Thematic Promises Human-Level Quality of Insights From Analyzing Conversations at Scale With Beta Launch of Thematic CA

Thematic, an AI-powered analytics company delivering insights from feedback data, today announced the beta launch of its next generation of Conversational Analytics (CA), a tool leveraging multiple large language models for a strategic analysis of customer conversations. 

Available in beta for current customers, Conversational Analytics is a major expansion from Thematic’s core offering of analyzing reviews, survey responses and community comments. With the power of generative AI, Thematic now provides companies with deep understanding, metrics and trends in hundreds of thousands of customer conversations.  

This means Thematic can analyze conversation data alongside other feedback sources, such as surveys and social media, for a holistic view of customer needs. According to G2 reviews, Thematic can deliver both a detailed and strategic understanding of feedback, while maintaining the speed and scalability synonymous with AI. 

How exactly does Thematic Conversational Analytics help? 
Support and sales conversations are rich sources of intelligence into issues and opportunities, and so businesses need a fast and affordable way to leverage all the calls, messages and emails. Until the arrival of LLMs, it was too hard to deliver a strategic quality level of analysis at scale.  

While Manual QA teams can look at only a small portion of conversations and introduce their biases, traditional conversation analytics will only capture certain keywords or phrases and miss the full context of the interaction. 

The most critical business insights and customer issues arise in daily interactions at the frontline. Yet these conversations are difficult to scan for insights. “Reading support conversations can be likened to reading a very dry textbook on a dry topic. The good signals and rich value get lost in all the noise,” explained Robert Dumbleton, Head of Data Science R&D at Thematic. 

To address this gap, Thematic’s research, engineering and product teams came up with a new generation of Conversational Analytics. Now, a set of LLMs are tasked with a hierarchy of jobs to analyze every conversation across every channel and uncover hidden insights. 

Dumbleton described a hierarchy of AI models working in partnership with each other. After the conversation data is cleaned, LLMs summarize every single conversation to share what it was about, convey important details, and whether it was resolved. Then the AI tags the conversation with themes, customer intent and categorizes its resolution and other insights, such as customer satisfaction. Separately, with a cutting-edge combination of Machine Learning and LLMs, the AI isolates and links to specific excerpts that relate to the conversation summary.  

Where other solutions require users to craft rules for theming conversations, Thematic’s AI does it on its own. Then, a human takes the next step, reviewing the categorizations and guiding the AI to make refinements in a no-code Theme Editor tool.  

This human-in-the-loop approach to validating the analysis is what sets Thematic Conversational Analytics apart from traditional AI models that need to be extensively trained for the task. Their customers say it’s like looking under the hood to see if everything is working and has helped them to build trust with their leadership in making decisions informed by Thematic’s AI.  

At its core, Thematic has a deep understanding of how to apply new AI technology, including LLMs, to perform different tasks that matter to you. For instance, applying AI to deliver the best in class extraction of themes from text, without any prior training. Thematic has created a set of generative AI models to clean data, redact PII, summarize conversations and themes, analyze quantitative data and track ongoing trends.  

All of this is achieved without using customer data to train models, except for that customer alone when they’ve made theme refinements. Their focus on security, compliance and data security means Thematic particularly appeals to the needs of enterprise organizations. 

Once the data sources are plugged in, Thematic AI gets to work to help teams across the business understand exactly what’s happening in their wealth of conversations — and make decisions accordingly. 

No need for fine-tuning or prompts 
LLMs are an easy solution for analyzing individual customer conversations. But analyzing many conversations at scale is not a trivial task. For proprietary business data, you shouldn’t use LLMs out of the box. They need to be directed, contained, and honed to deliver specific outcomes, which is an engineering-heavy task.  

With Thematic’s Conversational Analytics offering, you simply connect a conversation source and let the AI get to work. You can skip the coding, training and engineering. 

“Our mission is to get insights from feedback into the hands of every decision maker. With the rise of LLMs, the challenges in doing this at scale have shifted. Now, it’s not about understanding the language but building a solution that delivers consistent analysis across feedback channels, as well as making insights easily verifiable and accessible. Turning the latest AI advances into solutions is our superpower at Thematic, so we were quick to build an LLM-based solution that works out of the box, is secure and delivers analysis companies can trust,” shared Alyona Medelyan, CEO and co-founder of Thematic. 

The company has already tested the offering with multiple enterprises, with some already signing up for the beta version of the tool.  

Currently, Thematic works with companies like DoorDash, Atlassian, Woolworths and LinkedIn so that every team gets quick and easy access to insights from customer feedback at scale.  

Source: Thematic

Related Media