Will chatbots replace your call centre or customer service agents?

Alex is a chatbot. She was formally introduced to the world in March 2016 and has now communicated with more than 1.2 million people. Her creator, the Australian Tax Office, says she gets it right (first contact resolution) 81% of the time.

Alex understands over 37,000 variations of intent, which are the different ways the client questions are asked. There are over 500 questions that these intents map to and these questions and variations are growing every day.

The ATO says Alex, which is based on software from Nuance Communications, exceeds industry benchmarks of 60-65% of first contact resolution. Impressive, right?

Well, I’m not sure where this statistic comes from and the ATO wouldn’t say. How reasonable is it?

In my view that’s hard to say considering the immaturity of chatbots and also what constitutes a successful interaction – I’ve had many experiences in testing chatbots (including Alex) and I’d be hesitant to say the communications were successful 50% of the time. Yet, the more you use conversational user interfaces – whether in written text or spoken – the better they get.

Regardless, Alex’s apparent success and ongoing use is something that many organisations are pursuing. And it raises a serious question.

Can you use chatbots that leverage machine learning and natural language processing to become your primary point of contact for customers? Or even further, can they replace your call centre or customer service agents entirely?

The argument for using chatbots is, on paper at least, compelling. The cost per conversation compared to a call centre is, in whatever currency you calculate it, better, especially for larger organisations with huge call volumes.

Deployment and decommissioning is faster and arguably easier than using humans. They can be always available on whatever channel the customer wants to communicate. And there is less chance they’ll go off script – meaning they will represent your brand in the way you tell them too, every time.

The argument against only using chatbots is also, however, strong. In all the research we have done across Asia Pacific, it is clear only a small majority of consumers want to only use digital channels to interact with organisations. We still highly value traditional forms of customer service.

Further, the current form of chatbots and the technologies backing them are still developing. There are many times when chatbots and other conversational user interfaces simply can’t understand the language or pronunciation being used. Or they just don’t get the question being asked of them.

It’s important to remember that humans aren’t average and the diversity in how we communicate with one another is immense. The bots are improving rapidly, but they aren’t nearly close to human yet.

At this juncture in time (the middle of 2017) I wouldn’t be recommending anyone aim to replace all of their call centre or customer service agents with chatbots. That’s not to say chatbots shouldn’t be used. They absolutely should be on your technology roadmap if not already being deployed. Just make sure you have adequate human support and hand off at the ready.