Artificial Intelligence (AI) is big news currently. ChatGPT has brought AI into the mainstream news and caught the public’s imagination. Microsoft has been a long-time investor in ChatGPT and is incorporating it into its search engine Bing and enterprise and productivity solutions. Meanwhile, Google reacted by launching Google Bard.
Most people who try these public-facing AI engines react with amazement at their depth and breadth of knowledge and their ability to interact conversationally with users.
Organizations are now scrambling to see how to utilize AI to deliver better customer service at a reduced cost. Expectations are enormous – surely AI will lead to a revolution in customer service as customer-facing bots respond knowledgeably and helpfully to questions and requests?
If organizations carefully and intelligently incorporate AI into their customer service models, we will see this technology’s benefit. But there is also plenty of opportunity for organizations to disappoint their customers through the poorly thought-out deployment of this new technology.
We have been here before – many organizations have already deployed chatbots on their websites and digital channels. And, often, customers hate them! Many people see chatbots as a way for the organization to avoid talking with them and a barrier to engagement. How often have you typed that you want to ‘talk with a person’ – only to be told that the bot ‘didn’t quite understand your question’?
So why will this new generation of AI-powered chatbots be any better? The answer is that they may not be! AI is as good as the underlying knowledge the bot has access to. If you submit a question that has been asked millions of times before, the answer will likely be accurate. If you ask a question specific to you, and the bot has no access to your data, it won’t know the answer.
Privacy considerations may also be a restricting factor in a customer service environment. The flagship public AI platforms are accurate because they access a huge amount of data. And they store the content of every interaction to learn from experience. No organization can allow its private customer service interactions to go into an uncontrolled, accessible public data repository.
And worse than not having an AI-powered chatbot is having one that isn’t helpful or doesn’t provide an escape route to the customer when it runs out of steam. Whether AI-powered or not, any chatbot needs to understand its limitations and offer human intervention when it reaches its limits.
When a chatbot hands a customer over to a human agent, it must also hand over the full context and history of the interaction. That way, the agent can pick up where the bot left off and use the information gathered.
And here is where AI can make a difference – in helping human agents perform their roles more effectively. The concept of AI assisting people in their work is something we see rapid advances in – Microsoft has made a raft of announcements regarding adding what they call ‘Copilot’ functionality alongside their core desktop productivity tools. The same can radically improve customer service – by using AI to analyze the content of a conversation and do background research to prompt the agent regarding what can be done to assist. In this way, the customer still receives the ‘human touch’ while benefiting from the use of AI.
Using AI as an Agent Copilot also benefits the organization – time spent resolving a query and the time required to train new agents can be significantly reduced – as the agent is learning on the job – learning from AI.
Hence, correctly deployed, AI has a very important role in customer service. Incorrectly deployed, it’s going to be a big disappointment. We will surely see examples of both scenarios.
Kevin Ross, Director of Geomant UK
kross@geomant.com
