Home > Columns > Executive Interviews

PolyAI Executive Interview

Click image below to view the ebook

Sheri Greenhaus, Managing Partner of CrmXchange, sits down with Nikola Mrkšić, Co-founder and CEO of PolyAI, to explore the transformative impact of voice AI on customer service. They discuss how PolyAI's technology enhances contact center efficiency, reduces stress for agents, and improves overall customer experience through seamless voice interactions.

Sheri: The CrmXchange audience is interested in, "How can I help my customers? How can I make my job better?" There is some fear of ‘robots' taking over their jobs.

Nikola: The audiences you described are exactly who we work with. Rather than automating their jobs away, we're helping them become AI implementers. Everyone talks about AI, but we don't see much change in daily lives yet. People say contact centers will be where AI's impact is felt first, but it hasn't happened that quickly. For us, it's about supporting customer experience executives and contact center managers dealing with churn, budget cuts, and headcount reductions. Contact center agents face a tough job, dealing with abuse from customers over issues that aren't their fault.

We help by implementing voice assistants to handle simple to moderately complex tasks, freeing up human agents for more complex issues. This makes everyone happier—P&L owners save money, and agents focus on more rewarding tasks. Our vision isn't to replace contact centers but to have AI work in tandem with human agents. Many tier-two agents are becoming the first scaled AI operators, monitoring AI performance and ensuring efficiency.

Sheri: How did you get into this space?

Nikola: I entered this field almost accidentally. At the University of Cambridge, we worked on voice assistants, and after two years, our previous company was bought by Apple. We realized contact centers were a challenging but promising area because few dared to implement new technologies there. We've been at it for six and a half years now.

Sheri: I saw some demos on your site. When a customer calls, the response isn't from a live agent, right? It's your solution answering and helping. If the automated agent can't handle it, it passes the call to a live agent?

Nikola: Exactly. When the automated system gets stuck, like struggling to collect a VIN number, it hands off to a human agent with all the gathered information. For instance, if you're making a midterm adjustment for your insurance, and we've confirmed your identity but can't parse the VIN number, we'll transfer you to a human agent.

Sheri: Can the system learn from new issues it encounters and apply that knowledge to future calls?

Nikola: To some extent, yes, but it's complex. The voice assistant involves multiple machine learning components. One part is speech recognition—transcribing what you said correctly. The next part is understanding what to do with the transcribed text. It learns in a few ways: humans label data, and it measures user reactions through reinforcement learning, although this is challenging.

So, can it learn? Yes, it can, but in different ways. One method involves humans labeling data, providing the system with clear examples. Another method is measuring user reactions, a process known as reinforcement learning. This approach is complex and challenging, requiring substantial effort. It's not fully self-learning; it involves a mix of failing gracefully and then learning from those failures with human oversight. Contact centers want control over these systems. They don't want a black box run by generative AI without understanding its decisions. Instead, they prefer a structured approach where humans guide and reprogram the system based on observed performance and policy changes.

Sheri: Who in an organization typically starts the training process?

Nikola: That's a good question, and I wish there was a universal answer.

Our work typically starts with what we call the "contact center plus one" approach. We identify who in the company is responsible for customer experience (CX), service levels, and ensuring that metrics like abandoned call rates and response times are met. The person who feels the most pressure to maintain these service levels usually appreciates our support the most because we can provide that extra ‘labor’ they need.

Within organizations, the people who work on these projects are typically subject matter experts. They come in all shapes and sizes and have usually been with the organization for many years. They are smart, agile, and accustomed to the fast-paced, stressful environment of the contact center. They know where everything is—information, processes, procedures—and how to implement changes. These are the people who are highly respected and sought out for their knowledge. Often, they are tier-two agents or manage tier-two agents, handling more complex issues. They work closely with executives to help us onboard and integrate our solutions. By the time we've been working with an organization for three to six months and have launched the voice assistant, there's typically a subject matter expert at PolyAI ready to function as a tier-two agent.

Sheri: When a company says they have good results, what does that look like?

Nikola: Cost savings are a significant benefit, but I'd rank that as the third most common reason companies are satisfied. The primary benefit is customer service improvement. For example, when COVID hit, many companies lost staff and struggled to hire replacements, leading to long wait times and high abandoned call rates. Implementing our solution can initially automate 30% to 50% of calls, which is substantial. We avoid frustrating touch-tone IVR systems and ensure that if a customer wants to speak to a human, they can do so without excessive barriers. Our goal is to build trust in our voice assistants by making them lifelike and not obstructive.

When you successfully automate 30% of calls, agents no longer need to take back-to-back calls constantly. They have time to wrap up each call properly, aren't rushing, and aren't stressed by long queues. This reduces the overall stress in the contact center, making the environment more pleasant and efficient.

