How contact center AI helps businesses and customers

Clients shopping for contact center as a service (CCaaS) products have a keen interest in how providers are integrating artificial intelligence (AI) into their solutions. Without a doubt, AI puts big performance gains on the table – for example, fine-tuning large language models (LLMs) with industry data improves self-service prospects for consumers and helps agents too. But understanding how contact center AI helps businesses and customers is a much bigger story.

In fact, one of the best views on how contact center AI helps businesses and customers comes from focusing not on the technology directly, but on the key challenges that it solves. Consumer demands are higher than they have ever been. There’s pressure for service channels to be available 24/7 and offer personalized features, which creates much more work for brands.

What’s more, companies that lack the right tools to keep up with those expectations can soon find themselves overwhelmed. And in the worst cases all of that incoming unstructured information sits in siloed systems, which employees have to respond to and deal with. “While positive for consumers, the proliferation of channels is a challenge for agents to manage and it affects the ability of brands to forecast their business,” Darren Rushworth, a member of NICE’s global leadership team, told TechHQ.

The good news is that AI can help. And NICE is one of a number of CCaaS providers that are implementing solutions to address not just one, but many potential bottlenecks that have emerged as contact centers have evolved beyond voice to deliver customer service on multiple channels. Call volumes and durations are relatively straightforward for businesses to forecast, and operators need to have confidence in those numbers to hit metrics such as customer satisfaction (CSAT) and cost-to-serve.

Joining up data and finding patterns

However, once you start adding other channels into the mix and try to account for blended interactions, getting those resource predictions right becomes much harder. That data needs to be joined up, which is another challenge when contact centers find themselves relying on more and more discrete systems added over time – a scenario that Rushworth refers to as the ‘Frankenstack’ problem.

Also, an agent’s work isn’t over when a call finishes. There’s admin to perform and details to update, which can involve multiple backend applications. Helping service providers today are solutions such as robotic process automation (RPA) software, which can quickly streamline repetitive tasks. Early chatbots were supposed to make things easier too, by solving common customer queries. But first-generation FAQ bots lacked the power of modern AI. Often they fell short of the mark and could end up leaving customers frustrated.

Today, however, it’s becoming a very different story. NICE’s services enable more than a billion transactions per quarter. And huge, aggregated datasets (anonymized and stripped of sensitive information for security) turn out to be highly effective food for data-hungry modern AI, epitomized by LLMs such as OpenAI’s GPT-4. Large, domain-specific datasets allow CCaaS providers to offer advanced chatbots that are much more capable of dealing with customer queries.

Also, guardrails can be included so that these ChatGPT-like solutions know their limitations and can hand over to human agents when the chatbots recognize that they lack the information to answer queries they’ve never come across before. And those knowledge gaps can be filled in by learning from the new data that follows.

Making more accurate projections

Tracking back to the forecasting issues that have built up as contact centers support more and more channels, AI can contribute there too. The ability of algorithms to find patterns in vast amounts of data can be utilized to create much more accurate projections of how many resources will be required and at what times. And it highlights how contact center AI helps businesses and customers on multiple fronts.

Rushworth explains how NICE has used AI to identify 16 key behaviors that point to strong CSAT scores and positive interactions between agents and customers. And, in his view, this has led to a much fairer assessment of contact center agent performance. All transactions can be analyzed, rather than just relying on a small sample of interactions where a manager may be listening in. And the AI-powered system can be used to offer real-time guidance to agents, bring up knowledge, provide a playbook, and automatically schedule training. “Employee satisfaction has gone up,” Rushworth adds.

Flexible working tool

Employee engagement management is another area where AI is benefiting staff by giving team members the ability to work to more flexible schedules. Solutions can automate forecasting and incorporate shift bidding to account for when agents are available. It also gives contact center operators a way of reaching out to team members. For example, if a never seen before pattern triggers an unexpected rise in customer enquiries and the contact center needs more agents to cope with the spike in traffic.

Joined-up big data can be used to educate search engines, as another way of smoothing the workload faced by operators. “81% of the time, people still go to Google with their queries,” Rushworth points out. “It’s another self-service digital channel.” There are many ways in which CCaaS providers are extending the contact center AI landscape and making solutions even more capable to help businesses and customers.