An agent-centric view of customer service remains the gold standard for the customer service experience. We even hear about super-agents that see performance boosts with advanced, contextual AI. Human understanding and human empathy enhanced with advanced technology is a great vision, or is it?
We are twenty years into a CCaaS revolution built on a grand promise of better and more efficient customer service, yet last May, the American Customer Satisfaction Index reported customer satisfaction was at an all-time low. The Qualtrics XM Institute reported that from 2021 to 2022, industry average NPS scores declined across almost all industries, with an average decrease of four points.
Is anyone shocked? “Customer service” has become an oxymoron. Few people, if any, expect a great experience when forced to call a toll-free number, and they won’t call until they’ve run out of options.
During the holidays, my bank became concerned about potential fraud and asked me to call them (there was no fraud). It took three days to reach someone because my bank’s humans enjoyed the holiday weekend. When I did finally reach an agent, she wasn’t able to authenticate my identity. So, instead of addressing the fraud concern she transferred me to the wrong department, which required me to restart the entire process.
Could a bot have done this better? Yes! The fraud verification really should have been a simple text message requesting a confirmation. A bot would not have required time off for the holidays. The bot could have authenticated me easily with a PIN code or voice biometrics. And, even if all that failed, it seems unlikely the bot would have transferred me to the wrong department.
Around that same time, Frontier Airlines made headlines that it was discontinuing live agents for customer service. There was a collective scoff from the contact center industry that this was doomed to fail, and it might. But the more I process it, the more they might be on to something.
The challenge with digital and self-service customer service solutions are they can be used to decrease or improve service — either way they cut costs. The key for improvement is to adopt the right technology. I fear the airline is more interested in costs, but I contend that the right technology, including the latest AI, could be an upgrade from a live customer service line.
AI has been getting much better, and what we are seeing from ChatGPT-3 is groundbreaking. A bot that seems to understand human language as well as its predecessors have promised. My attitudes toward bots are changing, and this is humbling for me. I’ve been known to chant “operator” or “representative” repeatedly to self-service bots in the past. We’ve all been there, a place and time when IVRs were nothing more than a nuisance and delay — one that turns customer CX jaunts into customer journeys.
Now, the bots have a leg up over the human agents. Recent breakthroughs enable bots to truly understand customer intent. They can authenticate callers with technologies such as voice biometrics instead of insecure trivia like a mother’s maiden name. Bots don’t take off time for the holidays, speak multtiple languages, and unlike call centers don’t require much in forecasting because they can scale with demand.
Automation by contact center vendors have focused on enabling the agent to be more productive. That certainly was a reasonable step. We've all been been frustrated when agents don’t have access to the right screen or information. But perhaps our contact center efforts should shift to bot-first, and only supplement with humans by exception.
Consider my recent agentless experience with Farmers Insurance for roadside assistance. Their bot sent me a link, and then guided me through the service request form as I entered the information from my smartphone. It was easier than correcting, repeating, and spelling all the information verbally. The bot then dispatched a service vehicle, and the bot let me track the service vehicle's progress on a real-time map. I had a delightful service experience, and it’s hard to believe a human agent could have improved it.
The contact center industry keeps beating the empathy drum. I didn’t need or want empathy, I needed a jump-start. Farmers invested in the right technology and now it can sit back and watch its bots serve its customers — with reduced operating expense to boot.
The goal is quicker resolutions, not to make agents faster and better. There’s a technical element, but the bigger barrier is acceptance and attitudes. There’s the stubborn belief that customers want to talk to an agent above all else, and also an aversion to bots. Both need to (and will) change.
We need to shift away from bad AI with a human backstop to AI designed for all use cases. Designed for all, there will be failures, but they should be exceptions. We need to change our conviction that customers want to talk to human agents and instead focus on solving issues. I have changed my attitude, and I’m all in favor of my bank’s fraud department replacing its human agents.
As far as our attitudes and biases against bots, consider a recent story about a mental-health nonprofit under fire for using bots for counseling. The bot’s messages were composed by AI (and supervised by humans). The responses were rated higher than those written by humans, and more responsive. But, when people learned that the messages were written with AI, they felt disturbed by the “simulated empathy.” Their biases were stronger than their needs.
Attitudes toward bots are changing. Bots can solve problems quickly, easily, and securely. We have to re-learn what bots can do as we have already been taught how useless they can be. While I did everything I could to avoid a bot before, I find myself increasingly intrigued by the bot option, and even occasionally annoyed when I get a human.
I’m not kidding. I called to activate a credit card and got a human. This is a job routinely done by bots, so I was annoyed that the card provider was doing it manually. This experience proved to me that customers will adapt to favor bots, and contact centers should too.
Humans are prone to failure and require breaks. High turnover requires constant training. Hold times are a reality of human-shifts in contact centers.
The transition to excellent CX via bots will happen. ChatGPT has provided a glimpse into the future of AI. Effective customer service requires excellent conversational AI, very specific transactional skills, and tight integration with back-end systems as well the contact center itself. It also needs to know when it can’t help, and the appropriate actions to take when this occurs. These are not trivial to accomplish.
For example, the simple task of authentication may require a 360-degree view of the customer involving access to historical, CRM, ecommerce, and marketing data, or perhaps biometric samples, and 2FA will likely require use of multiple channels.
The joke goes that for the past 20 years we have been about 10 years away from great AI. I’m revising that estimate to five years away — at least in terms of mainstream adoption. I am seeing some encouraging progress from companies like NICE, Five9, and Cognigy.
"Bot by design" is my new creed, and for the revolution to happen we need bot frameworks that have predictive and LLM technologies. They need to be easy to train and customize with both proprietary and standard data. The CCaaS priority needs to be centered around managing customers and bots, not agents.
We are at an amazing time in CX evolution. First, came the cloud and a plethora of new channels, now comes the AI revolution.
Dave Michels is a contributing editor and analyst at TalkingPointz.