By
Jarrod Davis
|
Date Published: June 13, 2024 - Last Updated August 09, 2024
|
Comments
During a recent brainstorming session about the future of AI, the idea came up again of Generative AI making service more human like in the future. Now, being the contrarian that I am, I couldn’t help but challenge that and note there’s a huge assumption behind that statement, namely, that human-like is both good and desired. I don’t know about you, but I’ve had no shortage of awful customer service experiences in person, on the phone and via chat, all with actual humans. So, what do we actually mean with human-like service and should be your goal when implementing CX technology? Let’s discuss!
What is Human-Like
Most people tend to define human-like in terms of what it’s not: i.e. not an IVR, not a terrible bot, not a rigid experience lacking common sense. That helps a little. The reality is, contact centers have spent years teaching humans to act like robots. Follow this script, stick to this specific process. Don’t make exceptions. Don’t talk too long. It’s truly hard to not enjoy the irony of a human centric activity like service being treated like a mass manufacturing project. The question becomes a bit philosophical here as technically those awful script-driven examples are human experiences, but humans who are being forced to act like machines. What does that count as? The worst of both worlds!
Dictionaries describe human-like as something that has or suggests human-like qualities, but boy that includes a lot of potentially bad ones. Yes, on a superficial and technical level, an LLM can generate more “human-like” natural language, but we’re not only talking about that. I’m going to suggest a working definition of the goal generally intended when “human like” is said.
- Communication and comprehension is effortless like speaking to a person.
- Some personalization is used (addressing person by their name, mentioning relevant information about the customer)
- Communication is contextual, not standardized or copy/paste responses.
- Common sense is used vs. rigid processes.
- Empathy is expressed by both sides
At the risk of coming off cranky, there’s not a single point there in which humans cannot and do not regularly fail at, in fact often on purpose due to training. So, let’s all agree: “Human-like” service is actually a set of behaviors within the bounds of an interaction, and even humans don’t always exhibit them.
Service Design & Planning for Failure
Service design plays a critical role in the deployment of AI in customer service. Don’t just blindly accept “human-like” as a project goal. When considering human-like qualities in AI to aim for in your AI project, it's essential to focus on the aspects of human interaction that customers truly value. I’m going to repeat that, it’s about customer value. This includes being personalized and contextual, understanding, and the ability to adapt to unique situations—qualities that can mostly be emulated by AI. You may be asking yourself, “How can AI adapt to a unique situation” because there will always be edge cases or complex issues you cannot model or automate. That’s a given.
You’re not trying to avoid failure though, but rather to plan for it. In doing so, you can ensure that an AI Agent for example, quickly recognizes its limits (after X number of failed intent recognitions for example), gathers as much information as possible and provides a warm handover to the agent. In the end, that’s not much different than when a customer has to be transferred to a senior agent or one who specializes in the issue. It’s the “how” which makes all the difference.
Customers Want Automation and AI (sometimes)
Technology, especially AI and Generative AI, offers the opportunity to transcend previous service limitations. It enables a level of consistency, reliability and personalization that human agents often struggle to achieve. You absolutely can and should partially or fully automate many low-value processes that even humans don’t want humans for (e.g. reset password, activate new card, change address, cancel reservations). Honestly, that’s a given now, it’s no longer up for debate because for example, having to talk to a person and explain the details of changing a reservation or activating a new credit card has more friction and higher customer effort than a few clicks.
I’ll say it again: Don’t treat customer experience as a mass-produced item, it’s not a manufacturing process.
My Conclusion
The goal of implementing AI in customer service isn't to make interactions as human-like as possible, but to reduce friction and customer effort and thereby help both customers and agents. "Human-like" should mean leveraging the best parts of human interaction—empathy, understanding, and adaptability—while employing AI and automation when it’s the best solution to a concrete problem, not for it’s own sake. Remember, if AI is the solution, what is the question?
In the end, whether or not service feels human-like is less important than whether it is effective, efficient, and solves the customer’s issue. By focusing on outcomes rather than methods and feel-good buzzwords, businesses can ensure their use of AI in customer service truly enhances the customer experience, instead of replacing one frustration with another.