Extracting Business Insights From Your Calls and Meetings

Before we begin discussing how to extract business insights from call and meeting recordings, let's take a step back and discuss the challenges associated with this process.

Imagine you're on a call, listening to a conversation between colleagues about the future of your company. One participant suggests that an investment should be made in the launch of a new product. The discussion then leads to an idea about a competitor entering the market. But then you realize that someone made an off-hand comment about 'the blue stuff'. 'What blue stuff?' you ask yourself. In your mind, you go back over the conversation to try and figure out what was being discussed. You can't recall any mention of 'the blue stuff' at all. This appears to be a moment of great importance, but what exactly was said? What if you missed something? And if so, is it still relevant?

Without historical context (memory of previous conversations), your ability to understand what was discussed and why it was being discussed is severely limited. Our limited attention spans and memories cannot hope to keep track of every previous conversation, including all the ones that may have mentioned 'the blue stuff' before. This is why it is important that our AI systems create searchable indexes of our conversations that can be mined for historical context.

Transcription has been around for decades, and it works extremely well when it comes to translating a conversation into a human-readable format. But unfortunately, raw text transcripts are not built for mining insights. They alone do not explicitly identify important topics discussed in the conversation, nor do they identify the sentiment, facts, actions, complaints, accusations, or any other insights that might have been expressed in a way that machines can understand. They only translate human speech into a readable format.

Going beyond traditional transcription is not an easy task. Meetings can be complicated to capture, with so many participants talking at once. But this is a challenge that today's machine learning systems can solve.

Let's now discuss how to turn the traditional transcription process into something more valuable: one that automatically extracts business insights from conversations and builds a searchable knowledge graph from a company's calls and meetings.

Transforming a transcript into searchable knowledge involves applying many different natural language processing (NLP) approaches, for instance, to identify the important topics in the conversation, or the sentiment expressed by each speaker. One must identify actions taken and plans made during the conversation, and other key facts. Finally, the conversation data needs to be indexed into a searchable format.

Once this is done, it becomes easy to look for important topics discussed in previous conversations using graph exploration of prior conversations, people, or companies related to a topic. One can then perform a sentiment analysis on each speaker's statements over time to identify any changes in their opinion about a topic. We can also use our index to identify any common topics discussed by multiple speakers and to find when subjects were first (or last) mentioned.

Going back to our earlier example, let's say someone mentions 'the blue stuff' in a call or meeting. With a conversation index, an AI system can search for all prior conversations where someone mentioned 'the blue stuff' and identify anyone else who mentioned it in a prior conversation. This information can be used to better understand who is talking about 'the blue stuff', and why.

The ability to search for previous conversations that discussed a topic is incredibly powerful. It allows us to quickly narrow down our focus to only those conversations that actually matter. 

The goal of a conversation index is to allow us to perform more intelligent searches based on topics, people, sentiments, actions, and so on. It allows us to filter out conversations that don't relate to what we are looking for while highlighting the conversations that do. It also allows us to focus on conversations where certain people mentioned something important about the topic of interest.

So how can you extract the business value from your call and meeting recordings? The best option is to employ AI-based solutions to automatically process and AI enriches your call and meeting recordings. These systems use machine learning techniques to turn raw transcripts into actionable information by automatically identifying important topics discussed during each call and meeting, extracting the sentiment, facts, actions, complaints, accusations, and any other insights that might have been expressed in a way that machines can understand. Solutions such as Hyperia build a searchable knowledge graph by combining these topics into a single narrative that is accessible across your business, making it easy to look up information on specific customers, conversations around problems or issues, and more.