Box Inc. wants its customers to do more with all the files they’re storing on its content management service. Today, it announced a couple of new tools powered in part by machine learning technology to give customers more ways to extract value from the data — and more reasons to pay Box for extra services.

Box Skills is a software framework that uses machine learning, a set of technologies that allow computers to learn akin to way the brain does rather than being explicitly programmed, to do tasks such as computer vision for image analysis, video indexing and sentiment analysis from audio. The capability was announced at its BoxWorks conference this week in San Francisco, as part of Box’s goal to make Box “the most intelligent cloud content management platform,” as Chief Executive Aaron Levie puts it.

The San Francisco-based company today previewed three skills that are still in development. The first, called Image Intelligence, was actually announced in beta test in mid-August. Using machine learning technology from Google LLC’s cloud platform, it detects objects and concepts in image files (right) and analyzes text with optical character recognition to compile metadata for image catalogs, eliminating the need for manual tagging of receipts, driver’s licenses and the like.

Another skill, Video Intelligence (pictured at top), uses Microsoft Corp.’s Cognitive Services to do transcriptions and detect topics and people to make quick sense of videos without manual viewing. A third one, Audio Intelligence, uses IBM’s Watson AI technology to create text transcripts from audio that can be searched and used in other ways.

That last one, in fact, demonstrates how users can create custom skills that chain together several skills. For instance, one skill could combine Watson’s speech-to-text and natural language understanding services on IBM’s cloud to analyze customer service calls, determining which were angry or satisfied to provide better call-center training. Another skill, developed by document capture Ephesoft Inc., can detect information in, say, a contract and extract it to custom metadata “card” in Box that automates a loan application or onboarding of a new employee.

Box hopes to encourage outside software developers, system integrators and in-house enterprise developers to create their own skills or chains of them. “Any developer can create a custom skill,” Jon Fan, Box’s senior director of product management for enterprise products, said in an interview.

“Since Box is the place where content is stored, it makes for a logical place to derive insights, find patterns, automate repetitive tasks, etc.,” said Alan Lepofsky, vice president and principal analyst at Constellation Research Inc. “Box’s current approach is not to build any AI system of their own, but instead leverage features from Google Cloud Platform, Microsoft Azure, IBM Watson and others. This will allow them to stay ‘AI-neutral.’”

The ultimate goal for Box, said Fan, is to persuade customers that “as you put content into Box, you can get more value out of it than you can anywhere else.” That’s a pitch Box needs to resonate with more customers, since it faces rampant competition from Dropbox Inc., Google Inc. and many others that offer cheap file storage.

In the last quarter, Box has introduced new products such as Box Drive, an unlimited cloud drive for large enterprises, and Box Elements, a set of software tools to embed Box content into applications built on its platform. And in January, it revamped its three-year-old Notes real-time document creation and sharing tool.

New products

The strategy shows some signs of working. New products have been helping to drive more spending by customers. During its second fiscal quarter reported in August, some 60 percent of the 39 six-figure new deals worth over $100,000 that it struck with new customers included new products, according to the company.

Still, Box continues to lose big money in the pursuit of growth. It will need to create an ecosystem something like an enterprise version of Apple Inc.’s App Store if it hopes to veer toward profits at some point.

“Each of these skills does help Box become more of a destination, versus just a place where files are stored,” Lepofsky said. “For example, instead of just storing audio files, now customer support people can have conversations around the areas where the sentiment in the calls was negative. Box Skills should be attractive to developers wanting to build applications on top of Box, as they provide pre-coded functionality, such as image tagging or video transcribing, removing the need to code those features yourself.”

Box hasn’t provided any pricing information for Box Skills. Only the image recognition skill is available in beta test currently, while the others will go into public beta early next year.

The company also showed off Box Graph, which it describes as a network of content, relationships and activities intended to power new experiences for individuals and enterprises. As an example, Box introduced Feed, a personalized activity feed that uses machine learning to surface the most relevant content and insights, such as what content teams are working on and what’s trending within the company: a sort of Facebook news feed for the enterprise.

Box said it expects to provide other applications from Graph, such as search, personalized file and app recommendations and security services such as threat detection.

Images: Box