A businessman in a suit, running across the finish line
PHOTO: Shutterstock

This is part 2, in a 4 part article series sponsored by Alfresco.

In our previous article in this series, we concluded by saying that you should start small and begin to leverage artificial intelligence (AI) now rather than wait. It’s not rocket science, and it is not as expensive or complex as you might imagine. In this article, we will get practical and help you to consider exactly how and where you should target your first use of AI for content and process management.

AI can be applicable across a wide variety of use cases, some of which are common across industries, such as automating of retention management or customer segmentation. Other use cases are industry- or process-specific like claims processing or customer on-boarding and management. But all of these use cases, no matter how generic or specific, can be truly transformational to your operations.

As Bernadette Nixon, the CEO of Alfresco, recently said in an interview with Deep Analysis, “If you take a look at how you can apply AI in the process arena, you can leapfrog and get much more than process optimization. You can even get AI to design some of these processes.” At Deep Analysis, we agree that AI is a tool to leapfrog your transformation efforts. But true transformation will usually come through a series of baby steps involving some quick wins, rather than a single big bang project. So, if you are looking to use AI for transformation in your organization, we suggest that you take the following approach to getting started.

Identify an Opportunity

Don’t boil the ocean and get overly ambitious; instead, you should look for an opportunity that would represent a quick and visible win for your organization. For some people, a great starting point may be auto-categorizing some documents, and for others it may be identifying risk or increasing the speed and accuracy of a business process.

Quick wins should be low risk, narrow and focused in scope, as well as inexpensive. They should also have a high level of impact if your project is successful, for as the saying goes, success breeds success. We recommend you brainstorm potential quick wins with your team and make sure you identify one that all team members understand the potential benefits of and are enthusiastic about.

Related Article: How to Rethink AI

Categorize the Opportunity

There are many ways to categorize your opportunity, but we find the best starting point for content and process management is to decide if your opportunity falls into one of the three buckets listed here.

Corpus AI - Most established firms have millions (or, in some cases, billions) of stored historic documents sitting in legacy systems and document repositories. Few have any real insight into what is in these documents or what value (or risk) that these documents carry. AI systems can be trained to analyze this corpus of data for legal discovery and identify risks, duplicates and redundancy, as well as improve basic access to large legacy repositories of files.

Active File AI - Active files and documents in, for example, an ECM system can leverage AI to ensure that intelligent capture, document classification, summarization, and insight are applied, whether to improve capture efficiencies or to simply apply accurate classifications to meet increasing regulatory oversight needs.

Activity AI - Though a focus on activity has been primarily the concern of web content and customer experience (CX), that is beginning to change. There is a growing interest in analyzing who engaged with content and how they engaged with the content and when. For example, M&A (merger and acquisition) processes can be protracted and document-heavy. Understanding the patterns of activity across multiple deals can give an insight into how investor A behaves (including their interests and business processes) versus investor B.

Clearly, some of the uses we touch on here are more ambitious than others. We recommend that a good place to start is with some kind of auto-classification to identify risk, compliance and redundancy or to improve on current process efficiencies.

Identify the Data

Identify the data you will need and its locations. AI needs data to function, but it doesn’t necessarily need a lot of data. What it does need is the right data. It needs that data both to learn from and to actually process your chosen task when it is up and running to extract insights and drive automation. If, for example, you are thinking of using AI to reduce your storage costs and identify records, files and documents for deletion, then you will need both permission and access to those records.

This may seem obvious, but gaining permission and access to the relevant data is often more of an internal hurdle than many expect. Often issues of privacy, ownership or security arise, issues that may have to be addressed before you can get your hands on the data you need for your AI project. Once you have access to the data, you may need to clean it up and sort through some of it to get a better understanding of what you will be working with. If you have targeted your quick win well, then this step should be straightforward; if not, it can be a showstopper.

Choose Your AI Tool and Approach

At its simplest level, AI can extract data quickly and then use that data “intelligently” to transform a task or an operation. Different tasks will typically require different AI models, systems and approaches to do each job. For example, the same AI that you would use for sentiment analysis of social digital media is unlikely to be the same AI you would use to classify documents as records for deletion. There are many different techniques and methods used in AI tools, from decision trees and random forests to natural language processing and neural networks. But thankfully there are many AI tools and systems available today that have been built and, in some cases, pre-trained to meet specific opportunity requirements.

For instance, Alfresco now offers the Alfresco Intelligence Services module for its Digital Business Platform (leveraging AWS Comprehend, Rekognition and Textract). This new module provides enterprises with a method for intelligently extracting important content often locked away in multiple documents, scanned images, videos and photographs. Alfresco is seeing demand for its Intelligence Services for two use cases, Claims Service Automation and Citizen Services. Claims Services Automation helps insurance companies improve file intake and orchestrates claims processes in a secure and regulatory-compliant manner. Alfresco’s Citizen Services capability enables government agencies to receive and manage information and service requests from citizens. It includes content management, process design and orchestration, as well as governance.

Conclusion

Identify a quick win opportunity for AI. Categorize that opportunity as one that will determine your approach and the data you need. Locate and get access to the data, then look for a suitable AI tool to meet your needs. This may seem pretty logical, but all too often this sequence happens in reverse, with organizations getting excited about an AI tool and then looking for a place to use it. That seldom works out well.

Remember that AI can find and extract value from your content, and use that to improve your business operations and processes. Every AI project, no matter how small, should be driven by your specific business requirements, so find small and manageable opportunities, quick wins that will gain great visibility within your organization, and start there.

There is a bit more to it than these four simple steps, but a quick win should always be your starting point. To quote Bernadette Nixon again, “Get quick wins along the way. They are the stepping stones to your grand vision.”