AI in customer support: A practical guide to understanding and implementing AI

Artificial intelligence (AI) used to be a conversation topic reserved for the early adopters and the most tech-savvy among us. But since the launch of the AI chatbot ChatGPT in November of 2022, AI has rapidly transformed from a niche interest to a household name. 

These days, it feels like you can’t go anywhere without hearing something about AI. Airlines, classrooms, and even fast food chains have all been experimenting with different ways to use the technology for peak impact and efficiency.

All of those headlines fuel a pervasive sense of urgency that often comes along with these advancing technologies, and it’s easy to feel like you need to hop on the AI bandwagon immediately or risk getting left behind.

But when it comes to using AI for customer support, it’s worth a pause. Your customer support team has the most direct and frequent communication with your customers. That means any change you make to your support team’s processes — whether good or bad — will have a direct impact on your customers’ perceptions of your business as a whole. 

Needless to say, using AI on your customer support team is worthy of some thorough understanding and consideration before you sign up for the latest and greatest tool that promises the world. This guide will help you dig deeper into AI and understand just how, where, and why it fits into customer support.  

What exactly is AI? 

AI, machine learning, automation, deep learning — the tech industry is no stranger to buzzwords. And particularly for those who lack knowledge of their inner workings, it’s easy to use them all interchangeably. 

So, let’s start with a quick rundown of what exactly artificial intelligence is. IBM puts it best when it describes AI as machines “that mimic human intelligence and human cognitive functions like problem-solving and learning.” Put simply, AI is essentially a computer that’s able to function like a human brain. 

AI is pretty broad, and all of the other terms we mentioned are actually subsets of AI. Essentially, think of them as different ways to use or implement AI. Here’s the gist:

Machine learning 

This is the ability of computers to learn without being explicitly programmed to do so. With machine learning, the computer teaches itself to analyze information, pull out patterns, and then make predictions or decisions.

This term is the one that’s most easily confused with AI, as most of the recent AI advances have hinged on some type of machine learning. 

Example: Netflix recommends shows it thinks you’ll like based on what you previously watched. 

Deep learning

Deep learning is a more complex or involved subset of machine learning. The easiest way to understand what makes deep learning different from machine learning is by considering the type of data each is able to work with.

Machine learning algorithms need structured and labeled data. Basically, the algorithms need neatly packaged data to work with — or, at the very least, the algorithm needs to organize data into that format itself.

In contrast, deep learning doesn’t need that same data pre-processing. Deep learning algorithms can take in a big ol’ mess of unstructured data and make sense of it. 

Example: Researchers for self-driving cars are using deep learning to develop detection and recognition of traffic lights. 

‍Automation 

Along with AI, contact center automation is the one you’ve probably heard the most about. You could save time by automating this or automating that. 

Automation is the process of setting up software to perform certain tasks without repeated human, manual effort (or at least, with as little human effort as possible).

While some automation can use AI to get the job done, the intent of the two is actually pretty different. With automation, you’re providing explicit rules and instructions for the software to do what you need it to do. But with AI, you’re trying to get the computer to make decisions and take action without your explicit input and direction. 

‍Exaxmple:You create an automation trigger so that all of your email attachments are automatically saved to your Google Drive.

‍

The growing hype around AI

For quite a while, AI was talked about either with wistful longing for a far-off advancement or with a sense of fear or impending doom (“The robots are coming to steal our jobs!”). 

But as things often do with technology, they moved quickly. AI is no longer a remote possibility — it’s here and happening now. Consider this:

• 35% of businesses worldwide used AI in 2022. While that might still seem like a modest number, it’s a four percentage point increase from 2021.
  
  
• Another 42% of businesses say that they’re exploring AI for future use.
  
  
• ChatGPT set a record for the fastest-growing user base, reaching 100 million monthly active users only two months after its launch.
  
  
• Online searches for terms related to AI have skyrocketed year-on-year. 

It’s safe to assume that AI isn’t a passing fad or a flash in the pan — its star is rising. And it leaves a lot of businesses with a big, pressing question: What the heck are we supposed to do with it? 

