Machine Learning Strengthens Interactions Omnichannel Offerings

After the recent news of Interactions’ acquisition of Digital Roots, a provider of AI-based social media customer engagement tools, Opus Research posted an initial assessment of the technology. We’ve since had an opportunity to learn more about Digital Roots from Jay Wolcott, the company’s founder and CEO.

With the Digital Roots acquisition, Interactions has added impressive new capabilities to its toolbox. Chief among them are powerful natural language analysis tools driven by machine learning. The Digital Roots application is currently focused on the social media channel. In an ever-growing sea of consumer posts, it’s humanly impossible for brand managers and customer support agents to find and shift through all mentions of their brand. It’s even more challenging to determine how to handle these posts.

The Digital Roots software can connect to major social media sites and the web to automate this discovery process. The technology scans enormous numbers of consumer posts at scale. But the real trick is picking out those that are actionable. Wolcott explained that the product’s machine learning algorithms are trained to make this determination by leveraging user feedback.

Posts that warrant a response have something in common, and it’s the job of the algorithm to figure out what that is. Out of the box it does a great job at surfacing actionable posts based on massive amounts of training data compiled over almost a decade.

It turns out that actionable posts generally fall into two main categories: marketing/pre-sales questions or requests for post-sales support. Marketing related posts might consist of someone asking about a specific brand or mulling the purchase of a product. Comments related to post-sales support are often complaints or direct requests for help with a product issue.

Surfacing and categorizing actionable posts through natural language analysis is only one side of Digital Roots’ capabilities. The other side is its ability to generate possible responses. Just as the algorithm learns to identify actionable posts by seeing which ones warranted a human reply, it hones its recommendation skills by analyzing the responses penned by human agents. Over time, as humans rate the quality of the algorithm’s responses and suggest improvements, the algorithm gets better at generating usable suggestions.

Though the Digital Roots application currently focuses on analyzing social media posts, these same machine learning technologies could be applied more broadly across other channels. Automatically detecting when a customer is asking for help, and what kind of help they’re requesting, are valuable tools for any channel. Providing busy call center agents with reliably good responses for a variety of customers and issues is another great benefit.

Interactions’ key differentiator has been its demonstrated ability to blend elements of AI with human interaction to support Virtual Assistance over the phone. Its acquisition of Digital Roots brings additional analytics, natural language understanding and machine learning into the mix and applies it to social networks. The next steps involve integration of these formidable new skills across all the channels used by Interactions’ customer base. With Digital Roots’ core technology, it has already demonstrated the ability to add new knowledge domains and verticals quickly. That will be key to future success.



Categories: Conversational Intelligence, Intelligent Assistants, Intelligent Authentication, Articles