Though it doesn’t always get the credit it deserves, the customer service industry is a crucial part of the way technologies actually change the world.
Long after funding has been secured and code has been written, contact center agents will be answering questions about resetting passwords and using obscure features, making sure that promising new products really are changing users’ lives for the better.
Like every other industry, customer service has to adapt to changing conditions, and that’s the subject of this piece. I’ll draw upon my years of experience in technology and entrepreneurship to talk about some important emerging trends in customer service. Before drilling down into the one technology that’s most on everyone’s mind these days: artificial intelligence (AI).
Where Customer Service is Heading
The hockey great Wayne Gretzky once said, “I skate to where the puck is going, not to where it has been,” a quote that was popularized by one of the most innovative titans of the 20th century, Steve Jobs.
The point is that it’s not possible to win if you’re always playing catch-up. The best way to succeed, especially in a field as fast-moving as contact centers, is to read the proverbial tea leaves so as to anticipate what customers will want in the next three, five, and ten years.
With that in mind, let’s start by looking at where the puck of customer service seems to be heading. For context, I make no claim that this list is comprehensive; instead, I will focus more on the trends that would be impacted by AI as this is my own particular area of expertise.
The Rise of Self-Service
We tend to think of customer service as a very hands-on endeavor, with agents working through problems step-by-step over the phone or via text. This is certainly true, but in the grand scheme of things, what’s most important is ensuring that our customers feel empowered to use a tool or product with confidence.
Agents can be the best way to do this, but the forward-thinking contact center will invest in ways to give customers what they need to solve their problems on their own. This means having a robust ecosystem of helpful content, which could be videos on YouTube or TikTok, FAQs, knowledge base articles, webinars, and the like.
...providing them with tools instead of fixing their problems for them, that's only going to increase their loyalty to our businesses...
This is not an area contact centers can afford to neglect. According to Zendesk’s survey data, 69% of customers want a way to resolve their issues without needing an agent to intervene, and 63% of customers begin by looking at the resources the company has made available.
On the face of it, this may seem counterintuitive. After all, aren’t we making ourselves less relevant by adopting this approach?
But our job is to give our customers what they want.
If that means providing them with tools instead of fixing their problems for them, that’s only going to increase their loyalty to our businesses and brand(s).
Automation
Delegating tasks to machines is nothing new and has been driving progress in a variety of industries for at least the past few centuries.
In the context of customer service, this can manifest in a few different ways. If you’ve written a bunch of articles and trained a large language model (LLM) on them, for example, a customer can just ask their question directly and often get a useful reply.
There’s much more to say about this, but I’ll leave it for the bigger section on AI to follow.
Personalization
Today’s customers are increasingly on the hunt for service that is tailored to their specific needs, i.e., personalized customer experience (CX).
If more and more customers are taking it upon themselves to try to solve their own problems – and the data suggest they do, as we’ve seen – then it’s only the more complicated issues that’ll rise to the level of a customer service agent.
In such situations, no one is going to want to feel like an agent is reading a script to them or treating them like they’re just another ticket that needs to be marked “Done.”
Personalized CX begins with training agents. There’s an art to making a frustrated, distressed person feel like they’re being heard, and it’s something that good contact center agents (and managers) know how to cultivate.
It’s also something that AI can help with. Through techniques like retrieval augmented generation, LLMs are already getting better at “grounding” their replies.
Put another way, LLMs are becoming more and more able to utilize context when they generate their outputs. Some of that context could be information about a customer, about their past interactions with your product or customer service team, etc.
Messaging and Social Media
Finally, an article like this would only be complete with a discussion of the way social media has changed customer service.
In pursuit of the ever-elusive competitive advantage, brands are increasingly turning to Twitter, Facebook, Instagram, TikTok, and other digital networking services to find their customers.
It’s no surprise, therefore, that more brands are interacting directly with their clients through these platforms.
Pulling back a bit, all of this is part of a broader move towards what is sometimes known as “omnichannel support,” meaning that customers can reach agents through the phone, email, messaging services like WhatsApp, and messaging an agent directly inside an application.
And as you may imagine, this is also a prime spot to integrate AI. Chatbots, in particular, are extremely well-suited to help answer basic questions, provide context and resources, or route a customer to an agent when the situation requires it.
There are services that are able to train LLMs specifically on your contact center’s past messages, dramatically increasing the speed with which your agents can formulate replies.
AI for Contact Centers: Implications, Strategies, and Best Practices
Now that we’ve covered some of the major emerging trends in contact centers and customer service, let’s turn to a more direct conversation about AI.
What is Artificial Intelligence?
There’s a longstanding debate about the nature of intelligence and how it arises out of certain kinds of complex, adaptive systems. We will simply define AI as using machines to do tasks that once would’ve required human thought. This won’t satisfy the philosophers, but it’s good enough for our purposes.
Obvious examples in the modern world would be anthropomorphic robots, self-driving cars, automated theorem provers, and LLMs like ChatGPT.
None of these are general intelligences in the sense that they can operate in open-ended ways like you or I can, but each one can do one or many tasks that were once solely within the purview of human beings.
AI in Customer Service
Given how much customer service relies on text, it’s a natural fit for AI. In this section, we’ll talk about a few ways to use AI to improve CX as well.
Sentiment analysis, which refers to machine-based techniques for identifying whether a piece of text is neutral, positive, or negative, is a good place to start.
Using sentiment analysis, it’s possible to figure out how people feel about a product or categorize incoming tickets based on their tone. This can help identify crucial areas of customer dissatisfaction that should be addressed in future product releases and prioritize those issues most in need of fixing.
Real-time machine translation of audio (and now video) is also worth mentioning. This is actually one of the original uses of machine learning, and after decades of progress and research, it is remarkably good.
If you’re fortunate enough to have built a global brand, you’ll need a way of making sure you can speak to your customer base in words they understand. Machine translation services make it possible to do this without hiring an army of translators.
Throughout this piece, I’ve made reference to ways in which LLMs can help supercharge agent productivity, and for the sake of completeness, I’ll reiterate some of that now.
Out of the box, LLMs are quite good at summarizing text, answering questions, and helping diagnose issues. These capabilities become dramatically better when the model in question has been finetuned to specifically operate on your documentation.
Finally, I’ll peer into the future just a bit to talk about some of the incredible work being done on building AI agents.
Today, LLMs like ChatGPT are already capable of using simple tools by i.e., creating API calls, but they still rely on humans to actually use the code.
But efforts like Auto-GPT, AssistGPT, and SuperAGI are actively attempting to remove this barrier. Though the technical specifics of the projects vary considerably, what they all boil down to is trying to give LLMs more of an ability to operate autonomously.
They have ways of taking high-level tasks (“book me an appointment”) and breaking them down into individual steps, generating the necessary output to achieve each step, and interacting with tools like the internet whenever required.
Closing Thoughts
Given how human-centric our work tends to be, injecting machines into dozens of parts of the pipeline may seem to make little sense. But this is not how I view the situation.
Rather, I see AI handling progressively more and more of the tedious, routine parts of the job, freeing up human agents to spend more time personalizing their responses, resolving the thorniest problems, and attending to overall customer satisfaction.
It’s still too early to say for certain how AI will ultimately go. The LLMs we’re using a decade from now might be at human-level performance across the board, or they might be 10% better than what’s available today.
Regardless, I anticipate that AI will have a large and growing impact on the future of our industry. As customer service and contact center professionals, we have to keep an eye on these developments, and prepare ourselves to meet the puck where it’s going.