Take on any CX challenge with Pipeline+ Subscribe today.

Getting Past the AI Hype

Getting Past the AI Hype

Getting Past the AI Hype

How AI can realistically improve the CX.

Aside from artificial intelligence (AI) being the most important technology trend for contact centers to get behind, it’s also evolving at a pace that few organizations can keep up with. The benefits are undeniably attractive, but the complex nature of AI offerings makes it difficult for buyers to know what they’re really getting.

Contact center leaders face difficult priorities, as they will only fall further behind if they hold off on AI. But once starting their AI journey, they will be dealing with a lot of unknowns. To make those priorities easier to manage, this article focuses more on the reality than the hype, along with some guidance for evaluating AI offerings from the vendors.

AI Evolution for Improving CX

Before considering what’s real with AI, contact center leaders need to be current on two forms of evolution that are driving most of today’s innovation around improving customer experience (CX).

First is the emergence of Conversational AI (CAI). Chatbots have long had a bad reputation for self-service, and early versions offered little improvement over IVR.

With CAI, today’s chatbots are more conversational and human-like, making them more acceptable for customers, and offering a far better CX than IVR. A key reason is the ability of these chatbots to understand intent and context, which enables them to handle a wider variety of inquiries, thereby taking more workload away from live agents.

The second major evolution is Generative AI (GAI), where rather than responding to queries from customers, these applications generate their own responses, either via text or voice. This takes AI to an entirely new level for customer engagement, where bots can provide long- or short-form replies without supervision from humans.

GAI certainly has a long way to go to be on par with human responses, but it’s constantly improving, and the CX use cases will only keep growing.

Another important factor to consider for both CAI and GAI would be the way AI learns from its mistakes. As current missteps are rectified, they won’t occur in the future. And in time, these applications will earn enough trust to be used in more complex scenarios.

The key here is to deploy AI initially for operational use cases rather than customer-facing use cases. This way you can mitigate the risks during the learning curve, and in time, you’ll be comfortable enough to start using them with customers.

At that point, it’s best to start with automating simple inquiries like password changes or account balances. And then progress to more involved customer interactions that agents will be happy to have chatbots take on.

  • “AI technology is definitely ready for consumption, however, CX leaders need to start small and grow big. Don’t jump into a six-figure project immediately. Begin with a proof of concept and understand their data readiness requirements and learning model. Determine if you have enough volume and agents to justify scaling the AI technology. Very large organizations have incorporated AI into their customer experience strategy successfully because they’ve had the economies of scale working on their behalf.”
  • —John Heiberger, Contact Center Consultant and President, Lucilium

Knowing the Hype from Reality

These forms of evolution are very promising – hence the hype – but we’re still early days, and CX leaders need to proceed with caution, not full speed ahead. In particular, here are two factors to consider, especially in terms of how vendors are touting their AI capabilities.

1. Accuracy is everything with AI

The outputs of both CAI and GAI are only as good as the inputs they are working from. When tapping data from customer conversations for building a knowledge base, speech-to-text accuracy is critical, and a common metric is word error rate (WER).

Another form of accuracy – especially for GAI – is being factually correct. If the inputs from customer interactions are ambiguous, erroneous, or incomplete, you will very likely have similar outputs. In AI parlance, “hallucinations” are often the result, where what the bot is telling the customer is inaccurate, or even seems made up.

It’s unrealistic to expect 100% accuracy, so you need clarity from vendors for how they make things as accurate as possible, and what metrics they use to demonstrate this. Perhaps just as important, will be to determine their capabilities for handing off to live agents when bots are getting into trouble with accuracy.

This last one can be difficult to gauge, but it makes all the difference for CX when you can do that handoff just before the bot has reached its limit for providing accurate responses. When the handoff is too early, the agent is taking on more work than necessary, and if it’s too late, the risk increases for chatbots to ruin CX and possibly do harm to the brand.

2. Protecting your customer data

Data is the oxygen that powers AI, and since the contact center is a metrics-driven operation, it’s ideally suited for adopting AI. That said, customer data must be handled with care, especially when it comes to training the large language models (LLMs) that most AI applications are based upon.

