AI in the Contact Center: What Can Go Right and What Can Go Wrong

AI is seen by many in the general public and media as being used to oust humans from their employment. A survey of over 200 US businesses for ContactBabel’s “Inner Circle Guide to AI, Chatbots & Machine Learning” finds that opinions on whether AI will replace agents are evenly divided.

Respondents from large contact centers were more likely to feel that AI would replace human agents, while those in small and medium operations doubt that this will be the case. It is worth noting that the belief that AI will replace agents has strengthened over the past few years.

Virtual unanimity was found when the question was asked as to whether AI would support human agents, with almost all respondents agreeing or strongly agreeing that this would be the case. It’s our view that for the foreseeable future, AI in the contact center will be used far more to assist and augment agents than to replace them.

How AI is Used in the Contact Center

AI in the contact center is best known for powering chatbots. While this is a useful and growing part of the customer contact mix, AI also can directly assist frontline contact center agents to do their jobs better. Here are some ways AI might be able to help:

Using real-time analytics on calls to improve the quality of interaction, agent behaviors, and customer outcomes

AI can provide suggestions about the next best action, pull up relevant information from the knowledge base, make suggestions based on customer history and sentiment about optimal cross-selling and upselling opportunities, and even suggest the style of conversation that this customer may prefer

Updating relevant systems and initiating the correct business processes

AI-enabled robotic process automation (RPA) can look after some of the work that an agent has to do within a call, and also afterwards in post-call wrap up. The bot can create the call summary from the relevant data gathered and email this to the customer, disseminate any relevant data to various databases, and kick off required processes. Within the call, the attended bots can be triggered by the agent’s desktop activities, rather than requiring them to activate the bot manually. This can include populating fields, compliance reminders, and cross-selling opportunities, making things quicker for the agent and customer.

Using sentiment analysis to support outcomes

This analysis can quantify customer and agent emotions within interactions, whether on the phone or through an alternate channel, for the purpose of uncovering processes, behaviors, and situations which cause strong levels of positive or negative sentiment that could affect business outcomes and customer experience. Sentiment analysis is generally thought to be of most use at an aggregated level rather than in judging particular individuals; it can identify those processes, interactions, and subject areas that are causing customers the greatest stress and negativity, and help agents view any changes over time. Some businesses do decide to look at sentiment at a team and individual level, but great care must be taken not to attribute negativity to a specific agent rather than the topic or product under discussion.

What Can Go Wrong

In any technology implementation there will be risks of failure. AI can cover a vast amount of territory, and it has the potential to be misunderstood by business owners. Here are some thoughts to consider:

• Expectations of what the AI implementation can actually achieve must be closely managed. There may be an expectation from senior management that implementing AI means that headcount and cost will immediately begin to drop, but in the majority of instances this will not happen.
• AI in the contact center is relatively new. With it being so popular, there is a shortage of skills, support, and resources within the industry as a whole.
• Businesses data assets must be in place before implementation of AI, as this is a technology that relies upon having large, clean pools of actionable data. Without this in place, it will be virtually impossible for any AI implementation to get close to its potential.
• Always have a well-designed and clear path out of AI-enabled self-service (e.g. a chatbot) to a human agent. Trapping a frustrated customer in a self-service session runs the risk of training them not to use self-service again.

In the AI world, knowledge management is not something that is a part-time job, or one that can be handled by amateurs. Consider developing more full-time, expert roles to support knowledge bases and to enable understanding of data models and flows across the entire enterprise. AI experts have to understand both data and also the real-life business/customer issues.

Explain to employees why you’re implementing AI, how it will affect them, what it means for them in their role, and how they can help. There may be concern that their jobs are going to be replaced by bots, so educating them about what the contact center of the future will look like, training them thoroughly and getting their opinion and feedback will help to get them onboard. It is important to emphasize to agents that any AI implementation is about making their jobs more interesting and effective.

While some businesses may look at AI as a way of decreasing their employee costs, this is not a strategy which their customers recommend. The following chart, from our recent survey of 1,000 US customers, shows that there is still very much a preference for communicating with other people rather than through automation. In order to gauge the level of acceptance and expectation around fully automated customer contact, customers were asked whether automation or human assistance would be preferable to the customer base in circumstances where the customer effort, time and outcome were exactly the same. As can be seen, there is still a very strong customer preference for human contact over automation.

Where AI Might Work Best

In the longer-term, most businesses surveyed believe that AI will be used as a key part of handling customer interactions in most businesses, but the question is: how?

The use of AI should be focused on use cases where the AI does a better job than a human, whether that’s being quicker, more accurate, available 24/7 or able to see patterns in data that no person could see.

It’s our view that people call other people not necessarily because they want to hear a friendly voice, or that they’re Luddites who won’t countenance automation, but because they’ve found through experience that this is the most effective way of making sure their issue is resolved. While AI-enabled automation may handle much of the simple work, customers will still seek out a live channel for complex or emotional interactions. And we’re starting to see that the real benefit from AI comes not from replacing agents, but from assisting them with the repetitive and time-consuming parts of their job, meaning that both agent and customer have a better experience.