How conversational AI is shaping customer service [Q&A]

Conversational AI has seen great advances and a massive increase in adoption in the last year or so such that it’s increasingly finding its way into our everyday lives.

We spoke to voice AI industry expert PolyAI's co-founder and CEO, Nikola Mrkšić, to discuss how the technology is shaping customer service and reducing costs, as well as where generative AI tools fit into the picture.

BN: Why has conversational AI become so popular in service environments?

NM: One of the biggest benefits of conversational AI in a service environment is its constant uptime. Any time, day or night, that a customer calls, AI can answer because algorithms don’t need breaks. Not only is AI available 24/7/365, but it also has the ability to resolve many of the FAQs that contribute to high call volumes.

When AI handles these common inquiries, human agents have more time to focus on higher-level calls requiring empathy and a human touch. Taking some of the weight from human agents has become especially important in recent years as contact centers have faced significant staffing challenges, in both retention and attracting new hires. Having AI-powered voice assistants on the frontlines decreases the amount of staff needed to handle customer calls. Since advanced voice AI can answer many common questions, and some systems can contain more complex issues, call abandonment rates decrease and customer satisfaction scores improve.

BN: How closely related is this to the use of generative AI tools like ChatGPT?

NM: Conversational AI uses Natural Language Understanding (NLU) models to understand the caller's intent and can communicate with the caller in language so natural, it's hard to tell that the AI isn't a live agent. With various layers of complex technologies, conversational AI-powered voice assistants can also cut through background noise, decipher colloquialisms and dialects, navigate conversational pauses and handle multiple questions posed in quick succession. In short, conversational AI builds interactive systems for fluid dialogues and are typically built to be organizationally specific.

Generative AI, meanwhile, trains on huge data sets to create completely new content. Generative AI can power intelligent voice assistants -- in fact, we just announced our PolyAI Jupiter platform that’s entirely generative and lets anyone build a voice assistant in a few minutes, for free -- but are more expensive and resource-heavy to utilize for specific use cases. You're also likely to see more hallucinations from generative AI, which is when the answer provided is simply wrong

What you're going to see a lot of is a hybrid of generative's deep learning creativity with the safeguards and boundary setting of more traditional, retrieval-based AI models, which are great at finding answers to questions, but have no character. This RAG (or retrieval-augmented generation) framework will help mitigate risk while allowing for a more expressive interaction between human and machine.

BN: What challenges do businesses face in rolling out conversational AI?

NM: As with implementing any new technology, businesses incur start-up and operational costs associated with rolling out conversational AI. However, those ready to take the plunge and invest in this tech will see long-term benefits unlocking transformational value for both operations and customer experience, in very short time.

Unlike live agents, AI-powered voice assistants don't incur onboarding or training costs and leaders don't have to worry about an AI assistant churning. AI-powered voice assistants can dramatically reduce the call volume -- largely those with mundane or routine issues -- live agents must handle, reducing personnel stress and turnover. Also unlike live agents, AI never forgets to take the opportunity to upsell. With each personalized exchange, AI can introduce special offers tailored to the caller's tastes and needs. In the end, investing in AI-powered voice assistants generates significant ROI more quickly than many other automation efforts.

Another challenge that businesses may face when implementing conversational AI is the issue of trust. Previous exposure to legacy systems -- traditional interactive voice response (IVR) systems requiring callers to speak keywords to move the conversation forward -- makes many people wary of AI’s potential, largely because speaking to a bot historically meant frustration or repeatedly asking for a 'representative!' We’ve all been there.

Unlike IVRs, however, conversational AI-powered voice assistants can detect nuances in the language people use, so interactions with these voice assistants are productive and often resolved without needing to connect to a live agent. As callers are increasingly exposed to helpful AI able to resolve their issues, their trust in the technology will grow.

BN: What questions should businesses be asking before implementing conversational AI?

NM: There are several questions for businesses to consider before implementing conversational AI on the voice channel. For example, how will the voice callers hear sound? While an upbeat and friendly voice will likely be appropriate for most calls, the voice response for someone reaching out after the death of a loved one should be more gentle and compassionate. In other words, the voice's tone should match the content of the conversation, as well as align to the brand.

Next, businesses should ask how they can leverage call data, which provides a wealth of opportunities to personalize customer calls. Data can add context to speed up issue resolution, ultimately improving the overall customer experience.

Finally, businesses should ask how conversational AI can improve agents' work experiences and the customer experience. If AI handles the bulk of FAQ calls, call volume requiring human intervention will decrease and with it, the typical stress experienced by live agents. Businesses should assess how this will affect their org chart, and consider developing AI competency centers where experienced agents can oversee continuous training and refinement of voice AI systems.

AI can also improve customer experiences. With no wait times and highly capable automation containing most calls, happier, less frustrated callers are less likely to be disrespectful to live agents, further improving agents' job satisfaction and retention rates. Better retention means reduced costs and a more knowledgeable, tenured workforce.

BN: How will this impact the role of humans in customer service?

NM: Contact centers will always need humans because some calls require a level of empathy and understanding that AI just can't provide. Additionally, AI will always require human oversight, and its increased usage will give rise to in-house AI Competency Centers, where knowledgeable agents will oversee AI operations and outputs. Ultimately, frontline AI voice assistants and experienced live agents will work hand-in-hand to deliver the best customer experience possible, driving high satisfaction scores and business success.

Image credit: limbi007/depositphotos.com