Sheri: Have you heard directly from clients about reduced stress?

Nikola: Yes, we have. We've received tokens of appreciation, like muffins, from clients thanking us for giving them enough time to grab a muffin between calls. Our voice assistants have even won "Employee of the Month" awards from different clients. Just this morning, a European bank told me how much they love their voice assistant, Mia. It's heartwarming to see our technology being embraced in an area that typically doesn't get to invest in much beyond CCaaS and workforce management solutions.

This technology often elevates subject matter experts within the company, giving them opportunities to present to the board and take on leadership roles in AI transformation. One thing I didn't anticipate was how wonderful contact center people are—their emotional intelligence and resilience are remarkable. Our technology helps reduce frustration on both sides of the call, allowing agents to handle more interesting work and providing greater job satisfaction.

Sheri: With AI being such a buzzword, how do you differentiate yourselves?

Nikola: The original company in this space was Nuance, which built the first generation of voice assistants for large American contact centers. Those early systems often reduced agent numbers at the expense of customer experience, creating a lot of frustration. While Nuance did the best they could with the technology available at the time, today's conversational AI companies usually focus on chatbots for online platforms or text-based interactions like web chat or WhatsApp. While these can work well if integrated properly, they often lead to agents juggling multiple customers simultaneously, which frustrates end users.

There's a misconception that older demographics prefer phone calls while younger people avoid them due to social anxiety. In reality, people generally prefer talking because it's a natural and evolved form of communication. They appreciate being met with a knowledgeable agent on the other end of the line. Our focus is on creating voice assistants that provide a human-like, engaging, and efficient experience, fostering trust and satisfaction.

When you look at many conversational AI companies, a lot of them focused on chat and ignored voice, thinking it would phase out over time. However, voice isn't going away—it actually grew during COVID. People want digital solutions because they're cheaper, even if they involve human interaction, and they are also easier to automate. Companies like low-cost airlines in America and Europe tried to eliminate phone support, but I believe they'll eventually revert because customers will start turning to competitors.

Another trend is agent assist tools. Companies use these tools to address high churn rates, offshore challenges, and the need for better performance. The idea is to use software to enhance agent capabilities, but it’s incredibly challenging, especially for voice interactions. The software needs to provide information instantaneously, almost like an Olympic discipline, for agents to use effectively during live calls.

Our approach focuses on mastering voice interactions with voice assistants. We aim to make them as seamless as possible, even if we're only automating 20-30% of calls initially. That’s a huge win for contact centers because it reduces stress and improves efficiency. Our goal is to build trust and ensure that users have a positive experience with our voice assistants, gradually shifting their attitude towards automation.

We recognize the value of integrated chat experiences in certain contexts, like apps where quick, convenient support is needed. However, for most situations, people prefer voice interactions. They find them more natural and reliable, especially when compared to chat systems that can be disrupted by something as simple as a page refresh.

Sheri: I agree. If you're going to implement a solution, you should do it right from the start to avoid frustrating your customers and incurring higher costs in the long run.

Nikola: Exactly. When we started, I didn't fully appreciate the significance of putting a voice assistant at the front line of customer interactions. It requires a high level of trust because it becomes the first point of contact for customers. It took us time to develop voice assistants that sound human and provide excellent service. Companies that have never used automation before are now adopting our technology because they see the value it brings.

For instance, Metro Bank in the UK placed 14,000 test calls into our system before allowing it to handle 2% of their calls. They listened to every automated call for months to ensure quality. This level of dedication shows how much effort both sides put into co-creating a solution that truly represents the brand.

Sheri: Is there anything else our audience should know about your company?

Nikola: We work with brands across various industries, delivering ROI in different forms. For example, we collaborate with FedEx in 17 countries to automate calls and provide detailed data that humans might miss. This helps in extracting meaningful insights and improving customer service. We also work with hospitality groups, home services companies, and dental offices to ensure they never miss a call, boosting their business.

(Note: At this point in the interview, Nikola  played a recording of a customer calling to make a restaurant reservation. The ‘attendant' sounded as good as, if not better than, a live person handling the reservation.)

Our voice assistants are customized for each client, considering language, dialect, tone, and style. This personalization makes the interaction feel unique and enhances the brand experience. We've seen significant results, such as a 4% revenue increase for a restaurant group by ensuring they never missed a reservation call.

Sheri: You recently won an award at Enterprise Connect?

Nikola: Yes, we won two awards: Best Innovation in Customer Experience and Most Innovative Use of AI. Our solutions range from simple calls to complex interactions like troubleshooting hardware, scheduling appointments, and arranging non-urgent medical transport. We thrive on solving difficult problems with AI to make life easier for our clients and their customers. Our ultimate goal is to change the perception of AI from something people hate to something they love interacting with.