Your customers want and need to be heard and to have our insights gathered, analyzed, and if merited, acted on to enable success for your organization.
But these are uncertain times, riddled with angst. We as customers are both demanding and nervous. Nothing personal; it is today’s zeitgeist.
At the same time, the technology underpinning customer engagement and customer experience (CX) is increasingly being built with AI (artificial intelligence) on a digital, networked virtual foundation.
Can we harness AI effectively to provide organizations with deep customer insights without missing our humanity?
To provide insights into how best to listen and understand customers like us and utilize that knowledge, we recently had a virtual conversation with Dave Singer, who is Global Vice President, Go-to-Market Strategy, Verint.
Q. What are contact centers hearing from customers in the way of issues but also in the tone and urgency? Are these changing, and why?
The types of issues that contact centers are facing hasn’t really changed. Think billing questions, delivery delays, challenges with the website, etc.
However, customer expectations around the speed to resolve issues have significantly increased. Customers today want excellent service and quality products, and they want them fast. Tomorrow feels like a long time to have something physically shipped to your home.
It’s the same with resolving customer issues. And everything and anything is increasingly an urgent matter.
In addition, there is a growing divergence between on-demand customer service in today’s digital age versus tolerance for delays in the “real” world.
It’s realistic now that a consumer can be shopping in-person, realize a product is out of stock and then find it online and have it delivered before they return home. Today, people expect the physical world to move at the same fast pace the digital world offers. This trend is creating less tolerance for delays.
Q. What channels and means are customers choosing to express their comments and concerns to organizations, are these changing, and why?
People are getting increasingly used to digital channels to engage with a brand, and at the same time, these digital channels are becoming increasingly available to them.
But it’s still the case today that a certain level of importance and emotion will inform a customer to speak to a live agent or escalate the interaction from digital to voice.
“...the breaking point when customers switch between a digital channel and voice is increasing.” —Dave Singer
For example, if your rent payment is processed in error and you’re at risk of being evicted, you will likely want to speak to a live agent until it’s resolved because the issue is of high importance and impacts your livelihood. With the high costs of renting and low availability of suitable places to live these days, your angst would understandably be off the charts.
What’s changed is the threshold for what’s considered highly important. Today, you can buy a car entirely online without ever speaking to a live person, even considering that this purchase is relatively high cost and therefore high importance.
Which engagement channels customers choose to express their comments and concerns to organizations comes down to three types: Immediate digital (web chat app), deferred digital (web/email form), and voice (calling into customer service).
Today, there’s less differentiation between these channels, but the breaking point when customers switch between a digital channel and voice is increasing.
Before, channel hopping was more difficult because you had to move between voice and mobile devices, but today, mobile devices handle every modality, making it easy and seamless.
Q. Conversely, what channels and means are contact centers using to hear, understand, and act on customers’ concerns?
There are many channels and means for companies to collect, analyze, and act on the voice of the customer (VoC) to improve and enhance CX, including feedback channels and customer surveys, to name a few.
It used to be that a customer was quick to use social media to get their issues resolved quickly. Or companies “forced” channel hopping. For instance, when you receive the text “We’ve noticed fraud on your credit card, please call this number,” forcing the customer to switch from text to voice.
Consumer-channel behaviors are evolving, and they should be able to engage on whatever channel (voice, text, email, etc.) at whatever time they want. Brands shouldn’t force their customers into one channel or another, it should be the customer’s choice.
If you get a text message alerting you of suspected credit card fraud, you should be able to authenticate and approve within the same channel. Channel hopping is decreasing as mobile apps are increasingly becoming an action channel versus just a communication channel.
Smart Social Listening
Listening to social must be done carefully as there are bad actors that are using the channel to shill for companies, tear down others, or game them for special treatment.
But is it actually possible to gain meaningful accurate insights from social? And can artificial intelligence (AI) help?
Here’s Dave Singer’s response:
“Thankfully, there are generally more good actors than bad actors, but toxic tactics still exist and can damage a brand’s reputation,” says Dave.
“This is what makes AI so important. For example, there is a new phenomenon called ‘review bombing,’ in which a large number of people (or a few people with multiple accounts) purposely post negative reviews, resulting in biased ratings against a company.
“This is when AI-powered analytics can come to the rescue and identify these types of events or behaviors by examining content and looking for patterns, like bursts in negative comments phrased exactly the same way.
“Similarly, with fraud detection, robust AI-powered analytics and capabilities that are designed to provide rich metadata and risk signals can help companies quickly and easily identify and filter out fraud callers versus safe, legitimate callers. Thereby reducing costs and call handling time in its contact centers.
“Overall, getting good insights is possible, but it takes some work. You can’t take it all at face value.”
Q. Discuss the insights into agent performance garnered from listening to customers for quality assurance (QA). What types of insights can be gained, is this changing, what channels and means are best to obtain them, and why?
All customer engagement channels are equally valid, but the interactions are different and therefore the customer behavior must be measured differently as well.
Customers often share feedback differently when speaking versus emailing customer service. For the customer service agent, the tools and protocols are all the same too; automated quality and post-interaction feedback, for instance, all work for the same end goal: to improve quality and CX.
Yet, to be effective, different standards must be applied depending on the engagement channel to ensure quality.
For example, there are two kinds of insights you can look at. These are: (1), agents following the performance standards “script” and (2), agents listening to the customers and tweaking the scripts slightly to be more customer-friendly in the moment.
In the first scenario, Agent A scores 100% by internal performance standards for following policy and procedure perfectly. Whereas in the second scenario, Agent B scores 92% on quality because they changed the script yet is more aligned with customers’ expectations. By looking at both, organizations can guide overall agent performance better to be more aligned with customer expectations.
