Credit by: Niraj Patel
It’s all about the implementation. Retailers and consumers are increasingly open to AI and machine learning. But you have to get it right, which is essential because there are so many ways to get it wrong.
DMI’s research on consumer behaviors backs this up. Our recent special report (What Buyers Want — a DMI Consumer Survey) found that one-third of respondents had encountered AI, while the remainder either had not encountered it or could not say for sure. Of those who were aware of AI, just over one-half were either satisfied or very satisfied with the experience.
So, a significant minority of consumers know they’re dealing with AI and a small majority are fine with it. But they don’t prefer dealing with an AI chatbot over talking to a human. We posed eight customer service scenarios to survey participants and asked whether they’d prefer dealing with a human or an AI chatbot.
Even in the most readily automated scenario, ensuring a product is in stock in the customer’s size before purchase, survey respondents preferred human assistance over a chatbot by a factor of 4-to-1. In the most complex scenario, where the customer was mischarged and needed a refund, the human-to-AI preference leapfrogged to 19-to-1.
5 AI Implementation Tips
If you’re getting ready to ramp up AI in your retail customer service operations, keep these tips in mind:
Optimize the agent side before the customer side. AI chatbots can make life easier for customer service agents, speeding customer and product information to them so they can handle more service requests. This strategy lets you work the kinks out of your AI engines behind the scenes, where they are less likely to annoy consumers.
Seek the path of least resistance. Humans don’t always resist AI in chatbots. For example, people can be more comfortable talking to a bot about personal issues (health, finance, etc.) they’d rather not reveal to a person. In a more common scenario, you can automate simple queries like basic product information that a bot can deliver more efficiently than a human. As our survey revealed, resistance to AI happens on a continuum. You’re much better off embracing the low-resistance end of that range.
Start small. It’s tempting to take on ambitious AI projects that transform large, complex processes. That’s usually a bad idea because you’re dealing with so many variables. AI is a learning process that optimizes positive outcomes and discourages negative ones. You start out solving easy problems and let your AI engines evolve to handle increasingly complex challenges
.Optimize for satisfaction. Frustrations rise when an AI chatbot cannot handle the customer’s request. You have to build your AI engine to recognize when the bot is in over its digital head and needs to hand things off to a human agent. Resistance usually happens because somebody botched this implementation.
Deploy trust mechanisms. Fingerprints and facial recognition are mechanisms to establish user trust. If you’re creating an automated interface that has to pull in sensitive customer information like a credit card number or a Social Security number, you need to add these kinds of mechanisms to reassure people that their data is secure.
Keeping it Human While Encouraging Automation
To a large degree, AI in customer service is a process of adoption rising as people become more acclimated to automation. Think about when gas stations started allowing people to pay for gas at the pump. Initially, we all asked for receipts because it felt necessary. Then we started seeing receipts pile up in our cars and realized we didn’t necessarily need one every time.
In years to come, many more components of AI will become automated and only the most complex issues will require human intervention. The human part of AI will be more about overseeing, measuring and monitoring increasingly complex AI systems.
We’ll always need humans in the loop of customer service because commerce is about people getting what they want. A bot can’t understand why that matters