A long time ago, in a galaxy far, far away. It was the time of self -service and the emergence of new entities called ‘chatbots’. Many were trying to get hold of the new technology and a few more trying to understand where it came from.
Apart from its intergalactic battles, light sabers and force-wielding Jedi – the Star Wars universe gave us two impressive bots – R2D2 and C3PO. They were an integral part of the saga and proved resourceful in helping their human masters save the day. Indeed, the concept of, ‘bots with a dash of artificial intelligence’ has been fascinating for many years. While by simple definition, a bot (or robot) is any machine that performs tasks replacing human effort, it is the advent of artificial intelligence that is kicking up a storm as we find ourselves at the cusp of redefining machine history. It is essential to understand that artificial intelligence is a vast concept that will impact various sectors in different ways but it has definitely arrived in the customer service arena with natural language understanding (NLU) in the form of chatbots.
Certainly, chatbots have raked up a storm of interest but amidst that fan-fare, we see many losing out on the critical element of bot design and how to use it for bringing out the best in their business.
In one of my earlier posts, I had written about virtual assistants (a.k.a. bots) for contact centers, now called customer service chatbots. It’s also important to know what we should expect from these assistants and what would be the best design protocol of a good bot.
- It’s complicated? It is important that the customer and bot relationship remains simple. This basically means not offering too much at a time for the customer to experiment with. And as a result, not confusing the bot with an array of human behavioral attack. It would be best to start with simple interactions, keeping the bot language crisp ‘n clear and a clear defined logic. Gradually, as better adoption sense is gauged, one can move onto adding more complex features and service offers.
- Assign it a job: Bots are here to serve a purpose for your business. They should cater to one segment/bunch of queries /services most frequently asked by one’s customers. It follows the rule of thumb of self-service – let the frequently asked questions be automated. While, it’s fun to ask a banking bot, “What’s the weather today?”, it may actually not serve any business purpose. It is always a good idea to have a focused bot that wins a customer’s trust than attempt to set up a magician that may end up becoming a joke! After all, the intent is to create customer stickiness and offer a valuable self-service medium.
- Everyone loves predictions: I cannot stress how significant it can be to have predictive capabilities in self-service. A bot that could maintain my previous interactions and throw me suggestions or proactively address my queries would surely be a winner.
- Do not leave me stranded please: Indeed, a bot may not have an answer to all your questions. It is important to not leave your customer stranded here with abysmal error messages. I personally feel, as a Step 1, the bot should ‘dumb’ things down and prompt some cues if the customer is lost. And if nothing works, the customer should be able to reach out to an agent and have the context of the interaction kept intact. This way a customer experience is never abrupt. It is a common practice for IVRs, why not have it for bots as well?
- Natural language Understanding: NLU does act as a powerful tool for a good bot. It helps capture customer intent and reply accordingly. The success of a bot would depend on how appropriate responses have been given to the customer and not the level of customer traffic at your bot. And this is where NLU plays a critical role.
- What’s in a name? Shakespeare would be disappointed, but, I feel a bot a should have a catchy name to give it a more humane feel (apart from the NLU of course). It would ultimately represent the face of one’s business and a robust call flow along with a name can be the recipe for a trusted advisor.
- Analytics: Full-fledged business intelligence may not be a must-have for day 1 but analytics is key in making sure the bot actually works, and looking closely at the bot’s performance. Logging what users write and how the bots responded, and how users respond to the answers shortly after go-live should be mandatory, to avoid a bot’s failure.
At the end of the day, the motive of a bot should be to make a user’s life easy and benefit a business in the best possible way. With the right technology combined with the right design, it should not be difficult to achieve the same.
Latest posts by Ayesha Borker (see all)
- So, what is the future of work? - July 17, 2017
- Conversational systems are the future and there is no running away from this dialogue - January 4, 2017
- Using Voice Biometrics to Secure your IVR and Shorten the Call - December 1, 2016