Beyond the big promises: Where does AI fit into customer support? 

“Easily scale your customer support!” “Simplify work for your team!” “Be totally hands-off with your forecasting!” “Make your support team AI-powered!”

There’s no shortage of lofty claims about AI’s capabilities in the context of customer support — most of which are focused on enabling support teams to be more streamlined and strategic while simultaneously being saddled with less stress.

That means AI is gaining steam, particularly as leaders and support teams continue to grapple with the challenges of the past few years. Some estimates state that a whopping 95% of customer interactions will be through channels supported by some form of AI as soon as the year 2025. 

But the ambitious promises about AI and its benefits also mean it’s been positioned as the panacea for any and all support team problems. It doesn’t matter what your team is struggling with — roll out an AI tool and all of your worries and roadblocks will vanish.

That rush to implement AI is what most typically leads to disappointment in its capabilities and results. Leaders quickly realize that staff burnout, inaccurate forecasts, or other operational inefficiencies are complex, nuanced issues that require more thought and effort than simply throwing AI at the problem.

In many cases, AI actually ends up being an extra tool or layer of complexity that organizations need to roll back until they can resolve the deeper dilemmas and identify how to best use AI to optimize their systems and processes. In short, it takes time to understand how AI can be a help and not a hindrance.

Yet, support leaders can’t get so hyper-focused on their own team’s efficiencies that they forget about what’s at the heart of the customer support team: the customer.

Support leaders need to balance the team experience with the customer experience, and many customers are understandably still skeptical about AI’s ability to meet their expectations. Consider these statistics from a survey conducted by contact center provider, UJET, and reported by Forbes:

• While 78% of customers have interacted with a chatbot in the past 12 months, 80% said using chatbots increased their frustration level
  
  
• 78% of customers needed to connect with a human anyway when an automated service channel failed to resolve their issue
  
  
• 63% said that their interaction with a chatbot didn’t get them a resolution
  
  
• 72% felt that using the chatbot was as waste of time

While support leaders might be enthusiastic about the role of AI in support interactions, customers are a little more wary.

How should customer support teams be thinking about AI?

Does this mean AI is all hype and no substance — or that customers want you to kick it entirely to the curb? Absolutely not.

In fact, the majority of customers are optimistic about AI’s potential impact on the customer experience. It just requires some careful balance, with 77% of customers saying they believe that positive customer experiences do still need some element of human touch. 

AI absolutely does have a role in the future of customer support teams — likely a large one. But much like with other business changes or initiatives, organizations are best served by tapping the brakes and figuring out how to implement AI on their support teams in the most meaningful ways. 

But what does that actually look like? 

Most companies that are building AI-powered tools in the customer support space would have you believe it’s all about cutting costs and driving efficiencies. And after several years of global economic uncertainty, why wouldn’t it be?

That’s why you’ve probably noticed so many chatbots stepping in for what used to be human-to-human interactions — providing basic answers and pointing you toward the answer you might be looking for in the help center.

To be clear, this kind of deflection can be a really useful part of your AI strategy because it will free up your agents to focus on more complex customer queries. But remember, chatbots can also be a huge source of frustration for customers who aren’t getting what they came for.

An even bigger problem with this approach to leveraging AI is that it fails to recognize what’s at the heart of customer support, which has been and always should be creating amazing experiences for your customers.

Assembled user Gershwin Exeter, Vice President of Global Services at Thrasio, put it best when he said, “Why deflect when you can engage? We want to interact more with our customers. We want them to feel our appreciation and dedication toward them. Blending AI with humanity is the balancing (because it is a constant effort) we strive to achieve to ensure our customers know how much we value them and their brand loyalty.”

This is where we believe AI can really change the game for support teams — by driving more quality interactions at a fraction of the cost.

“AI companies are missing the big picture,” said Kaytlin Louton, GTM for New Products at Assembled. “AI can and should be leveraged to answer customer questions live (deflection) when that drives the best possible customer experience.” 

