The future of generative AI models in conversation intelligence

Generative AI applications like OpenAI’s ChatGPT and Google’s Bard are already transforming the way people work. At Tethr, we’re currently exploring ways to incorporate generative AI into our conversation intelligence platform to enhance employee capabilities and reduce workloads in the contact center. Our platform already leverages machine learning models to surface insights from customer conversations, and we recognize that generative AI has the potential to help contact center leaders find and act on those insights even more efficiently.

Below, we’re sharing an overview of the AI models deployed in our platform today, the generative AI applications we’re testing, and our overall outlook on the future of generative AI in the contact center.

<span class="anchor" id="ai-definitions" data-anchor-title="Definitions"></span>

But before we go further, let’s get aligned with some quick definitions of terms we’ll be using in this article.

Artificial intelligence (AI): This is a broad field that uses computer systems or machines to perform tasks that have historically required human intelligence, such as understanding natural language, recognizing patterns, solving complex problems, and making decisions.

Machine learning (ML): This is a branch of AI using computer models that learn from data and improve over time without being explicitly programmed.

Generative AI (GenAI): This is a branch of AI using models that can create new content, such as text or images, by learning from and emulating patterns in existing data.

Large language models (LLMs): LLMs are a part of ML and AI specifically focused on processing and generating human language. They are not necessarily a subset of GenAI, but often are. They're trained on vast amounts of text and are good for natural language processing tasks like translation, summarization, or conversation.

<span class="anchor" id="ai-in-tethr" data-anchor-title="Examples of AI in Tethr"></span>

Where we are today: Examples of AI deployed in Tethr’s platform

Conversation categories

Currently, Tethr employs machine learning to sift through conversation transcripts, pinpointing specific sentiments, events, or behaviors based on the dialogue between customers and agents. These identified themes serve as a structured lens through which contact centers can discern common hurdles, monitor agent performance, and unearth valuable insights into customer experiences. When analyzing larger groups of conversations, categories provide a coherent structure that makes the analysis of common trends and issues more manageable and insightful.

Predictive scores

Tethr’s newest proprietary language models–CSATai and sentiment analysis–help businesses better understand how customers felt about their service interactions and why they felt that way. With these models, Tethr tags specific points in time where strong emotions are detected and provides an aggregate positive, negative, or neutral sentiment score and CSAT score for each interaction. With this, you can now reliably estimate the satisfaction and sentiment of every customer who interacted with the contact center, not just those who completed a survey.

QA automation

Tethr’s QA automation is engineered to automate the mechanical aspects of the quality assurance process, efficiently handling routine checks such as marking 'yes' or 'no' for standard scorecard items like "Did the agent perform the proper greeting?" This automation significantly trims down the time spent on filling out scorecards, liberating QA managers to channel their focus towards more nuanced analyses, personalized agent coaching, and strategizing for long-term improvements. 

<span class="anchor" id="near-term-use-cases-genai" data-anchor-title="Tethr's near-term use cases for GenAI"></span>

‍

Tethr’s near-term use cases for generative AI

GenAI advancements have primarily been focused on understanding diverse text bodies, and the output is, by its nature, still unstructured. In the contact center, conversations are often highly repetitive, with the same issues or queries occurring frequently, often with only minor variations. This repetition contrasts sharply with the diverse content that GenAI advancements have been targeting but presents a unique opportunity for innovation. 

At Tethr, we are combining the generative and creative capabilities of LLMs with our proven structured analysis methods tailored to the repetitive yet nuanced nature of contact center interactions. This fusion transforms the unstructured data from LLMs into organized, actionable insights, enabling more intuitive and responsive customer interactions. By doing so, we aim to build deeper understanding and trust between businesses and their customers. We invite our customers and those considering our platform to join us as we explore the potential of GenAI to revolutionize the contact center landscape.

So what is Tethr looking at first?

Call summaries

After-call work (ACW), including summarizing call notes and updating the CRM, is an essential step in most contact centers today. But the more time an agent takes to complete this work, the longer they will be unavailable to take the next call. Conversely, if agents rush through ACW, they are more likely to make errors or omit valuable information.

Summarization is one of the most apparent applications of GenAI in the contact center, and Tethr is currently beta-testing this functionality. With automatic call summaries, contact centers can capture key details from each customer interaction while reducing the time required for ACW and the average cost per call.

Segment highlights

At Tethr, we're evolving how conversation transcripts are analyzed by introducing the concept of 'Segment Highlights.' This feature extends beyond our current functionality of category tagging by automatically identifying and segmenting significant events within a call. For instance, if an agent assists with rescheduling an appointment and sets up a recurring service, these actions are earmarked as separate highlights. These highlights also allow for quick navigation to specific segments, making the transcript review process more efficient.

The essence of both segment highlights and summaries lies in their goal of distilling the core activities and sentiments from calls, albeit with a nuanced difference. While summaries attempt to condense the content into a compact narrative, segment highlights provide a more structured and navigable breakdown of key events. A crucial aspect of this approach is grounding the GenAI results to the actual dialogues within the call, ensuring the derived segments and summaries are directly traceable to what was said. This not only enhances the accuracy and reliability of the insights but also aligns closely with the established workflow of Tethr categories, promising a richer, more task-oriented understanding of customer interactions. 

Automatic evaluations

At Tethr, we recognize the indispensable value of human insight in Quality Assurance (QA). While our AI-driven tools significantly expedite the evaluation process for objective criteria, we are looking into leveraging GenAI and LLM to help supplement the nuanced and subjective discernment inherent to human QA managers.

We are also testing a new auto-evaluation feature to automatically generate responses to more nuanced scorecard questions, such as “Was the agent courteous throughout the call?” In addition to answering the question, Tethr will also show the specific language in the call used to determine the answer, giving QA managers confidence in the accuracy.  

Auto-evaluations represent a big step forward in QA automation, allowing QA managers to reduce the time they spend on repetitive processes and focus more on strategic process improvements and agent training. 

<span class="anchor" id="looking-further-ahead" data-anchor-title="Looking further ahead"></span>

Looking further ahead

As GenAI continues to evolve, so will the ways we work. We foresee a future where GenAI acts as a helping hand, understanding the context of your tasks and utilizing LLMs to streamline operations.

Tethr's decade-long endeavor has been all about blending AI with the contact center domain, and GenAI is the next step on this path. Our existing models are the outcome of meticulous crafting over the years, and GenAI is poised to meld seamlessly with this established framework. Each model, whether longstanding or newly minted, has a unique role, forming a collaborative ecosystem to tackle the diverse challenges in contact centers.

We're just embarking on our GenAI journey, but we see many opportunities to elevate operational efficiency and customer engagement. We invite our customers and those considering our platform to venture with us into the promising realm GenAI opens up. With a focus on tangible benefits, we're prepared for the positive shifts GenAI can bring to the contact center.