CX professionals used to believe that people would talk about the same thing in the same way regardless of the channel. This is no longer true.
Today, consumers express themselves differently and will have different types of conversations based on channel, so you need to measure accordingly. You want to know if your customers call the contact center for specific issues versus using a chatbot for others.
The concept that companies are more likely to get high-emotion interactions on voice channels and more urgent requests on digital chat or messaging channels must be applied to measuring quality assurance too.
Q. There is anecdotal evidence of customers becoming more agitated and stressed, taking it out on agents over and above the actual substance of their comments or service or support issues, and yelling at them at a volume that isn’t warranted. Is this the case, why it is occurring, what are the consequences, and how can contact centers sort the proverbial “wheat from the chaff”?
Today’s expectation for immediate customer service has lowered the threshold to be disappointed.
This, compounded by the COVID-19 pandemic and political and economic uncertainty, means that the general public is rightfully living in a heightened emotional state. Unfortunately, the contact center agent is an easy outlet for escalated rage on occasion.
AI-powered analytics can help companies isolate the “noise” for better complaint management and agent confidence to provide a resolution.
Using deep learning models, AI can look through large volumes of engagement data and apply analytics to find patterns when spikes in similar complaints occur. Like (“I can’t pay my bill online”) versus one-off customer issues (“My package is lost”) and help provide suggestions accordingly.
That way, agents have more confidence in their road to resolution, whether solving for spiky or one-off problem behavior.
AI and analytics shine when used to identify the issue, determine how customers’ responses differ from the average behaviors and further suggest meaningful resolutions.
There has been much talk about “survey fatigue”, that customers are fed up with them and not responding, which risks distorting the real picture of customer attitudes/voice of customer.
So, we asked Dave Singer whether this is happening.
“Survey fatigue is real: consumers are constantly bombarded by surveys, and they take too long to complete,” says Dave.
“Phone surveys are also notorious for providing biased responses because a limited pool of people will tolerate holding on the line to take a five-minute survey, which can drive skewed survey results.
“Today, the best practice is to shorten and randomize feedback questions.
“If you ask one customer a short set of questions like ‘How was your experience?’ or ‘How was the resolution?’ from one section of your survey and another customer another short questionnaire from a different section, you can average the results across all respondents. And get a more accurate picture of customer attitudes.
“And it will reduce survey fatigue.
“Another option is to ask your customers just one straightforward question at checkout on the pin pad and randomize it.
“You will get a fuller picture across thousands of respondents. Your customers won’t have to spend longer than five seconds to say, ‘Yes, the cashier was friendly,’ ‘Yes, I found what I wanted,’ or ‘Yes, the price is reasonable.’
“Reducing the amount of time and questions decreases consumer frustrations and increases survey participation so brands can collect meaningful voice of the customer insights for continuous customer experience improvement.”
Q. Advanced AI, including large language models (LLMs) holds promises and pitfalls in customer service. Can AI help with understanding customers, including gaining meaningful quality insight on agent performance? And what are the caveats?
Yes, AI can help in these areas. The two biggest challenges are a need for more relevant training on AI-powered solutions and a lack of prompt engineering and ongoing monitoring to ensure your AI is delivering correct, timely, and relevant responses. The caveat is that your AI must be trained on current data relevant to your customer engagements to deliver intelligent analysis.
One challenge with using generic ChatGPT to identify contact center problems is that it’s trained on historic internet data and is unable to answer questions that require more current, or specialized, training data.
In addition, generic ChatGPT is not training on your unique customer engagement data. Training on the right data sets is critical. This also requires ongoing testing, monitoring and tuning to prevent AI hallucination that misleads AI to answer questions it can’t.
Q. In listening to customers there are often the seeds of excellent improvement and new product suggestions; loyal customers often want the companies they do business with to succeed. How can contact centers best pick these out and refer them to other departments for action, and in acknowledging customers?
The common theme is that you need to listen to all interactions across all channels. In addition, you must intersect analyses of interaction data with demographic and customer data.
Using AI and analytics to listen to and understand customers’ requests across every interaction is critical. It can signal common themes and marry them up against enterprise data to understand the demographic it is coming from so you can make accurate data-driven decisions.
For instance, if you see a trend of suggestions to move billing policies from A to B, but most of the responses are coming from customers delinquent on their payments, you might not take that recommendation.
Or you could try to understand the challenges that are causing these customers to become delinquent on payments. What if this request comes from your top customers and should be prioritized? Either way, you need to understand where recommendations are coming from and why.
Q. What are your recommendations for contact centers seeking to improve their understanding of their customers and bolster their agent performance from quality monitoring and assurance?
Listening to both the customer and the agent sides of every interaction is best. For example, what kinds of agent performance practices led to the best customer outcomes? How does your agents’ performance match up against defined policies and procedures? You can use this to refine policies and procedures to better serve your customers.
“...you need to listen to all interactions across all channels...you must intersect analyses of interaction data with demographic and customer data.”
It’s about understanding behaviors on both sides by analyzing the impacts of agent behaviors against customer outcomes, and then drilling down to determine which behaviors are driven by your training and policies or driven by the customers’ behavior. All of which can be used to evolve training and coaching, as well as policies and procedures that align to better customer outcomes.
When you coach and train your agents against policies, procedures, and protocols, it will be easier for agents to accept and adopt because they’re being asked to do what the customers want them to do anyway. When the agents are tuned into customer needs in this way, then more positive customer experiences can result.
If you would like to contact Dave, you can reach him on LinkedIn.