Keeping humans in the loop with contact center AI

Madeline Jacobson

March 13, 2024

From conversational chatbots to QA automation, AI is reshaping the contact center. Amid a period of staff shortages and mandates to do more with less, contact center leaders are turning to AI solutions to improve operational efficiency and deflect calls. So are the days of human employees in the contact center numbered?

The 25 million Google search results for “How to reach an actual person in customer service” suggest otherwise. 

In one study, 53% of consumers said they want to talk to a live agent over the phone when dealing with complex issues. That preference isn’t surprising. Human agents have contextual knowledge, can experience and express empathy, and can approach problem-solving in ways that even the most advanced AI solutions can’t.

Rather than thinking of AI as a replacement for human workers, contact center leaders should think of it as a way to augment human abilities. AI solutions can quickly complete repetitive tasks that would be incredibly time-consuming for humans, generate written responses for humans to review, and analyze data to help humans make decisions. When people and AI team up, contact centers can improve efficiency without losing the oversight and human touch that are essential in customer service.  

<span class="anchor" id="ai-transforming-cc" data-anchor-title="How AI is transforming the contact center"></span>  

How AI is transforming the contact center

Contact center operations look different than they did just a few years ago, with AI solutions running both behind the scenes and in customer interactions. Here are four ways AI is changing the contact center:

1. Automating repetitive tasks

Data entry. Agent scheduling. Quality assurance. There’s no shortage of repetitive contact center activities that are important but time-consuming when done manually. AI technology helps contact centers automate many of these routine tasks so that agents and managers have more time to focus on complex and high-value activities. 

2. Improving agent coaching and performance

AI-powered conversation intelligence software can monitor customer interactions and analyze the language agents are using to determine how they’re performing. This arms managers with specific, data-driven feedback for their team members and ensures agents are evaluated on their performance across every interaction, not just a few randomly sampled calls. 

3. Improving self-service

In a 2022 Boston Consulting Group survey, 95% of customer service leaders said they expect their customers to be served by an AI bot at some point in their service interactions in the next five years. Many contact centers already use AI chatbots to handle routine queries or navigate customers to the right resources. And with the rise of generative AI, chatbots are getting better at responding to natural language queries and handling a wider range of issues. This can reduce calls to the contact center, reducing operational costs and giving human agents more time to spend on the most complex customer issues. 

4. Allowing for more personalized service at scale

71% of customers now expect companies to deliver personalized interactions, according to research from McKinsey. AI helps contact centers deliver personalized service at scale by extracting actionable insights from customer data. For example, conversation intelligence software can analyze customer sentiment in interactions in real time, giving agents a deeper understanding of customers’ feelings and needs. 

“I believe this technology will give companies the ability to harness better insight into their customers and use this to enhance personalized service, anticipate customer needs more efficiently, and enable more proactive problem-solving to reduce customer effort,” says Ellie Bird, Head of Sales and Service Excellence at wholesale food supplier Brakes. “In turn, customers will benefit from quicker resolutions and feel more understood, which will lead to enhanced loyalty and brand trust.” 

<span class="anchor" id="ai-cc-jobs" data-anchor-title="Is AI taking jobs?"></span> 

Is AI taking contact center jobs?

The common thread behind the applications for AI is that it’s dramatically reducing the time required to complete what would otherwise be highly manual, labor-intensive tasks. But does that mean AI is eliminating contact center jobs?

Not necessarily. However, it is helping contact centers do more without increasing their headcount. For example, Tethr customer BCLC reported they were originally planning on hiring two additional QA managers but didn’t need to fill the roles after implementing QA automation. BCLC’s existing QA team has been able to shift its focus from manually auditing a small percentage of calls to analyzing difficult calls and coaching agents with data-driven feedback.

Because AI reduces the time and human effort it takes to complete routine, repetitive tasks, it’s changing the nature of contact center roles. In other words, it’s giving people more time to spend on the activities that people are best at.

AI applications are good at resolving simple issues like password reset requests, but there are still plenty of complex issues that are best left to people. Contact center employees have skills that can’t be easily replicated by AI, including:

  • Empathy: Contact center agents can understand complex human feelings and adjust their tone of voice and responses depending on the situation and the customer’s emotional state.
  • Problem-solving abilities: Generative AI can produce human-like responses, but it isn’t thinking critically the way a human can. Essentially, it uses statistical models to generate the most logical output based on the input. Unlike AI, humans can understand context, intent, and nuances in conversations. Contact center agents can adapt to different situations and use this deeper human understanding to effectively solve complex problems.
  • Contextual knowledge: Generative AI and large language models are only as good as their training data. Additionally, LLMs have memory constraints: the more context you put into a specific prompt, the more computing power it will take (and the greater the cost will be) to generate an output. Human memory doesn’t have these constraints. Contact center agents can apply both explicit knowledge (e.g., specific product information) and implicit knowledge (e.g., knowing how to diffuse a tense situation based on past experience) to successfully assist customers. 

<span class="anchor" id="examples-of-humans-and-ai-together" data-anchor-title="5 examples of AI and humans teamining up"></span> 

5 examples of AI and humans working together in the contact center

The best AI applications fit into contact center employees’ workflows without disrupting them, allowing them to be more productive and focus on the human work that adds the most value. Agents have more time to solve tricky problems and strengthen customer relationships. QA managers have more time for complex analyses and coaching. Contact center leaders hone in on the initiatives that will drive the biggest cost savings and CX improvements.

