How to Get Started with AI-Led Automation for Document Processing
4 min read
This post breaks down the initial planning and assessment process for applying AI-led automation to document processing within your operations.
To learn more, jump to the following link to register for the upcoming webcast “How AI and Automation Tools Enable Automated Document Processing for All,” featuring guest speaker and Forrester Principal analyst, Cheryl McKinnon.
In an earlier post called “Four Benefits of Applying AI-Led Automation to Document Processing,” we introduced IBM Automation Document Processing for the IBM Cloud Pak® for Automation. That post leads to a follow-on question: How can you get started with applying these new tools?
To tackle this topic, I’ll break it down into a basic three-step process:
- Discover and prioritize
- Evaluate existing resources
- Build test and prove the solution
Let’s get started.
Discover and prioritize
The discovery process needs to focus on what business priorities you want to address, including the associated tasks, uses cases, or processes you want to focus on first. These could be defined by a range of factors that ultimately answer one basic question: What is your most critical business need?
Leading factors for consideration should include the following:
- Strategic urgency: Do you need to find low-risk wins to help combat the threat of new disruptive innovators or digitally savvy competitors?
- Potential return on investment: Do you need to see quick incremental wins to help grow digital transformation projects?
- Staffing constraints, including skill gaps: Do you need to apply automation to increase productivity while creating more sustainable workloads for your employees?
- Risks associated with spiking customer demand: Are you concerned about future spikes in information-related customer requests?
- Scale and scope of deployment: Do you want to deploy the solution across your entire enterprise?
Evaluate existing resources
Depending on your organization’s priorities, you’ll next want to consider and evaluate how an AI-led automation solution will fit into your current document processing operations. There are two common scenarios to consider.
Scenario 1: No document processing solution, all manual document processing today
For this scenario, you will need to decide if you want a complete solution that provides robust no-code document modeling for classification and data extraction, in addition to integrated information verification. Or will you need a hosted service that supports custom software that your organization will write and maintain? Or do you need both?
Scenario 2: An existing OCR solution, but still lots of manual document processing
If you have an existing document processing solution that uses OCR and zonal templates, can that solution support the use cases and processes that you identified as priorities? If not, is it still productive and worth keeping in place for existing processes? If you answered “no” to the first question and “yes” to the second, you have some options.
For an organization with an existing document processing operation, some AI-led automation tools can be deployed alongside and even integrated with the existing processing workflows. The artificial intelligence solution can expand the range of supported documents for automated processing while also allowing front-line document submitters to quickly verify information in real-time.
Build, test, and prove the solution
The next set of steps will help to shorten the time required to thoroughly evaluate, test, and prove that a given solution will adequately meet your document processing needs:
- Based on your leading business priorities, choose the use case(s) and the corresponding document types for use in a proof of concept. This information will help qualify your requirements relative to the capabilities of a given AI-automation solution. Consider which types of documents would benefit from automated processing with little or no human intervention.
- Gather sample documents for training and testing. For an AI-led solution, you will need a small set of document samples. Systems that use deep learning will be able to recognize and differentiate the same types of information across different documents, unlike solutions that rely on only OCR and location-based zonal recognition.
- Engage with a solution provider like IBM Services or one of IBM’s certified Partners to document your goals, objectives, and priorities for consideration in building the proof of concept. An experienced solution provider can provide consultative expertise regarding how to optimally deploy an AI-led automation solution in either of the scenarios identified earlier.
- Work with the solution provider to configure and test the proof-of-concept solution in as close to a production-like scenario as possible, then measure the cumulative results. In addition to document classification and data-extraction accuracy, consider the total time, speed, and ease of system training required to build document models with the AI tools. Look for no-code options that allow business users to quickly add and train new documents in the system. Tools that leverage deep learning have the ability to transfer learned information types between different document types, simplifying and speeding the training process. In combination, these capabilities can reduce the total cost of maintaining the solution, reducing or even eliminating demand for highly skilled document experts and data scientists. Other features to look for include a complete human-in-the-loop data verification workflow and automatic error correction that contribute to produce high accuracy and total throughput.
Register for the upcoming webcast
However you decide to proceed, you are now on a viable path to getting started on your journey to applying AI-led automation for document processing in your enterprise. To learn more on this topic, you can register for the upcoming webcast “How AI and Automation Tools Enable Automated Document Processing for All” on December 2, 2020, at 11:00 AM EST, featuring guest speaker and Forrester Principle Analyst Cheryl McKinnon and Director of IBM Content Services Eileen Lowry.