Sergey Galchenko, Chief Technology Officer, IntelePeer – Interview Series

What initially attracted you to the field of computer science and AI?

I enjoy solving problems, and software development allows you to do it with a very quick feedback loop. AI opens a new frontier of use cases which are hard to solve with a traditional deterministic programming approach, making it an exciting tool in the solutions toolbox.

How has AI transformed the landscape of customer support, particularly in automating CX (Customer Experience) operations?

Generative artificial intelligence is revolutionizing the contact center business in unprecedented ways. When paired with solutions that help automate communications, generative AI offers new opportunities to enhance customer interactions, improve operational efficiency, and reduce labor costs in an industry that has become fiercely competitive. With these technologies in place, customers can benefit from highly personalized service and consistent support. Businesses, simultaneously, can contain calls more effectively and battle agent turnover and high vacancy rates while allowing their employees to focus on high-priority tasks. Finally, gen AI, through its advanced algorithms, enables businesses to consolidate and summarize information derived from customer interactions using multiple data sources. The benefits of utilizing those technologies in the CX are clear – and there is more and more data supporting the case that this trend will impact more and more companies.

Can you provide specific examples of how IntelePeer’s Gen AI has reduced tedious tasks for customer support agents?

The ultimate goal of IntelePeer’s gen AI is to enable complete automation in customer support scenarios, reducing reliance on agents and resulting in up to a 75% reduction in operation costs for the customers we serve. Our platform is able to automate up to 90% of an organization’s customer interactions, and we’ve collectively automated over half a billion customer interactions already. Not only can our gen AI automate manual tasks like call routing, appointment scheduling, and customer data entry, but it can also provide the self-service experiences customers increasingly demand and expect—complete with hyper-personalized communications, improved response accuracy, and faster resolutions.

Can you describe why AI-related services must balance creativity with accuracy.

Balancing creativity with accuracy and predictability is critical when it comes to fostering trust in AI-powered services and solutions—one of the biggest challenges surrounding AI technologies today. First and foremost, it should go without saying that any AI solution should strive for the highest level of accuracy possible as to provide the right outputs needed for all inputs. But creating a great experience with AI goes beyond just providing the correct information to end-users; it also includes enabling the correct delivery of that information to them, which takes a decent amount of creativity to execute successfully. For instance, in a customer service interaction, an AI-driven communications solution should be able to automatically match the tone of the customer and adjust as needed in real time, giving them exactly what they need in the way that will best reach them at that moment. The AI should also communicate in a life-like manner to make customers feel more comfortable, but not so much as to deceive them into thinking they’re speaking to a human when they’re not. Again, it all goes back to fostering trust in AI, which will eventually lead to even more widespread adoption and use of the technology.

What role does data play in ensuring the accuracy of AI responses, and how do you manage data to optimize AI performance?

Good data creates good AI. In other words, the quality of the data that’s fed into an AI model correlates directly with the quality of the information that model produces. In customer service, customer interaction data is the key to finding gaps in the customer journey. By digging deeper into this data, organizations can begin to better understand customer intents and then use that information to streamline and improve AI-driven engagement, transforming the overall customer journey and experience. But organizations must have the right data architectures in place to both process and extract insights from the massive amounts of data associated with AI solutions.

The IntelePeer AI solution uses the content and context of the interaction to determine the best course of action at every turn. During an interaction, if a question is posed by the customer that requires an answer specific to a business’s process, rules, or policies, the AI workflow automatically leverages a knowledge base that includes such business data as FAQ documents, agent training materials, website data, policy, and other business information to respond accordingly. Similarly, if a question or a request is made that the business does not want AI to respond to directly, the AI workflow will escalate the query to a human agent if required. The remaining interaction can be automatically added to the Q&A pairs to enhance responses in subsequent customer interactions or handed off to a supervisory authority for approval prior to incorporation.

With AI’s increasing role in customer support, how do you foresee the role of frontline agents evolving?

We at IntelePeer envision a drastic reduction in the reliance on frontline agents due to the evolution of AI technologies. With massive strides in AI-driven call containment, which continues to improve in quality and grow in volume, organizations today are able to automate up to 90% of their customer interactions. This allows them to optimize their frontline staffing and save significantly on operational costs—all while providing better experiences for the customers they serve.

While some tasks are automated, which skilled CX roles do you believe will remain critical despite AI advancements?

While AI will cut down on the number of frontline agents needed in customer service roles, a human element will always be needed in CX operations. For example, AI-powered communications models must be trained, configured, and managed with human oversight to ensure accuracy and the elimination of any biases. The human touch is also needed to align automated customer communications with the messaging and personality of the organization or brand they’re coming from, which contributes to customer comfortability and helps to foster trust in the technology. These more technical, AI-oriented roles will overtake typical frontline roles in the years to come.

AI hallucinations are a concern in maintaining accurate customer interactions. What specific guardrails has IntelePeer implemented to prevent AI from fabricating facts?

 Businesses need to implement generative AI today to stay relevant amid the ongoing revolution while avoiding a rushed and disastrous rollout. In order to do that responsibly, companies must start with implementing a Retrieval Augmented Generation (RAG) pattern to help their gen AI interface with analyzing large enterprise datasets. For automated customer service interactions, brands must create a human feedback loop to analyze past interactions and improve the quality of those datasets used for fine-tuning and retrieval augmentation. Further, in order to eliminate AI hallucinations, organizations should be laser focused on:

• implementing guardrails by analyzing customer interaction data and developing comprehensive, dynamic knowledge bases;
• investing in continuous monitoring and updating of these systems to adapt to new queries and maintain accuracy; and
• training staff to recognize and manage unidentifiable permutations ensures seamless escalation and resolution processes.

How do you ensure that large language models (LLMs) interpret context correctly and provide reliable responses?

 A haphazard approach to implementing gen AI can result in output quality issues, hallucinations, copyright infringement, and biased algorithms. Therefore, businesses need to have response guardrails when applying gen AI in the customer service environment. IntelePeer utilizes retrieval augmented generation (RAG), which feeds data context to an LLM to get responses grounded in a customer-provided dataset. Throughout the entire process, from the moment the data gets prepared until the LLM sends a response to the client, the necessary guardrails prevent any sensitive information from being exposed. IntelePeer’s RAG begins when a customer asks a question to an AI-powered bot. The bot performs a lookup of the question in the knowledge base. If it cannot find an answer, it will transfer to an agent and save the question to the Q&A database. Later, a human will review this new question, conduct a dataset import, and save the answer to the knowledge base. Ultimately, no question goes unanswered. With the RAG process in place, businesses can maintain control over response sets for interaction automation.

Looking ahead, what trends do you anticipate in AI’s role in customer experience?

At IntelePeer, we deeply believe that generative AI is a powerful tool that will positively augment human communication capabilities, unlocking new opportunities and overcoming long standing barriers. AI will continue enhancing customer service communications by streamlining customer service interactions, offering around-the-clock assistance and providing language-bridging capabilities. Moreover, trained on large language models (LLMs), virtual assistants will be able draw upon millions of human conversations to quickly detect emotions to modify its tone, sentiment and word choice. There will be more and more evidence that businesses that successfully use AI to enhance human connections experience see a significant return on investment and improved efficiency and productivity.

Thank you for the great interview, readers who wish to learn more should visit IntelePeer.