The Reality of AI in Customer Service: Balancing Risks and Opportunities

Written by Ian Landsman · 04.10.2024

The buzz around artificial intelligence (AI) in customer service is deafening. On one side, there are bold declarations that AI will automate all customer interactions, making human agents obsolete. Conversely, there are warnings that AI creates unacceptable risk by producing erroneous outputs and will diminish the personal customer experience.

The reality, however, lies in a balanced perspective that acknowledges both the potential upsides of strategically applying AI and the limitations and drawbacks of full, mass automation. As conversational AI and large language models (LLMs) continue to advance quickly, customer service leaders must take a harmonic approach, judiciously leveraging AI’s capabilities while simultaneously upholding the human expertise we know to value.

The Benefits: Efficiency, Scalability, and Empowered Agents

It’s difficult to deny the appeal of AI-powered customer service. Using AI chatbots and virtual agents to automate responses to common questions promises to drastically reduce wait times, boost productivity, and scale support operations in a cost-effective manner for businesses with high volumes.

Beyond handling basic Tier 1 queries, AI can potentially enhance your human agents’ performance. There are AI customer service writing tools that can suggest better phrasings for responses or emails, and machine learning models that can surface knowledge base articles relevant to the conversation at hand. Let’s not forget the power of time with regard to learning models. Over time, AI may even be able to extract insights from interaction data to predict issues and personalize service.

While some companies are still waiting to implement AI in their business, others are already realizing these promising benefits. Klarna, the “buy now, pay later” service, claims its AI customer agents now handle 83% of customer conversations automatically. This frees up human personnel to focus on more complex issues. The question still remains: Is full automation really the end goal?

The Risks: Inaccuracy, Complexity, and the Need for Humans

Even with all its potential, unchecked AI automation in customer service could severely backfire. We’ve seen it happen with big companies in the past (hey, Microsoft’s Tay). We’ve also seen more recent legal situations due to the use of AI, like when Home Depot used a wiretapping AI to tap into call center interactions with customers (without consent or notice). The legal landscape around AI isn’t one we’ll discuss today, but it’s crucial to know requirements are rapidly changing around AI implementation.

“Right now, the biggest risk is that language models will confidently give wrong answers that have real consequences.” – Ian Landsman, HelpSpot Founder & CEO

AI language models, while impressively fluent, are trained on broad data. Without specialized tuning, they can regurgitate inaccurate information that damages trust and erodes brand reputation.

When talking with our own founder, Ian, about this intersection of AI and customer service, he shared a recent example of AI gone awry that stood out. Somewhat of a rumored case, an airline’s AI supposedly sent a customer’s relative’s remains to the wrong destination. As Ian put it, “Every business has a different tolerance for AI errors, but the technology is still far from perfect.” In this given scenario, one would say the tolerance level is relatively low. Even here, we can see the potential of AI as exciting, and yet the risks of getting it wrong can be significant.

In our decades of supporting customer service teams with useful software, we’ve found that many customer queries are too complex or context-specific for current AI to fully resolve. Take the example of a maintenance department. While AI may help with the initial onset of a request or easily scheduling a repair, a physical person with the appropriate tools and skills is required to resolve the issue—this is all part of the customer’s experience.

Beyond functional limitations, there’s also the subjective aspect of brand voice and empathetic communication that human agents excel at. Quality customer service agents will continue to outperform AI in heightened emotional environments. An over-reliance on AI risks delivering tone-deaf responses that fail to make emotional connections, resulting in less happy customers.

HelpSpot’s Balanced Approach: Enhancing Humans, Not Replacing Them

Given both the potential upsides and risks, HelpSpot is taking a balanced, measured approach to incorporating AI. We’re not rushing into full automation. We want to carefully evaluate the technology and understand where AI can genuinely enhance customer service, not just fully replace it.

