Contact Center Generative AI: Use Cases, Risks, & Predictions

The customer experience space is awash with generative AI (GenAI) innovation.

Whether it’s supporting supervisors, agents, or customers, lots of use cases catch the eye.

Yet, it’s not all sunshine and roses, as DPD’s GenAI chatbot disaster recently highlighted.

As such, we’ve pulled together some of the best minds in contact center tech to discuss the biggest wins and risks of GenAI implementations. They are:

• Michael McCloskey, CEO of Bright Pattern
• Lorrissa Horton, SVP/GM and Chief Product Officer at Webex
• Michael Wallace, Americas Solution Architecture Practice for CX at AWS
• Martin Taylor, Co-Founder & Deputy CEO of Content Guru
• Jonathan Rosenberg, CTO and Head of AI at Five9
• Brett Weigl, SVP and GM of Digital, AI, and Journey Orchestration at Genesys
• Elizabeth Tobey, Head of Marketing, Digital Solutions at NICE / Lauren Maschio, Product Marketing Manager at NICE.
• Ben Neo, Head of Contact Center and CX Sales, EMEA at Zoom

Below, each expert also shares a prediction on how GenAI will impact the contact center space over the coming year.

Which GenAI Use Cases Are Contact Centers Starting to Implement?

Auto-Summarizations & Insight Generation

Maschio: Generative AI has changed the way contact centers summarize their interactions. It is a key ingredient to automating the process of what has traditionally been a manual, time-consuming task.

Indeed, contact centers can now instantly generate objective and digestible post-interaction summaries with the relevant context for the next agent. Those summarizations include customer sentiment and intent insights.

As a result, the next time a customer interacts with a brand's customer service department, their experience will be connected to their previous interactions, allowing agents to never miss a beat and help facilitate a seamless interaction.

Finally, it saves CX organizations valuable time and money, reducing post-interaction work and average handle time through automation.

Pinpointing CX Improvement and Automation Opportunities

Rosenberg: Five9 anticipates more brands will harness Gen AI technologies to aggregate data for a clear picture of recurring issues and areas for improvement to enhance the services offered by human agents.

However, it’s crucial that CX leaders call AI what it is, or they risk breaking customer trust. Attempting to pass an automated response off as a human, especially in high-emotion scenarios, is a mistake. Instead, companies need to strike a balance between human and automated services.

GenAI-Augmented Live Agents

Neo: Zoom introduced AI Companion this year, a GenAI assistant that helps Contact Center agents complete tasks quicker, thus improving their engagements with end customers.

Such tasks include auto summaries to reduce wrap-up time, suggested next-step actions, live transcription, sentiment analysis to ensure you steer the conversation positively to help the customer, and many more features.

GenAI-Augmented Virtual Agents & Live Coaching

Wallace: Customers are very interested in using generative AI to augment virtual agents.

The early ask was to "replace agents" with generative AI. Yet, as customers have learned more about capabilities and risks, we see more of our customers using GenAI to augment and empower bots to understand and process ambiguous information where they previously could not do so.

Generative AI also surfaces more valuable insight into your CX performance and analytics. With the introduction of AI-powered contact summaries and Agent Scorecards, contact centers may access better insights into CX trends and agent behaviors.

Task Automation

Weigl: GenAI can assist customer-facing employees in working more efficiently through task automation and - in some cases - fully take over certain tasks, data lookups, and other time-consuming elements of customer care delivery.

The technology also provides employees with more timely customer context and can help them find an answer quickly by making suggestions and delivering answers proactively.

Lately, administrators, supervisors, workforce planners, and quality managers can all benefit from GenAI-powered assistants that help to interpret data, spot areas for employee improvement and learning, and automate routine tasks.

Customer Intent Analysis

McCloskey: GenAI makes it easier to monitor customer intent. As such, contact centers can more easily determine why customers are calling and what keywords/topics are the main contact drivers.

Such data is critical for contact centers to spot their demand drivers and take targeted actions – either via process fixes or conversational AI – to lower contact volumes and customer effort.

Moreover, that data is excellent to funnel to execs so they can see where their service costs are coming from. They may then offer more support in overcoming customer journey point points.

