Today, we showcased three of the initial skills on stage at BoxWorks:
Image Intelligence, in beta since August, adds structure to images, the second most common file type in Box. Upon upload, images are automatically contextualized with object detection, handwriting and text recognition. These insights are added to the content as metadata, visible alongside the image in Box Preview and indexed for search.
Audio Intelligence renders a text transcript and detects topics that can be easily searched throughout Box to find an audio file by its specific words or topics.
Video Intelligence provides transcription, topic detection and facial recognition for one of the fastest growing files types in Box. Not only does this improve discoverability of a video by its content, but it allows users to navigate through video files in a non-linear, efficient way according to the information or people featured in each frame.
With Skills, individual pieces of content will become agents to accelerate your business processes, with automatic, intelligent metadata that can trigger workflows, establish retention policies, apply classification and so much more.
Box has done the heavy lifting to bring intelligence to your content, all while ensuring you maintain control of exactly what happens to your data per Box's security and compliance standards. But we didn't stop there.
While these initial skills are applicable to any business in any industry, we wanted to extend the Skills framework so our customers can leverage machine learning to solve their unique business challenges.
That's why we developed the , a collection of tools to allow independent software vendors, systems integrators and enterprise developers to create custom Box Skills. Box Skills Kit
Skills Kit empowers an ecosystem of providers and customers to BYOML (Bring Your Own Machine Learning) to Box. These tools allow developers to train or group existing skills, or develop entirely new capabilities to intelligently process content in Box. At BoxWorks, we showcased several potential custom skills, including:
A skill to process call center recordings at scale, allowing quality assurance teams to mine a database of call recordings according to customer sentiment or topics discussed. This skill grouped IBM’s Watson Speech to Text and Natural Language Understanding services.
A skill to analyze forms, like contracts managed by a legal team or invoices by an accounting department. This example skill extracts specific information according to an organization's particular needs, allowing processes to be kicked off from that company's specified metadata. This skill was developed with Ephesoft, a machine learning startup.
A specialized image recognition skill that uses a company’s proprietary data stored in Box to train its own machine learning model for recognizing images of its own products.
And while Box Skills allows you to extract value from your content one file at a time, we've been working hard to develop a way for you to better understand your content at scale, giving you insight into the relationships between your content and your people.
We're thrilled to introduce Box Graph, our own machine learning model that actively interprets and then maps how your organization works.
Box Graph understands the relationships between content and other content, content and people, and people and people. Graph continuously learns from everything you do in Box: the shared links you generate, the colleagues and customers you collaborate with and the Box Notes you create to take meeting notes. It also learns from activity on other business applications integrated with Box, like a file shared on Slack or an updated Office document.
Graph is a living and improving model under the surface of your Box experience, which we leverage to power predictive, personalized and contextualized experiences across Box.
The first feature powered by Graph is Box Feed, a personalized activity feed that curates and surfaces the most relevant updates, insights and content for each Box user. With Feed, every employee will be able to see the content that's most important to them, in a single view. All while respecting existing privacy and permission settings.
This comes to life in 3 ways:
Surfaces active content you're working on. This could be a document you've been @mentioned in, a presentation you’ve commented on, or a Product Spec you've contributed to that has been significantly updated since you last reviewed it.
Recommends relevant content from people you work with closely. Imagine if you could get access to the top performing sales pitch if you're in Sales. With recommended content, you're served the most relevant files from people you're connected with, helping you get your work done fast.
See what content is currently popular within your company. With trending content, you can easily discover what content has been viewed, shared and downloaded the most, even if you're not connected to the person posting it.
We're so excited to share these intelligent technologies with our customers and watch as they use them to fundamentally change the way their businesses work. With Box Graph, organizations can turn the problem of exploding content volume into an opportunity: now, the more content you have in Box, the more you can do with it. And Box Skills helps you make sense of it all.
Box Skills and Skills Kit will be in beta in early 2018. Customers can sign up for the beta here, and partners or developers who want to build on the framework can sign up here.