The AI capabilities in HelpSpot’s next-gen platform aim to empower human agents to be more efficient and effective. For example, we might use AI to surface knowledge base articles and draft complete response texts, but agents would review AI suggestions before finalizing any customer-facing answers.

While our team is constantly evaluating where and how AI might enhance our existing software, we’re not losing sight of our primary goal—to elevate customer service management with simple and easy-to-use software, making CS teams more efficient. Ian describes it best, “We see huge value in using AI as a force multiplier to make our customers’ support staff faster and smarter, but we also believe the human touch is irreplaceable for more complex situations that require deeper product expertise, emotional intelligence, and real-world context that AI can’t fully replicate yet.”

The Evolving AI Landscape and Need for Continuous Adaptation

It’s undeniable that the AI landscape is moving at a rapid pace. Since 2017, the number of artificial intelligence startups has more than doubled. Language models are becoming more capable of comprehending lengthy queries and contexts. AI developers are implementing methods to estimate an LLM’s confidence as it generates outputs. Advances in constitutional AI may improve factual accuracy over time.

We’re building a foundation for AI today, but we have to remain flexible to take advantage of AI’s continued evolution. A few years from now, AI may be reliable enough for certain industries like e-commerce to enable secure, fully automated service. Or we may still be enhancing humans with AI assistant capabilities.

Therein lies the need for a balanced, adaptable approach. Different business models and customer bases will have varying thresholds for AI automation risk tolerance versus the drive for cost-efficiencies. Specific customer pain points may always demand empathetic human resolution. That’s why HelpSpot believes in striking the right balance.

Striking the Right Balance: A Hybrid Model for Customer Service

There’s little doubt that AI will fundamentally transform customer service. But it’s overly simplistic to frame it as an “AI vs. human” binary paradigm. The future will involve hybrid models that harmonize the strengths of AI and human agents.

An AI system trained on a company’s data may accurately resolve most issues automatically for certain types of customer queries with limited scope and stakes, like billing inquiries, product availability questions, and returns/refunds.

However, human oversight and judgment will remain crucial to success for uniquely complex cases and situations requiring deeper domain expertise. One approach could be for an AI to collect all the context for a support ticket and suggest potential resolutions based on the knowledge base, then pass it to a human agent to finalize the resolution.

AI could also benefit data analysis and extracting insights from large volumes of customer interaction data. Rather than replacing human analysts, AI could augment them by surfacing hidden patterns and trends that could inform operational improvements, product updates, and predictive service models.

The path forward is an ongoing evaluation of where AI can optimize and enhance specific customer service workflows while still keeping humans in the loop for oversight and quality control. The most successful businesses will likely cultivate hybrid workforces where AI elevates, rather than replaces, human employees’ skills and expertise.

Embracing the AI-Powered Future, Responsibly

Don’t think of it as an “AI takeover”—it’s about embracing responsibility. The future is undoubtedly AI-influenced when it comes to customer service, but that doesn’t mean AI will be solely driving the ship anytime soon. As incredible as tools like language models become, there will always be uncharted territory where human intuition, judgment, and empathy must take the lead.

That’s why HelpSpot is taking a pragmatic, balanced approach to AI adoption and implementation. We want to tap into AI’s superpowers for scalable, efficient service operations. But we’ll never lose sight of the irreplaceable value that human customer service representatives bring to the table. It’s about optimizing for the best of both worlds—the continuous innovation of AI, complemented by the authentic emotional intelligence that only humans can deliver.

If you’re not already using HelpSpot and want to see how two decades of dedication to improving customer support can enhance your team, now’s the time. Request a demo today!

Image of Ian Landsman
Ian Landsman
Ian founded HelpSpot in 2005 with the goal of making every interaction with your customers simple and efficient.
Case study logo or photo

We trust HelpSpot with our most important communication—and they've made a huge difference to our business.

Teagan West, Customer Service Manager

The Silent Partner

Your customer service at scale
Amplify your support with HelpSpot: the streamlined
solution for scaling customer service effortlessly