Supervisor Support & Generative Audio

Horton: GenAI helps deliver insights to supervisors on how they can support an agent's well-being.

Stress and burnout rates are especially high among agents in contact centers, which see twice as much turnover as any other profession.

Cisco has developed new GenAI capabilities to improve agents' well-being by enabling automated breaks, such as a Thrive Reset, and real-time coaching after difficult interactions.

Also, contact centers can deploy technology to enable smoother audio quality, even when caller bandwidth is low. Webex offers this with its new Webex AI Codec.

GenAI IVR Transformations

Taylor: Although GenAI use cases are applicable across all channels, the biggest efficiency gains typically come during voice interactions.

Indeed, GenAI can make ‘wait time’ productive, gathering information before a customer interacts with an agent - such as who is calling and the nature of their query - to segment and prioritize calls.

GenAI can "listen" during the call and populate the agent's screen with relevant information from approved sources. This prevents the agent from needing to look up the data manually, which could otherwise form 10-20 percent of the interaction, and reduces the time required to train agents.

Share a GenAI in the Contact Center Security Risk, and Note How Contact Centers Can Avoid It.

Adversarial Attacks

Neo: One security risk associated with implementing GenAI in contact centers is the potential vulnerability to adversarial attacks on AI models.

These attacks effectively manipulate input data to deceive AI systems, leading to incorrect or unintended outputs.

When it comes to contact centers, attackers may attempt to manipulate voice recordings or generate synthetic voices to mimic legitimate customers and gain unauthorized access to systems protected by voice biometrics.

By staying vigilant, regularly updating models, and employing additional security measures, contact centers can minimize the risk of adversarial attacks and ensure the robustness of their GenAI applications.

Moreover, ongoing monitoring and auditing of AI systems can help identify and address potential security vulnerabilities as they emerge.

McCloskey: Generative AI's ability to imitate human communication in almost any style prompts serious concerns about automated social engineering attacks.

Preying on users' existing susceptibilities to such attacks can result in disclosing sensitive data or engaging them in security-compromising behaviors.

GenAI Hallucinations

Rosenberg: The main risk to be aware of is AI hallucinations, which is the tendency for generative AI applications to make up information that is just not true.

Human-in-the-loop techniques and data aggregation -  which combines the output of the LLM across many conversations – help mitigate this risk.

For example, AI summary technologies will use these LLMs to summarize every customer interaction, but live agents should review them before submitting the summary to the case history.

Still, this saves a lot of time for agents, thus producing a great ROI, but also minimizes the risk of hallucinations by involving human intelligence.

Such an approach must underpin all AI strategies in 2024. Then, GenAI will empower businesses to create more personal experiences at scale while improving reliability at the same time.

Wallace: The most common concern we hear is: “What about hallucinations?”

So, how do we avoid them? By limiting the LLM's ability to access incorrect data, we can control what it will respond with.

For instance, with Amazon Q in Connect, we give Q access to the customer knowledge base, and answers can only be formed from that knowledge base.

The process (Retrieval Augmented Generation) allows users to update and re-train the information the model can use while limiting the model to information the user defines as safe.

Data Security

Horton: One of the key areas I recommend customers ask vendors about is: Where does my data go when I am leveraging GenAI, and how do I ensure it is safe?

While the benefits of GenAI in the contact center are immense, remember that these capabilities are not foolproof.

As such, contact centers must ensure their systems only leverage data individuals already have permission to access based on that specific data source's privacy and security rules.

In addition, there's always the risk that an AI model produces inaccurate suggested responses or summarization notes, so agents must play an active role in reviewing AI-generated content.

Deepfakes

Weigl: Generative AI enables fraudsters to create content that trips up conventional identification and verification (ID&V) techniques.

Thankfully, fraud detection technology will advance in lockstep to counter risks like deepfakes.

Also, multi-authentication factors will be as useful to prevent ID theft and impersonation. An excellent example is using a text confirmation code routine in addition to voice print.

Moreover, contact centers will use additional predictive signals to detect fraud, given that stolen identity can mean bad actors pass through ID&V in some circumstances.

For instance, communications from an unusual location might trigger a signal even if all other factors check out.

Loss of Data Sovereignty

Taylor: The biggest security risk concerns where LLMs process information.