The Dark Side of Chatbots

Chatbots have been around for years — and have frustrated consumers for pretty much all of that time — but with the rise of generative AI, these robotic helpers are once more in the limelight.

If you work in the customer experience space, you know there’s a palpable frenzy around integrating generative AI chatbots into customer support workflows.

Gartner predicts that by 2025, 80% of customer service and support organizations will apply AI to their organization. And while chatbots aren’t the only use of AI in customer service, they’re probably the best known and most visible application.

Support teams are keen to reap the benefits of generative AI (like cost savings and improved coverage), but chatbots bring a whole host of challenges. Any team looking to integrate chatbot-type tools into their customer experience need to be aware of their limitations.

Here’s why: a full 50% of consumers said they often feel frustrated by their interactions with chatbots.

Would you even consider launching any other tool, workflow, or initiative that has the potential to irritate half of your customer base?

Generative AI can absolutely help in customer support, but if the only form it takes is a chatbot, then both support teams and customers are missing out. That’s because chatbots have a dark side — one that’s not often talked about. 

The ghost of Clippy past can haunt even today's smartest implementations of AI.

Primary downsides of chatbots in customer support

Customers across all industries express their frustrations for chatbots — a full 40% of chatbot interactions are labeled as negative experiences. 

Yet customer support teams across the globes keep implementing them. 

Whether it’s due to pressure from executive teams to implement AI, the promise of cost savings, or just a desire to use flashy new technologies, chatbots keep on growing. 

But they aren’t working. 

There are at least three major reasons why chatbots continue to frustrate consumers and disappoint customer service leaders:

• Their lack of empathy

• Their inability to handle complex requests

• Their creative limitations

Improvements in generative AI may help in some ways, but there’s still a ceiling to what chatbots can do. 

Chatbots lack empathy

Generative AI has an uncanny ability to apply an emotional character to a response. It can take an input and make it more friendly, professional, or silly. 

Adding warmth to text, however, isn't a stand in for human empathy. It’s just the result of the tremendous progress made in the realm of natural language processing (NLP.)

Despite this progress, a chatbot still cannot genuinely understand a customer's emotional state and cater to it. It can't meet customer's needs by framing options and solutions to suit them.

For instance, even a well-trained chatbot can cheerfully tell a customer whose business is at a standstill because of a critical outage that they're sorry things aren't working like they should and suggest an article about troubleshooting.

Here’s why that hurts your business: 

While a perfect customer journey might involve never needing to contact customer support, the reality is that your support team is often the primary point of interaction a customer will have with your brand. 

This reality highlights something crucial: customer service interactions are a powerful opportunity to build loyalty. They’re a chance for meaningful connections with customers.

And forcing those customers to interact with a chatbot whose ability to connect is limited squanders those opportunities. 

Even if you have the best intentions when implementing a chatbot, a large portion of your customer base likely has a different perspective. One study found that as many as 46% of customers already believe chatbots are being used to stop them from reaching a human. 

Pause and think about that. 

When a customer starts a conversation with a chatbot, chances are they’re already gritting their teeth and bracing themselves for a bad experience. They’re not seeing an easy way to get help. They’re seeing an unnecessary hurdle. 

When combined with AI’s limited ability to convey empathy, overcoming this existing negative perception is incredibly difficult. 

Chatbots struggle (or fail) to handle complex requests

Chatbots have always struggled to handle complex requests. 

Even with improvements in conversational AI, this still remains true. If your products contain technical elements or your requests are layered, a chatbot will struggle to pick up on the nuance and parse out actionable answers. 

That’s because they lack the ability to think critically. They can’t analyze a particular issue from multiple perspectives the way a human can. They’re often bad at knowing when to send an SOS and pull in human help. Even your newest customer support agent is able to understand when something is over their heads and will require help from a team lead — but your chatbot will blindly charge ahead. 

These failures are often glaringly obvious to customers, which is why 75% of consumers agree that chatbots aren’t able to handle complex questions. 

A related challenge is that chatbots are hindered by an inability to multitask effectively. 

Customers often explain multiple problems — or multiple aspects of the same problem — all at the same time. This can overwhelm a chatbot. They’ll often latch on to one aspect of the problem and focus on trying to resolve that piece, even if it’s not the root of the issue or the logical way to approach things.

Compare that to a human being who can easily parse out, prioritize, and address a myriad of issues at once.

Bots often simply add up to more friction between a user expressing their issue and finding a solution. This increases customer effort, shoots frustration through the roof, and hurts key CX metrics like full resolution time. 

Chatbots have creative limitations

Chatbots aren't very creative. 

This isn't a dig at their pottery skills (I'm sure they’d make a great ashtray if they, ya know, had arms…). It’s simply an outcome of their training model. Customer service chatbots simply reflect back already known and documented solutions to customer problems, based on past tickets your team has handled and past solutions you’ve documented. 

What they can't do is generate novel solutions and take logical leaps to apply them to a customer's problem. 

This means they'll often confidently offer incorrect solutions to customers whose problems are outside their defined scope — an understandably frustrating experience. Haven’t you ever had a customer service agent explain a solution you already tried or knew to be incorrect? 

This issue is especially problematic for fast growing companies. 

When you’re in the high growth stage, it’s not uncommon for your documentation to lag behind your product’s evolution. This means new features become unsupportable by your bots, because their training model lacks context on these new features and issues. 

When 86% of customers say they'll leave a brand they trusted after only two poor customer experiences, that means chatbots represent a real risk to your brand reputation and customer retention.

There’s a better solution than chatbots

Generative AI has the potential to have a profound impact on customer experience. It should have a transformative effect on your support team. 

But chatbots are just one way to apply the power of AI to serving customers — and they clearly aren’t that great of a way. 

There are alternatives that provide the same efficiency and economic benefits, without putting your customer experience in jeopardy.

That’s why sometimes we talk about Stylo as being the anti-chatbot. 

Stylo is a Zendesk app that assists and augments your human team. It leads to the same benefits a chatbot can give you—automated resolutions to customer issues, cost savings, and improved scalability—without signaling to your customers that you don’t want to talk to them. 

There are a few key ways Stylo differs from a chatbot:

• Stylo only intervenes when it can actually help. Stylo learns from your Zendesk Guide knowledge base, and it only intervenes when it’s confident it knows the answer. This high-confidence threshold means you aren’t risking bad customer experiences. And when Stylo knows the answer, it automatically generates a unique reply to the customer. 

• Stylo takes a co-pilot approach. Instead of trusting a robot to assist your customers accurately and effectively, Stylo assists your human team. It suggests personalized responses to customer issues, then your human agents can simply accept the suggested responses or rework it as needed.

• Stylo is optimized for accurate responses. Chatbots are a mile wide and an inch deep. They optimize for breadth of coverage, trying to answer every incoming inquiry. Stylo takes an accuracy and quality-first approach. It’s better to answer 30% of incoming tickets with 100% accuracy than 100% of tickets with 30% accuracy. 

Stylo does all of these things directly out of the box, meaning it’s a total plug-and-play experience. As soon as you install Stylo in the Zendesk marketplace, your agents will seamlessly be able to harness the power of AI to generate better answers for customers far more efficiently than they can today. 

Want to give it a try? Try Stylo out live here, or book a meeting to learn more.