Use AI To Tame Unstructured Content

The technology world is really buzzing about Artificial Intelligence or AI. Many of the big tech companies are pouring ridiculous amounts of cash into AI technology. The big question from a business standpoint is “how will AI benefit my organization?”

It can be hard to envision how ChatGPT or any of the other conversational AI tools will help your organization sell one more widget or win one more contract. For most organizations, however, there is an obvious opportunity for AI to play a role right now… today. That opportunity lies in “unstructured content” or what has traditionally been “document management.”

Unstructured content is a broad term that includes document management and other “things” that are important to an organization that often don’t fit neatly into prepackaged ERP software. Maybe your organization manages pipelines or satellites. There may be inspections, CAD drawings, technical specifications, issues, vendors, and specialized service providers all related to those “things.” Keeping track of the “things” and all the “stuff” that goes with them can become very chaotic very quickly.

AI can help you make sense of the chaotic unstructured content in your organization. It can help tell what a document is, what’s in it and even how it applies to your unique business processes. The trick to figuring out how AI can benefit your organization is to understand the current AI capabilities and how you may be able to leverage them.

In the unstructured content space, one of the clear leaders in AI development is M-Files. They have been building AI tools for years and have those tools ready for your organization to leverage. Each AI tool has a distinct purpose such as classifying content or looking for certain types of information inside documents and each tool can be implemented independently or in combination.

The AI that’s making headlines is quite different.  It uses libraries of data that humans pour through and “label” to provide context to the data. The human input guides the tool as it goes through processes of supervised learning with sample questions and more human feedback that is used for ranking best to worst outputs. The results of that process become feedback to further train the model. Those results are reinforced in reward models of up votes or down votes to refine the AI model even more.

It’s all very complex and time consuming. It’s still new and there aren’t many direct business applications (yet). By comparison, the AI models implemented for document management and unstructured content from M-Files are much easier to implement and much less time consuming. This type of AI is a rule-based implementation that enables a computer algorithm to mimic how people would interact with unstructured content.

To understand how these tools work, imagine it’s your first day on the job and you get an e-mail with an attachment. You open the attachment and across the top are the words TPS REPORT. Even though it’s your first day you would be relatively sure this attachment is a thing called a “TPS Report” which probably has importance to the organization. You may not know what a TPS Report actually is with respect to the business and you probably don’t know any processes associated with the report, but at least you know what it is. In effect, you have a rule for identifying a document but you lack rules to know what to do with it.

The AI in M-Files attempts to mimic the way a new employee learns about a business. It uses rules that are set up by the organization to help analyze and provide context for the TPS Report. Unlike an employee that comes with an enormous amount of life experience and intelligence, the AI comes with only a few built-in rules such as “this is what phone numbers look like” and “this is what social security numbers look like.” It doesn’t know what a TPS Report is. It needs to be given those rules by people within the business.  Even with just a few rules in place, AI tools can begin to ease the daily burden of dealing with unstructured content.

AI, like a new employee starts out with a few rules and builds on them over time. As the number of rules increases the tools become more capable. Weighted results prioritize rules and control the decisions the AI model makes with levels of certainty applied to the decision. For example, the rules can have thresholds where it only acts on things that look like TPS reports when the AI engine is at least 80% certain it’s a TPS Report.

In the end, the AI in M-Files won’t be able to write you a song about daisies or get into a discussion about why it’s not wise to tame a chipmunk like ChatGPT, but it can certainly help tame the chaos that often surrounds unstructured content within an organization. The use of AI in your organization doesn’t need to be years off into the future. You can start to realize returns from AI today with technology that already exist in M-Files.  You can start with a single process that is currently causing grief within the organization.  Contact TEAM IM to find out how AI can help your organization today.