Imagine if the World Cup was decided based on a tally of the steps every player took, ignoring goals actually scored. Well, with conversational AI, we’re kind of measuring in strides today.
AI has long tried to make the user feel that they are talking to a human. This is what drove the creation of Eugene Goostman, the computer that convinced judges that it was a 13-year-old Ukrainian boy and garnered international acclaim a few years ago. Or why we saw AlphaGo beat the GO champion of the world. These examples of AI are astonishing, no doubt, but have little relevance to the everyday AI agents that engage with and offer support to customers on a daily basis.
Being indistinguishable from a human has no influence on the overall success of a retailer’s AI so why have we become engrossed with creating human-like AI? In fact, what customers want most is fast and accurate resolution to their problems in a friendly, respectful tone. In a recent survey we conducted, in fact, not one respondent even mentioned anything related to an AI being “human-like,” yet we’re still seeing time delays, spelling errors, and humor come up throughout conversations to try and personify the AI.
The Authoritative AI
Like self-driving cars, what we need to focus on is making conversational AIs autonomous. AIs need to work alongside human agents to solve customer problems quicker and more efficiently on the customer’s terms. Here’s where the AI focus should be:
- Answering repeatable customer questions: AIs should integrate with order management systems to be able to provide order status and updates, answer policy questions and other FAQs.
- Having the authority to act: AIs should have the autonomy to do things like accept returns or issue a refund based on specific circumstances (i.e. accepting a return under a certain amount or within a specific timeframe) and the individual (i.e. loyalty status).
- Providing proactive support: By integrating with inventory, logistics and other systems, the AI can do things like alert a customer proactively when an item they are interested in is restocked, there is a delay in shipping or to recommend specific product care or usage tips.
Leveraging AI frees up human agents to focus on more complex issues. An AI can empower agents in these scenarios to work faster by recommending replies and actions to the agent, surfacing meaningful information, or gathering information from the user before an agent gets involved.
AIs Need to Earn Authority
Making AIs autonomous sounds great, of course, but trusting an AI to act in your company’s best interest and resolve customer experiences accurately takes time. AIs need to earn this trust over time. Let it “listen in” to cross-channel conversations and report how it would have responded, with no actual contact with the customer. This “supervised” learning will see human agents confirming or denying the AI’s suggestion and will boost the AI’s confidence in certain scenarios while expanding its ability to respond in others based on how human agents act.
Once the AI is “in the wild,” you always need to give customers the opportunity to connect with an agent or easily communicate an issue was not resolved. These conversations need to be elevated quickly to the appropriate channels.
WIth a focus on autonomous AIs, authority-based metrics should be the KPIs that are the focus during AI planning and design. Things like resolutions vs. escalation to human agents; time of issue resolution and intent classification scoring should be the focus over the traditional Turing Test believability factors.
Let’s shift the ideal AI from human-like to one that can resolve, not just respond.
Msg.ai specializes in applying artificial intelligence to customer service.Favorite