Published: March 01, 2021 | Comments
The pandemic has pushed many of us to up our TV-watching game, and for me that means revisiting old movies. Recently, my channel-surfing led me to War Games, a 1983 classic starring Ally Sheedy and Matthew Broderick.
In case you are too young to know the plot, and many of you are, the story follows a teenage boy who unknowingly, and through dial up, hacks into a military central computer. The computer asks, “Shall we play a game?” Not realizing the consequences, Matthew’s character, David, thinks it would be funny to choose the game “Thermal Nuclear War”. He quickly learns that the computer is programmed with artificial intelligence to help him learn, but the twist is it is also programmed to win at all cost, which results in the potential of the end of the world.
While not the first movie to take on artificial intelligence, War Games might be the first time I contemplated machine learning, and the concept of a computer making decisions outside of human intervention. As I watched it this time, with my experience as a contact center veteran under my belt, I was reminded that there is a central truth to the movie that still holds true: AI is only as good as the training it is given by a human.
I recently had a conversation with one of our customers, and we talked about their successful launch of a pretty sophisticated chatbot to support their service organization. When I asked him how well the bot did when first launched, he said, “Well, compared to where it is today, it was actually pretty stupid at the beginning.”
Chatbots, voice bots, and all forms of interactive digital interactions are quickly becoming a part of the communication and service strategy across almost all industries. What we have talked about for years at conferences and in webinars is quickly becoming the standard for customer service. In every case when we implement these new tech solutions, there is one constant: bots are only as smart as we train them to be.
In War Games, our heros had to find the supercomputer’s creator to do some last-minute, white knuckle consulting work before the computer started WW III. In real life, when planning a chatbot strategy, you must include plans for follow-up training and redesign. Be sure you build in reporting and analytics that help you understand success-rates for solutions, and also help to understand which questions are not being answered. In our work with customers, we find that no matter how much time you spend in your planning, design, and understanding of what customers are asking, you really cannot know how to train the bot until the data is accessible.
Customers ask questions that you will not expect. They will also ask questions in weird ways - using words and phrases you have not considered. The customer I mentioned above told me that he was surprised by the lack of clarity customers use when “talking” with a bot; when humans realize they are talking to a machine, they will often not utilize the same communication skills they would with a human, and their language gets choppy, and even rude.
The customer agreed with me that there is a deep need for human handlers to “listen” to the interactions to understand how to train the bot better. On the plus side, he also said he was very surprised to see the bot learn so quickly, and become even better than expected with finding solutions and presenting recommendations to the customer in real-time.
In the movie, David and his girlfriend end up saving the world by retraining the computer to help it understand that the only way to win was to not play the game of nuclear war. When it comes to customer service and building an effective chatbot, there is no longer an option to not play the game, but the best way to win the game is to ensure you are building a bot that never stops learning.