At the latest Box hackathon, themed "Hackers of the Caribbean," swashbuckling Boxers got their hands on an internal alpha of the Box Skills Kit. What ensued were prototypes of enterprise-grade solutions that leverage machine learning to automate some of our customers' most critical -- but painful! -- processes.
Since our first hackathon eight years ago, hackathons have been an opportunity for any Boxer -- from engineering, to product management, to sales, legal, finance, and beyond -- to step out of their regular responsibilities for a day and design a fresh way to blow our customers' minds. For our most recent hackathon, Boxers from around the globe submitted over 100 projects, including potential Box features, workflow automations and tracking, and deepened integrations.
Announced at BoxWorks 2017, Box Skills is a framework for bringing machine learning technologies to content stored in Box. The Box Skills Kit is a set of APIs, developer tools, documentation and sample code for building custom skills for Box.
Box Skills opens up a whole world of use cases for enterprise content. Upon ingestion, Skills will analyze documents, images, audio and videos with machine learning algorithms from companies like Google Cloud Platform, IBM Watson, Microsoft Azure and the hundreds of emerging intelligent solutions. Skills translates the output of these algorithms to rich, automatic metadata that is searchable, visible in the Box web application, and can trigger processes, policies, classification, and more.
As we get ready to release the Box Skills Kit in beta, we want to share three of the innovative skills built by Boxers that represent the kind of enterprise solutions we're exploring with Skills.
Our customers' problem: Documents, while core to many business processes, are often hard to find, organize and use. I-9 forms, resumes, purchase orders, Terms of Agreement, BSAs, NDAs -- these documents capture critical information or actions as part of hiring, employee onboarding, vendor management, purchase fulfillment, contract lifecycle management and beyond.
Our Skills hack: In this custom skill, the team leveraged the Ephesoft Transact API to classify the types of documents most common to our organization's workflows -- in this case, recognizing W2s and invoices. The team then determined specific fields that are often needed for later reference or automations -- like invoice number, date and invoice amount -- to be automatically extracted upon upload and added as metadata on the PDF.
What it could mean for enterprises: This custom document skill leverages machine learning to automate manual tagging processes for specific, repeatable processes and automatically structure these documents upon ingestion.
NDA Processing with Box Skills and Nintex
Our customers' problem: Beyond just applying tags to documents, most document-centric workflows involve moving data between systems. Oftentimes, this requires employees to manually review documents, extract the information they need, and then enter it into another system.
Our Skills hack: One example of this is updating CRM records to denote whether or not a customer has signed an NDA. In this custom skill, the team used a combination of a custom NDA document extraction skill, a workflow system, and Salesforce to automate this process. When an NDA document is uploaded to the folder, the skill calls the Ephesoft Transact API to extract some fields from the NDA, including whether or not there is a signature present, and apply those fields as metadata on the NDA. Then, using Nintex, a workflow automation software, the team retrieves the metadata from Box using the Nintex connector for Box, uses the "Company Name" field extracted in the skill, searches Salesforce for a record matching that company name, and when it finds the record, attaches the NDA to the record in Salesforce.
What it could mean for enterprises: This custom skill removes the manual work of classifying NDA documents and automatically attaches the agreement to the relevant record in Salesforce for seamless integration across the IT stack.
Data Loss Prevention Skill
Our customers' problem: Beyond benefits to productivity and processes, we see a whole world of opportunity for Box Skills in the realm of risk mitigation and security. Today, users in Box can apply classifications to files, labeling them as Confidential, Internal or Public, which can then trigger security policies like applying a watermark to the file in Box.
Our Skills hack: For this custom skill, the team leveraged the Google Data Loss Prevention API to analyze files uploaded to a folder, determine if there are any sensitivities about the file and even redact information in that file. In their submission, a sample text file containing social security numbers was automatically classified as "Confidential due to PII," which restricts a users ability to share the file, and the social security numbers were redacted from the file preview in Box.
What it could mean for enterprises: This custom skill helps mitigate the risk of sensitive information going undetected and unprotected in the volumes of unstructured content within an enterprise -- automatically recognizing and redacting sensitive data.
We're excited about the possibilities of Box Skills and the Box Skills Kit as the world of machine learning reaches enterprise readiness. The Box Skills Kit, upon which custom solutions like those above can be built, will be available in beta in a few months and you can sign up to receive more information by filling out the form here.