Generative AI, meet the contact center - the latest silver bullet to kill off the traditional operating model?

There are plenty of conversations about using generative AI in marketing and sales, but what about customer experience? Yes, the chatbot experience exists, but that's just one way to leverage generative AI. I spoke with twol experienced tech leaders to better understand the opportunities in areas such as customer support, the contact center, and digital customer success.

Mladen Milanovic is the VP of Automation at Presidio, a global digital services and solution provider. Milanovic has extensive experience working with contact centers and customer experience. When Presidio decided to build an automation practice, they decided the biggest synergies and impact were in the contact center. Milanovic says automation significantly impacted contact center efficiencies, including improving call center agent satisfaction and productivity. Most importantly, it increased customer satisfaction.

The contact center was an early adopter of AI, such as Natural Language Processing (NLP) and Machine Learning (ML). That AI tech is pretty much commoditized in this industry, argues Milanovic. The technology is proven, and there are many best practices and strong ROI. So, it's natural to start looking at what's next.

One area Milanovic sees interest in is agent-assisted technology. For example, sentiment analysis systems can provide feedback to an agent on how to handle a situation. Now, that's still NLP, but Milanovic believes that generative AI and Large Language Models (LLMs) will accelerate the adoption of this type of technology and provide even better results.

According to Milanovic, the contact center often provides the only interaction brands have with customers. And that means the interaction is critical, and the information is high quality and accurate. This is why Milanovic believes generative AI will be adopted in internal systems like agent-assist software, more so than in virtual agents who interact directly with customers - at least until the challenges with inaccuracy and hallucinations are addressed.

Milanovic argues that there is no way to effectively control, monitor, and manage AI agents in real time. There are ways to fine-tune and adjust, but the tools are technical and developer-oriented:

So what that means is in the contact center, you need to have an ecosystem or a framework that will allow a business user to easily track performance, and measure the performance of the virtual agent in real-time based on historical data, fine-tune responses and potentially jump in and augment or modify and adjust and retrain the agent faster. This is, today, a very technical job that requires a very scarce talent from the application development perspective.

He adds that new low-code/no-code tools are coming, but they are young and unproven, and it's unlikely a contact center executive will put something on the front end they can't control or measure. They will, however, experiment on the back end with tools for agents.

Generative AI requires a solid handle on company data

There are other ways that customer experience can leverage generative AI, including chatbots, and language translation, suggests Michael Ringman, CIO of TELUS International. TELUS International is a spin-off from TELUS focused on serving customers across several business lines, including many customer experience capabilities.

Ringman says that a lot of customers are asking how they can take advantage of new tech like generative AI to move them forward. With a large customer base, the conversations are always interesting, Ringman says and the answers depend on where the customer is in their business portfolio:

I think what really matters is the data, right? Getting access to data, understanding the knowledge of your data, and leveraging that to its fullest is really the sort of point of the spear to get going on Gen AI and leveraging it to better capabilities across the organization. You know, you can't just put Chat GPT as your new chatbot on your internal or external support site and expect that to take away all of your call volume. It's just not going to work like that.

Companies shouldn't focus on building LLMs but instead need to get a better handle on their data. Ringman argues that there is an untapped well of voice of the customer data. His company uses tools like Google CCI, Bard, and other extensions to do speech-to-text to harvest and mine the data.

He adds that many customer support teams and bots (the choose your own adventure kind) work with stale knowledge bases. LLMs and gen AI can have a more free-flowing conversation, but if the data sets aren't well-tuned, there's a greater chance of hallucinations:

That's usually where we are starting with those customers is let us help you get into some of this untapped data because the data that you're tapping today, especially around customer experience, is typically around average handle time and typical traditional call center sort of statistics and not really getting to the heart of the matter which is meeting where your customers where they want to be met and understanding what they're looking at.

That’s not necessarily anything new just because gen AI is out there, he points out:

It's the same challenge that call centers have had for quite a while. The interesting part is that gen AI gives us so many more tools in our tool chest to go and tackle some of those more complex problems and really start to offer customers personalized services. Really get in and start to give them a more human-like feel when they're interacting with some of those bots and really provide a differentiated experience."

There will always be humans in the loop

Ringman and Milanovic both agree that there will always be a need for human involvement in customer experience.

When Presidio advises clients on using generative AI, the firm recommends giving agents additional training. Milanovic says that it's important that agents feel okay with disagreeing with an AI recommendation instead of entirely relying on it - there have been incidents where complete reliance on AI has led to bad outcomes.

Ringman notes that IVR was deemed to be the end of the 'traditional' contact center, then chatbots were going to take over. He suggests that while the best intents are there, we still need people:

Just because we now have these large language models so that the interaction with the bot can be more real, more life-like. Just because they've got access and can start to understand more datasets doesn't necessarily mean they're always going to be able to solve all of those challenges."

As for the need for responsible AI, Milanovic states:

Everybody is talking about that right now, but without an overall framework leading the organization on how those tools that are in a very specific stage of development should be used. I think that we need to really create and help our clients create those frameworks and ensure that we do have checks and balances the around that.

Meanwhile Ringman posits:

It's all about building those data sets and understanding that reducing bias across your data sets, reducing all of the corruption that could potentially lead into those datasets. So again, this is where I would say the human-in-the-loop aspect of that is going to continue to be around. The jobs are going to change, but they're still going to want to be people in the loop in regards to that.

Milanovic also argues that as long as human labor stays cheaper (most contact centers employ agents overseas where the cost of labor is low) and the costs of leveraging LLMs and generative AI remain high, adoption will be low:

As technology becomes more accessible and more commoditized, we're going to see an increase in adoption, right? But right now, AI is a hot topic together with security in boardrooms, and everybody is just reading [article] titles. Nobody reads the article, in general, to understand the limitations.

My take

The contact center is still a key component of customer service and experience, and there will be a continued focus on improving how agents support customers. Milanovic and Ringman agreed that the first part of leveraging generative AI will be to support agents, but it will also shift to the front-end customer experience as confidence grows. Chatbots, or virtual agents, are improving (Conversica shows us how this can happen).

But I agree with Ringman that the data should come first. We never seem to stop talking about eliminating content silos within the organization. And until we do, organizations won't understand customers as well as they could, and they won't be able to leverage any technology to create the best experience, let alone AI.

Next, we'll look at a couple of additional ways generative AI can improve customer experience, particularly by supporting back-end workflows and processes.