Here are five ways AI and humans are working together in the contact center today:

<span class="anchor" id="ai-qa-evaluations" data-anchor-title="1. QA and agent evals"></span> 

1. QA and agent evaluations

Contact center quality assurance (QA) can be a notoriously time-consuming process. Traditionally, QA managers listen to a small percentage of each agent’s weekly or monthly calls, fill out a scorecard with 20+ questions for each call, and provide feedback to agents based on their scores. Because of the time required for manual QA, most contact centers only review 1-3% of their customer conversations, making it difficult to spot trends in agent performance or provide meaningful feedback.

Now, however, contact centers can automatically score the objective criteria on their QA scorecard using conversation intelligence technology. Conversation intelligence uses machine learning, a branch of AI, to determine whether agents meet QA criteria based on what they say in their customer conversations. By automating the most time-consuming part of the QA process, conversation intelligence makes it possible to score 100% of customer conversations and give agents more specific, data-driven feedback.

QA automation also frees up more time for QA managers to complete manual evaluations based on complex criteria, such as tone of voice and agent knowledge gaps. Combining these manual evaluations and automated QA scorecards in a conversation intelligence platform provides an in-depth view of agent performance, which QA managers can use to improve agent coaching sessions. 

<span class="anchor" id="ai-acw" data-anchor-title="2. After-call work"></span> 

New call-to-action

2. After-call work

After-call work (ACW) for contact center agents typically includes writing up notes on the call, logging a call reason, scheduling any follow-up actions, and updating the CRM. ACW is important because it ensures customer records stay up to date, contact center leaders get visibility into activities happening on calls, and agents close the loop on any customer issues that require follow-up. But the more time an agent spends on ACW, the less time they have to assist customers.

Fortunately, AI can significantly reduce the time agents have to spend on ACW. For example, conversation intelligence technology can automatically identify call reasons or required follow-up actions and document them in a CRM. Contact centers can even set up triggers in their conversation intelligence platform so that when specific phrases or actions appear in a conversation, the right teams are notified and can respond appropriately. Additionally, GenAI applications can generate call summaries using call transcripts as the input. From there, a human agent can verify for accuracy and make any necessary tweaks.

<span class="anchor" id="ai-customer-service-outreach" data-anchor-title="3. Customer service outreach"></span> 

3. Informed customer service outreach

AI can help contact centers determine when proactive human outreach can improve a customer’s experience. For example, one of our customers, Thrasio, uses Tethr’s CSATai to get a predicted customer satisfaction score for every closed customer service ticket. Tickets with predicted dissatisfied scores are automatically reopened and returned to a human agent. The human agent then determines if they need to reach out to the customer directly to address the issue causing dissatisfaction. This personalized human outreach can increase customer satisfaction and reduce the likelihood that a customer will submit a negative survey response.   

<span class="anchor" id="ai-response-generation" data-anchor-title="4. Response generation"></span> 

4. Response generation

Contact centers are beginning to test generative AI chatbots to improve self-service and deflect calls. GenAI chatbots may be able to handle more complex queries than rule-based chatbots, which operate on “if/then” logic and do not learn over time. But with GenAI, there’s also a risk that a chatbot will share inaccurate information or incorporate harmful biases into its responses. There have already been several high-profile examples of this, such as Air Canada refunding a customer after its chatbot shared incorrect information about the airline’s bereavement rate, and a Chevy dealership having to rescind its chatbot’s offer to sell a customer a new car for $1.

Contact centers can reduce the risk of their chatbots sharing incorrect or harmful responses by keeping humans in the loop. The level of human oversight needed for AI applications depends on the level of risk. For example, a virtual assistant for a clothing retailer that offers outfit recommendations is relatively low-risk–this chatbot could likely operate on the retailer’s website with a conversation intelligence platform monitoring its responses to identify opportunities for improvement. On the other hand, a virtual assistant that shares financial advice in a banking app would be much higher risk. In this case, it might make sense for a human to review all responses for accuracy before sending them to customers. 

<span class="anchor" id="ai-cx-insights" data-anchor-title="5. CX insights identification"></span> 

5. CX insights identification

Traditional methods for collecting CX insights have several limitations. Post-interaction surveys suffer from low response rates, sample bias, and limited nuance. Qualitative research methods, such as focus groups and customer interviews, deliver valuable insights but can be time-consuming and expensive. 

AI gives businesses the ability to quickly access CX insights from the data that already exists in their contact center. Machine learning models can analyze customer conversation data and pinpoint key CX insights, such as drivers of satisfaction and dissatisfaction, churn risk factors, and moments of friction in the customer journey. This gives contact center leaders clear actions they can take to improve the customer experience. It’s up to people in the contact center to take those actions. AI essentially provides the blueprint for CX improvements, but people must execute on it.  

<span class="anchor" id="ai-and-humans-final-takeaways" data-anchor-title="Final takeaways"></span> 

Final takeaways

The Boston Consulting Group predicts that when implemented at scale in customer service operations, generative AI could increase productivity by 30-50%. The important thing to keep in mind is that this increase in productivity doesn’t make human contact center roles obsolete. Instead, it allows people to reallocate their work time to activities that are highly impactful and more satisfying. Contact center roles may be changing, but people continue to be essential to delivering high-quality customer service.  

New call-to-action
Jump to:

Most popular articles