Defining conversation intelligence can be a difficult endeavour, with a broad range of offerings muddying the waters by claiming association with the technology. At its core, conversation intelligence involves extracting information from digital communications. Building on top of that with real-time analytics can reveal the true value that was previously being lost in human-to-human conversations, be that between customers and colleagues, doctors and patients, teachers and students, and many more.
In an exclusive interview, CX Today welcomes Surbhi Rathore, CEO and co-founder of conversation intelligence company Symbl.ai, to discuss what conversation intelligence really is, how it adds value, and where it's headed in the future.
“Some people confuse conversation intelligence with chatbots and virtual assistants or with sales call analysis tools,” says Rathore. “We believe that conversation intelligence really implies the presence of an artificially intelligent system that can consume, understand, analyze and/or reciprocate in any conversational interactions .”
A Horizontal Approach
End-to-end conversation intelligence requires the presence of numerous programmable layers - as Rathore explains. “Most businesses start with speaker-labelled speech recognition with basic formatting and accuracy boosting approaches to make the text more readable and presentable for a fine reading experience. To support that, conversation analytics are a low hanging fruit to understand all the ones and zeros from the conversation and how speaker ratios, pace, overlap and other such parameters are identified as the conversation progresses.
“The next layer is formed by insights and structure that can be automatically generated from the conversation like action items, questions, topics, outlines. These contextual insights help businesses understand conversations individually and in an aggregated way without the need for any prior knowledge of what to look for. For more advanced use cases, the third layer of domain intelligence lets developers build personalized vertical intelligence, provided you have some idea of what to look for. The key here is to make it developer-friendly without the need for extensive data training or manual labelling efforts while reducing the time to implement in scale. These three layers demonstrate the journey businesses go on, from simply having a voice stream as part of a file or live to actually generate and using the conversation data as a function of growth in their business.”
The real-world examples of the application of conversation intelligence are many, not least in agent coaching in the contact centre space. “Agent coaching is so fundamental, but so is coaching in general. Learning from conversations to ensure you can apply that intelligence to future discussions applies across other areas such as recruitment or training.” Compliance is another mainstay of technology. “That can mean a lot of different things based on where it is being applied to,” says Rathore. “Content moderation, for instance, which involves profanity detection and flagging the right content so that it does not touch primary channels.” Aside from its use to automate data entry and take the burden off of workers, Rathore also enthuses about the technology’s use in building more accessible experiences. “That’s not just live captioning, but also being able to index the content in smaller pieces for easy distribution and knowledge creation - tagging content for accessibility or easy navigation, for instance.”
Symbl.ai differentiates itself from competing solutions in the field by being an end-to-end conversation intelligence platform, and by focusing on understanding conversation data contextually - so that businesses can go beyond speech recognition easily and apply intelligence in any domain. “Most of the businesses or technologies out there have two problems,” says Rathore. “Either they're very domain-specific - built to only understand one type of conversation data - or their understanding is based on keywords or phrase matching. Our patented technology is about fitting any speech into a conversation structure which can then be applied to various other structured data points.” That approach strikes at the foundation of the issue. “If you transform speech data into a structure, which is then commonly used across several domains and applications to derive more insights, you basically solve the fundamental problem of understanding human conversations, rather than understanding every conversation differently. Once the conversation moves from unstructured to structured data, you can then do deeper analysis across any domain and build further statistical models that combine real-time conversation understanding with existing data in your system of records to make predictions more powerful and accurate.”
Multi-Industry Potential
Industries across the spectrum are waking up to the potential of the technology, with the customer support industry leading the way in terms of maturity, sales and marketing, healthcare, edtech and collaboration are all key verticals ready for disruption. “The first instances of converting voice to text happened in the contact centre space at scale - but there's so much that can still be done in call centres around real-time agent coaching and agent assist. For instance, dynamically changing dashboards based on what agents and customers are talking about.” Education is another ideal use case, especially during this new world of remote working. “Everyone has had to move to a digital medium and children are among the most impacted - because they don't know how to navigate the digital content which is getting created at such an abundance. Helping students be more efficient in consuming information rather than noting it down can really help accelerate learning and help them focus on being present.”
