I must admit: chatbots have been a tough hype to fight.
At the F8 conference of Facebook in April 2016 Mark Zuckerberg announced tools for developers to build bots on the Facebook Messenger platform. This was the real start of the chatbot hype.
Between the summer of 2016 and the autumn of 2017 literally every customer and prospect I talked to asked if we would address their customer service challenge with a chatbot. I would always reply: “No, since we can create more value for your customers than a chatbot can.”. And what followed after the “No” mostly fell on deaf ears.
This was how organisations looked at using AI to improve customer service. Then gradually the discussion changed, not the least with those organizations that had already gained some bot experience.
AI can best create value through augmentation and automation
As from the start of NoisyChannels we believed that AI can best create value through augmentation and automation. For AI to completely replace human customer service, at least one crucial condition needs to be met: a whole lot of data and a small number of variations. Putting your hands on large amounts of conversation data is unfortunately not the case in most business settings and certainly not in customer service, where chatbots are predicted to bring disruption.
In most customer service organizations with European companies, the amounts of useful customer interaction data are not expressed in millions but rather tens or hundreds of thousands. Even the best AI talents in the world will not be able to train algorithms with these datasets in order to completely erase the human customer service agent for a large variety of situations.
Most chatbots today can be very useful as either disguised FAQ’s or as a new interface for easy transaction presented as a conversation
Organizations that have pioneered with chatbots have by now come to the same conclusion: most chatbots today can be very useful as either disguised FAQ’s for a limited number of topics in a very narrow domain (“I have lost my credit card. Now what?”) or provide a new interface for transactions that are presented as a question and answer conversation with some options to select (“I would like to order a pizza quattro formaggi”). Moreover, while pioneering, these companies have learned that it takes a lot of work to get one chatbot up and running that is worthy of “talking” to their customers instead of their trained customer service experts.
Indeed, the worst that can happen is when the bot starts replying “Sorry, I didn’t quite get that.” or asking to rephrase the question hoping to understand what the kind customer exactly is asking.
So, for now, chatbots can handle a few customer requests. But this doesn’t mean all the rest should be handled by humans alone. Only focusing on complete replacement of people is missing out on a huge potential of human-machine cooperation.
The process from human-to-augmentation-to-automation is the only realistic answer to the disillusions generated by the disruptive human-to-automation state changes
The journey for the years to come is to let AI help people in providing better service. With smart support these human experts will not only provide better and faster service. They will also create better data. Organisations will gradually move from 100% human, through an AI-assisted human-led service (i.e. augmentation), towards new domains where bots, supported by more and improved data, start taking over the easy, repetitive and boring tasks of customer-oriented humans (i.e. automation). This rolling process from human-to-augmentation-to-automation is the only realistic answer to the disillusions generated by the disruptive human-to-automation state changes.
That is when real customer service will become more human.