The Future Of Conversational Ai Begins With Continuous Intent Discovery

Oct 17, 2023

Generative AI and LLMs have been all the rage over the last two to three quarters, and many companies are scrambling to get ‘something’ delivered.

While others are trying to figure out how to apply Generative AI and LLMs across the enterprise, we have been using AI models like GPT and Bard for over four and a half years.
What we learned was humbling. We realized we needed to combine our proprietary foundational model development (what we call dominant path analysis, DPA for short) with the best Open-Source AI Models, all while leveraging the state-of-the-art in Inference Computing and Open-Source Generative AI Research. This took time. Over three and a half years later and $25M+ investment in R&D and Engineering, we evolved our maturity and the capabilities we have developed to automatically build (in hours, not months) a custom set of AI models based on that customer’s own conversation data. Then these models automatically analyze 100% of ongoing customer conversations for each business segment a company has (sales, service, tech support, collections, customer retention, etc.).

This approach enables significantly improved accuracy to understand these highly specialized, complex conversations between customers with specific issues and agents who are trained to handle those topics. This is why Talkmap has been selected by the world’s leading enterprises to ‘supercharge’ all of their contact center platforms. Because accuracy is critical and Talkmap’s platform delivers it better than any other offering these organizations have implemented or evaluated. Importantly, Talkmap’s conversational intelligence integrates seamlessly into existing call center platforms, agent workflow, and system-level operations…improving accuracy and insight across all of an organization’s existing platforms and the customer interactions they handle.

As of last January, very few firms in the Telecommunications, Call Center, and Cloud Services industry were leveraging customer data to build AI models. The accepted best practice across the Experience Management, Customer Engagement, and Customer Relationship Management industries was to use pre-built industry vertical models.

But now the world has changed and the market has shifted. The Telecom and CRM industries are asking for Conversational AI models which are ‘always on’, in production, and dynamically and continuously learning what Customers are saying…all without hallucinations and with an accuracy that can be trusted.

Talkmap’s introduction of Continuous Intent Discovery unlocks the power of 100% visibility into each and every customer interaction. This delivers the critical missing link that enables timely and robust enterprise knowledge, insight, and intelligence that transforms a brand’s ability to deliver remarkable Customer Experiences and Employee Experiences.

With Continuous Intent Discovery, Enterprises now have the ability to discover new Client and Agent intents (including associated utterances), immediately and over time, from aggregated customer conversation data, without hallucination. Importantly, Talkmap also automatically and dynamically incorporates these new intents (labels) into new, highly accurate machine-learning models for ongoing conversation data enrichment and processing with no need for human interaction. This revolutionizes the efficiency, productivity, and profitability of Call Center Operations.

What does Continuous Intent Discovery mean for CIO’s and CTO’s? It means you never need to re-program your models or take them out of service. This changes all aspects of contact center transformation, digital transformation, marketing, customer acquisition and margin expansion across the Enterprise. From Cloud Data Services to Platform Engineering to Application Development to IT Operations.

With Talkmap, Enterprise IT, Cloud, and AI teams can easily deploy our conversational intelligence data services at scale, providing visibility and insight from their customer conversation data within hours (vs. weeks or months with pre-built or industry-specific models). With Continuous Intent Discovery, the intelligence around a customer’s preferences, desires, the context of their call, and their overall intentions become more clear, insightful, and actionable.

Moreover, our next major release will incorporate some of the new LLMs that have recently been introduced in the market. Additionally, this release will further advance the summarization of conversations, saving valuable Call Handling Time for agents while delivering high-quality call summaries that are both accurate and automated. This further enhances the value to Call Center Operators and their IT teams with their AI model development leaders.

What’s exciting is that we have already completed the research and proven this past September that we can automatically generate labels that have the highest accuracy in production that we have ever developed:

  1. With an ‘out-of-the-box’ accuracy rate exceeding 80%, and
  2. We can do it within hours (as opposed to months), and
  3. Without any tuning of the models (and all 100% secured in our private cloud)

All three of these are industry firsts. Impressively, all the while doing this with a SOC II Type II security posture. Previously, the only way everyone in the industry could get to the 80% target was to go through a manual tuning process, which is why we are so thankful to our enterprise customers for their support in helping us over the past three years evolve and mature our Research and Development to introduce Continuous Intent Discovery.

If you’re interested in learning more, please reach out to Talkmap’s product management and business development teams at [email protected]

Thank you for all the feedback over the years.

Myron Prevatt
Chief Product Officer