How does Intel, which expects the market opportunity for AI hardware to grow from $2.5 billion in 2017 to $10 billion in 2022, find new customer opportunities? With AI, of course. In a blog post today, Intel detailed a tool its IT Advanced Analytics team developed internally to mine millions of public business pages and extract actionable segmentation for both current and potential customers. The chipmaker says that its sales and marketing staff have used the new system to discover new leads faster and more accurately than before.

“Intel sales and marketing staff have traditionally used manual search and vendor tools in order to identify potential leads; however, these methods lack the ability to align with the internal language used by Intel staff to properly segment and tailor their outreach plans,” wrote Intel. “Additionally, in the era of globalized business, existing customers are often expanding into new domains, requiring sales and marketing staff to constantly keep current with changes in a wide variety of industries.”

As Intel explains it, the system focuses on two key classification aspects: (1) an industry segment ranging from verticals such as “healthcare” to more specific fields such as “video analytics” and (2) functional roles like “manufacturer” or “retailer” that further distinguish potential sales and marketing opportunities. The AI model acquires a constant stream of textual data from millions of sites, updating the multi-million node knowledge graph with gigabytes of data every hour, which then gets passed along to a set of machine learning models for segmenting potential customers.

Webpages are fed into a text classification model boosted by a pretrained, multilingual BERT language model to help scale across languages and classes. (BERT, which Google open-sourced in November 2018, enables developers to train a “state-of-the-art” natural language model on data that’s neither classified nor labeled.) Intel enriched the data it used to train the model by crawling tens of thousands of company sites with info from Wikipedia. And for companies without labels, it took advantage of a pre-existing Wikipedia corpus by employing semi-supervised learning, which entails combining a small amount of labeled data with a large amount of unlabeled data during training.

“Our customer segmentation system is only one of the thousands of AI applications that will improve enterprises in the coming years,” wrote Intel. “[We expect to] find new and exciting ways to harness cutting-edge technology to move, store, and process data wherever it is best suited.”

VB TRansform 2020: The AI event for business leaders. San Francisco July 15 - 16

AI-informed lead generation is fast becoming the norm rather than the exception. Roughly 87% of enterprise AI adopters say they’re using (or at least contemplating using) AI and machine learning for sales forecasts and to improve their email marketing, and according to real-time data warehouse company MemSQL, 61% of marketers believe AI is the most important element of their overall data strategy. Moreover, 65% of marketing professionals responding to a recent Salesforce survey said AI-powered insights would “make them more effective in their job.”