For some time now, industry prognosticators have hyped Artificial Intelligence (AI) and Machine Learning for their transformational potential. The possible use cases for AI are so broad, and their potential value to the enterprise so significant, it’s easy to dream the day away imagining AI-powered robots managing all of our challenging (and tedious) tasks.
Day-dreaming aside, we’re still waiting for the full impact of AI for content-centric business applications. To get a sense of how AI is driving the evolution of content management, it helps to understand where things stand now.
The State of Content Management in the Enterprise
Information is everywhere. Organizations are drowning in content, which is often stored across multiple systems that don’t talk to others. Research validates the pain organizations are feeling. Nuxeo recently surveyed more than 1,000 Sales, Marketing and Creative professionals in a variety of industries across the US and UK to learn more about their content challenges.
The survey results showed almost 75% wasted time recreating content they knew existed, but couldn’t find. More than half admit going rogue and using their own systems to store content because the information they needed was too difficult to find, or they found company-sanctioned tools too difficult to use. With stats like these, it’s hard to ignore the promise AI holds for making it easier for people to find the information they need to do their jobs.
The good news is that AI is beginning to be leveraged in meaningful and valuable ways to improve and enhance content management.
Metadata – the “information about information” – is a transformational area for content management. In early-generation document management and enterprise content management (ECM) systems, each stored document became the focal point for invoice processing, claims management, and other enterprise processes. Every one of those stored documents contained a small set of metadata attributes, or tags, typically limited to information such as filename, date created, author, and type of content. For most systems, once the metadata schemas were defined, they usually remained untouched because changes required tedious development work and mass updates to all content related to that metadata.
In a modern Content Services Platform (CSP), on the other hand, metadata schemas are both flexible and extensible. This is massively powerful when combined with AI, which can dramatically accelerate the creation and classification of metadata attributes. For example, let’s say you have an existing ECM repository containing customer agreements. These contracts are poorly managed, and the only relevant metadata attributes associated with these documents are customer reference numbers. By using an AI-infused CSP, an enterprise can identify and extract critical attributes at scale, such as project, customer contact, term period, etc. With this type of automation, additional security controls and provisions per privacy policies or regulations can also be enforced more expeditiously.
Amazon, Google, and other companies have built advanced AI engines, but these tools are based on publicly available data sets, which means they can’t deliver results specific to a business. Modern CSPs enable enterprises to leverage and create custom AI models based on their own data sets, which deliver attributes that are much more specific to the business.
Imagine you show a picture of a truck to one of the generic AI engines. The system recognizes that the image is a truck; it’s got four wheels, it’s blue, and it’s a Ford that is parked by a building. The AI will do a reasonable job of categorizing and classifying that – interesting, but not all that useful.
If you’re Ford, you want to know more Ford-centric specifics. For example: what model of truck is it? What is the exact type of alloy wheels are on that truck? What is the specific paint code of that blue? This is the type of information needed for truly domain- and business-specific intelligence and automation.
Classifying Content within Legacy Systems
As I noted above, publicly-available AI tools have proven valuable in identifying basic and generic content attributes, such as the difference between a contract and a resume. However, AI models based on data and content specific to an organization can be immensely more valuable. So, for example, if your business needs to know the difference between a personal life insurance document and a life annuity document and automatically apply the right contract language from your legal team, this can be incorporated into a specifically trained AI model, which in turn will deliver a much more detailed classification than could ever be possible with generic AI.
Modern CSPs with such AI capabilities enable organizations apply this to the mass of content stored in any connected systems or repositories. By using AI-classification, it is possible to quickly and accurately identify and organize different types of content, even at the scale of billions of pieces of content.
Read More: AI Does Not Have to Be a Zero-Sum Game
Only a Glimpse of What’s to Come
We see that infusing content platforms with contextual AI makes it easier to find content by automating the creation of high-quality metadata. This powerful combination of capabilities – the modern CSP and AI – enhances how users can identify and organize different types of content – and does so at a scale never before possible. The effect is a radically increased utilization of metadata within the enterprise – an enhanced capability of incredible value to today’s modern enterprise.
So, while self-driving cars and flying robots may be years off, the future of AI for content management is already here. It may be much more practical, but the possibilities the technology creates for business will create new opportunities we have yet to realize.