“However, there are some types of questions that should be handled by a human,” she continued. “High-volume, high-CSAT questions are an example of the type of question that should not be deflected. You know these interactions drive delight for the customer, so keep those questions around. For these scenarios, when support teams want to talk to customers, Assembled's building an AI product to supercharge support teams. Our product drives quality and speed of agent replies.”

Curious to learn more about how we’re thinking about AI in customer support here at Assembled? Kaytlin Louton’s inbox is open to you. Send her an email with whatever questions you have

So…what does it actually look like to use AI effectively?

Moving beyond the philosophical and into the practical, we think the use of AI by support teams should check two equally important boxes. AI should:

• Make your support agents’ lives easier
• Make your customers’ lives easier

That might sound oversimplified, but it really is the recipe for AI that supports both your team and your customers. Let’s take a closer look at each of these, as well as some potential ways to apply AI for both.

AI to assist your agents

Interestingly, one limited study of AI on a customer support team found that not all support agents benefited equally from AI. “It turns out that the company's more experienced, highly skilled customer support agents saw little or no benefit from using it,” explains the NPR interview about the study. “It was mainly the less experienced, lower-skilled customer service reps who saw big gains in their job performance.”

There’s still a benefit there — less-experienced agents achieved full proficiency a lot faster. But it does highlight the point that AI tools shouldn’t support one segment of your team and feel punitive or disruptive to another. Even if your high-performers don’t explicitly benefit from the addition of the tool, it should be a non-intrusive and welcomed addition to their work processes. 

With that in mind, support teams are implementing AI to assist their agents in a number of ways that streamline their work without dampening the customer experience. Here are a few ways AI is being used:

• Automatically prompting support agents with the information or context they need so they don’t have to search for it during a support interaction.
  
  
• Completely automating repetitive tasks or light-lift interactions so support agents can dedicate their time and expertise to more complex interactions. For example, offering chatbots for customers who need to do something simple like update their payment information or ask about your business hours.
  

Routing calls and contacts to the agent who’s best equipped to handle the inquiry (this is actually one of the most popular uses of AI on support teams). For example, one agent might specialize in technical troubleshooting while another is a whiz with billing questions. AI could automatically direct those contacts to the best agent, which saves your team and the customer time.

AI to assist your customers

It’s probably somewhat easy to think of ways AI could benefit your team. It’s when you introduce customers to the equation that things feel complicated — particularly with many support leaders fearing that the use of AI will make customer interactions far more impersonal and ineffective. 

Rest assured that AI can offer a boost to your customers, provided you use it to supplement or enhance your human support interactions and not totally replace them. Here are a few ideas:

• Similarly to the above, equip agents with the context they need — the customer’s intent, question, purchase history, contact history, and other pertinent details — so the agent can deliver a personalized experience. An impressive 71% of consumers expect companies to deliver personalized interactions. Another 33% of customers admit they’re most frustrated by having to repeat themselves to multiple support reps. AI can save them the hassle. 

• Use AI and machine learning to proactively deliver resources and other support content. For example, automatically delivering a discount code to a customer who is perusing your products or proactively delivering some support links to a customer who lands on your “contact us” page. That sort of perceived mind-reading is important, with 62% of customers admitting that they think a company should be able to anticipate their needs effectively. 

• Create self-serve support options like AI chatbots and knowledge bases with AI nudges that guide customers to the right information. This is important for customers who want quick answers without a live interaction. 81% of consumers expect more self-service options so they can find answers and information on their own when they want to.

AI is a tool, not a magic bullet

Gone are the days when AI was dinner party fodder for the tech pioneers and trendsetters. Today, chatter about AI is almost everywhere. And it’s not just the conversations about AI that are increasing — the actual implementation of it is too. 

Much like with anything else, some of the hype is well-deserved and some of it is overstated. Regardless, the relentless buzz instills a sense of urgency in support team leaders — they need to scramble to roll it out and keep pace or risk being left in the dust.

However, that rushed approach to AI is usually what leads to disappointment or confusion (and sometimes even a hefty amount of rework). Support leaders are better served by taking the time to pinpoint the ways AI can best empower their agents and their customers.

It’s time to start thinking about AI as one of many vehicles for better customer support — rather than the final destination.