Most consumers spending time online have come across chatbots and have learned to understand their capability as being rather limited. Most of us have not yet come across a truly impressive chatbot, however that is not to say that chatbots cannot already be built to be quite sophisticated. In reality the difference between a good chatbot experience and a poor one comes down to the long-term thinking of the company looking to deploy it.
To dispel a myth from the get-go, chatbots are not artificial intelligence (AI), just like many other pieces of automation or software not AI despite claims. A chatbot is actually quite a simple piece of software that takes in string of text (the question the consumer is asking) and tries to match it with possible buckets previously inputted question examples. If the bot determines with a high enough confidence (a simple probability calculation) that a question matches a certain bucket, it returns the one answer given to each bucket. From this description one can already see that the sophistication of a chatbot is directly connected with the number of questions the company has inputted into the application for the chatbot as past examples. However this is one of the first pitfalls that companies fall into when implementing chatbots – they do not put in place an ongoing method of educating the chatbot with more question examples, and instead hope that their initial efforts are sufficient.
Although our experiences with chatbots have on the most part not been great, chatbots do have the possibility of enhancing customer experience. Think about the times when you have called customer service of a bank or mobile operator and gotten different answers from different representatives. A chatbot would not make that mistake and therefore can therefore enhance accuracy and compliance in customer service, not to mention the speed of retrieving answers.
So how do you integrate a chatbot into your customer service seamlessly?
- Your organisation must have a long-term horizon and understand that this is an ongoing effort. The chatbots need to be trained over a long period of time based on real questions it is fed from real customers.
- Find a chatbot provider that has built a truly scalable solution. Since chatbots are hot now, several products are available that will do the job when being demoed with ten sample questions. However, the solution must allow for ongoing training so that it can reach thousands of sample questions and still work quickly through the databases they are stored in.
- You need to build a strong fallback option to live customer service when the bot fails to understand a question. This is more of a process mapping challenge as you need to make sure there is enough capacity in customer service to quickly take over chats in which the bot has given up. This process planning may include actually have a queue time for chatbots as well to ensure that live customer service representatives can pick up discussions within seconds; it would be terrible if customers had to wait for ten minutes to move from a bot to a live agent.
- Your KPIs should positively reinforce the trend of migrating from live customer support to chatbots while maintaining (or hopefully improving) customer satisfaction. Depending on the amount of training put into the bot prior to setting it live, your organisation may start from lower than 5% of queries being handled by the bot. However seeing that percentage tick up based on the new questions fed into the bot will be rewarding and will have direct impact on customer service agents being able to focus on more complex queries.
Hopefully over time we will see chatbot experiences get better and more delightful. The technology is certainly there, however as always with new tools the actual challenge lies with organisations being willing to put in the effort to actually implement it properly. Chatbots definitely have the capability of enhancing customer experiences, while also making companies more efficient.