Apply custom-trained AI models to Box with the Box Skills Kit

Last year, we introduced Box Skills, a framework for applying the world’s best AI from providers like Microsoft Azure, IBM Watson, and Google Cloud to your content in Box. With Box Skills, businesses can apply pre-trained algorithms from these providers to better organize, understand, protect and automate processes around their content in Box at scale. You can read more about Box Skills in our announcement here.

But many enterprise use cases for AI require a level of intelligence beyond the capabilities of these pre-trained algorithms. Oftentimes, businesses need a deeper level of understanding, like labeling proprietary brands or products in marketing images or recognizing unique keywords or phrases in audio transcripts.

Over the past year, many of the leading AI providers in the space have introduced sophisticated tools that enable businesses to create and train customized AI models, like a custom computer vision model that recognizes their products in images. These tools enable you to train and tailor pre-trained algorithms to solve for your business’ unique data and processes and even create entirely new algorithms.

Today, we’re excited to announce support for custom AI models, created using services like IBM Watson Studio, Google Cloud AutoML, Microsoft Azure Custom Vision, AWS SageMaker and others, via the Box Skills Kit. With this support, you’ll be able to integrate custom-trained AI models, tailored to your specific needs, and apply them to recognize and understand your unique business content in Box with a custom skill. This allows you to search your content based on terms and concepts, as opposed to based on file names or more specific data. These metadata values, applied via the custom skill, can also drive other Box functionality like automations and governance policies.

Box’s search results will display rich media results for content that has been enhanced by Box Skills

The ability to more deeply understand and extract information from business-specific data unlocks new use cases for Box Skills, such as:

  • A construction equipment rental company being able to apply customized image labeling services to automatically recognize and label specific equipment and part numbers in images captured and uploaded by field sales engineers
  • A media & entertainment company being able to recognize and label specific brand IP, like animated characters, in image, audio, and video content to make marketing assets more structured and easily searchable
  • A financial services company being able to automatically recognize and classify different types of forms uploaded as part of a client onboarding or loan origination process

Starting today, you can apply these customized AI models to content in Box by building a custom skill using the Box Skills Kit, which will be generally available in December 2018.

You can visit our website to learn more about Box Skills and request access to the beta access for the Box Skills Kit on our developer site.