There are two aspects to consider here when vendors tout their AI capabilities.

First would be how effectively they can harness your existing datasets of customer interactions to produce personalized responses or messages. The objective here is to automate customer service in ways that approximate how agents interact with customers. The better your data inputs, the better the vendor’s bots will perform.

If your existing customer datasets are limited – which is often the case – then you need to assess how vendors can help you build that up so bots can do a credible job with your customers.

The key here is to recognize that a vendor’s AI offerings are only as good as the datasets you have on your customers.

While AI-based technologies still have a long way to go to be on par with live agents, the biggest determinant of their effectiveness for contact centers will be the quality and usability of the data you’ve been collecting for all these years. As such, whatever level of hype you feel surrounds AI, this should not dissuade you from the capabilities you will gain when deployed effectively.

...you need clarity from vendors for how they make things as accurate as possible...

There’s a second consideration that has more to do with protecting customer data rather than harnessing it. When AI vendors build their LLMs, they’re largely drawn from publicly available sources. This may be the easiest, fastest, and cheapest way to develop AI applications.

But when vendors are all drawing from the same sources, there will be a sameness to the outputs when interacting with customers. That makes it difficult for contact centers to use AI to differentiate their CX and provide highly personalized interactions.

This is why your customer data is so valuable. And when vendors use this data to build your LLMs, you need to ensure this proprietary dataset is not being used by the vendor to train other bots: either for general use applications, or potentially for other customers who could even be your competitors.

In short, protecting your customer data needs to be a prime consideration when evaluating vendors, especially in terms of how they will be using this data to build your applications, and only for your applications.

Again, in AI parlance, transparency is the key concept you should hold vendors to account on.

Questions about technology aside, it will be just as important to understand their policies and principles around developing these applications. These typically fall under the banner of what’s called responsible AI. This term refers to a framework that businesses – as well as vendors – are adopting to articulate how they will approach AI in terms of legal compliance, ethics, and privacy.

The Reality for Your AI Journey

Getting past the hype is one thing, but it’s also important to see the reality as you get on this path with AI. The reality around making good buying decisions is to focus on use cases rather than trying to get a granular understanding for how AI works. AI will continue evolving at a rapid pace, and most contact centers lack the technical expertise to stay on the leading edge.

Better instead to build your business case on use cases that have clearly defined benefits and outcomes. For contact centers, there are three core use cases for AI; self-service automation, agent experience, and customer experience.

For each of these, there is a multitude of specific applications, such as intelligent FAQs for self-service, automated call summaries for agents, and personalized responses and offers for customers.

Legacy technologies can support these use cases to some degree, but with AI, not only can these things be done in new ways, but also at scale, for every customer and every call. That’s the real power of AI, and the ROI will follow when outcomes are measurable, such as shorter handle time, reduced agent turnover, and higher percentage of calls handled by bots.

Conclusion

In the course of my work as an independent analyst, there is definitely reality beyond the hype when considering the success stories from leading players like NICE, Cisco, Verint, Five9, Cognigy, RingCentral, and Talkdesk.

These vendors – and others – can readily cite specific use cases with metrics that quantify the impact of AI, both in terms of cost savings and operational efficiencies that help contact centers manage with today’s higher volumes.

As with any new technology, the hype will always be there. But once you get past that to these realities, the path for your AI journey becomes much smoother.

Jon Arnold

Jon Arnold

Jon Arnold is Principal of J Arnold & Associates, an independent analyst practice providing thought leadership about the business-level impact of digital transformation on the future of work. Core areas of expertise include unified communications, collaboration, cloud platforms (UCaaS, CCaaS and CPaaS), contact centers, and customer experience.

Contact author

x

CURRENT ISSUE: January 2025

Moving Forward: What Will 2025 Bring For Contact Centers?

View Digital Issue

SUBSCRIBE SUBSCRIBE

Most Read

Artificial Intelligence

Is CX Improving? Or Declining?

Artificial Intelligence

Looking Beyond NPS and CSAT

Customer Effort Index
Upland 20231115
Cloud Racers
Amazon Connect 20240826
RLZD Gartner Motivate Report
Verint CX Automation
Trends Forrester Budget Planning Guide