Automating Redaction Compliance to Improve Workflow Efficiency in Foreclosure & Bankruptcy Cases


Victor Diloreto

Date Published:
May 24, 2022


At Blackmarker we set out to introduce AI/ML to the task of redaction compliance, and in a sense provide a new age black marker. Using artificial intelligence/machine learning (AI/ML) to automate redaction compliance dramatically increases the accuracy of redaction compliance and reduces expenses.

The AI models target information within documents that are required for redaction when submitted to the courts. The information can include personal identifiable information, like names and social security numbers, but can also include loan numbers, stamps, and bar codes that could be traced back to an individual.

As you can see from this graphic, the introduction of AI/ML to redaction compliance can dramatically improve time/labor and accuracy.

We have seen how AI/ML solutions can impact law firms specializing in foreclosure & bankruptcy case preparation, and from this we have derived some best practices.

This article highlights the benefits of automation and reveals how to realize the full potential of automated redaction compliance. It will cover:

    • Challenges of AI/ML Introduction to Workflows
    • Benefits of Redaction Compliance Automation
    • Distributed versus Centralized Workflows
    • Best Workflows Using Compliance Automation

Workflow/Process Challenges

Introducing this technology to an enterprise, and the process automation it provides, could require a re-think of the roles and responsibilities of those tasked with redaction compliance.

It is like the scene in Hidden Figures, when Dorothy Vaughn (played by Octavia Spencer), sees the new IBM mainframe for the first time, and realizes that the current processes for factoring are going to change, and adaptation will be necessary. She picks up a Fortran computer programming book, as she quickly realizes that knowing how to run the machine is a better route than ignoring (or fighting) it.

Benefits of Accurate Redaction Compliance Automation

Time Savings

Money Savings

Better Compliance

Happy Employees

By creating a highly accurate and automated redaction compliance capability in software, an enterprise can relegate the function to a batch, or centralized process. A defining characteristic of batch processing is a lack of human intervention, with few, if any, manual processes.

This was a goal we sought for Blackmarker – offloading redaction to a machine, saving money and time.

Saving time and money will always be a worthwhile goal, but there is an additional benefit in that automated redaction compliance makes an enterprise more compliant. In sampling public court records, we found a 25% error rate for documents prepared manually.  Our automated process now approaches a 0% error rate.

This is where the Dorothy Vaughn effect is applied, as an AI/ML-based system must be trained. In the context of redaction, this means having the ability to mark documents, typically PDFs, by experienced people aware of the compliance rules/policies of the enterprise. In other words, the system must be “programmed” – not with rules but with examples. This does not require someone to learn Fortran — or any other software programming language — but to work within the system to mark the information for compliance against a finite sample of documents for the AI/ML to learn what targets are required. Targets in this context can be personal identifiable information, personal health information, or other sensitive information as defined by the policy being followed. Once that finite sample of documents is marked, and the performance of the system on new documents is at a level that is equal to, or exceeds that of an experienced human, the system can process similar documents unattended. It should be noted that AI/ML systems, like Blackmarker, maintain multiple models which are pre-trained for certain types of personal information specific to the foreclosure/bankruptcy field.

Another benefit of automating redaction compliance is that while the number of people involved in physical redaction is reduced, the smaller team tasked with training or “programming” the machine reduces the chance of misinterpreting the policies, which can lead to compliance escapes. In the end, the system has fewer moving parts, which translates to fewer mistakes.

Lastly, in our current times, we have seen a lot of churn in the job markets, induced by the pandemic and the other factors it brought on – remote work, tighter job markets, etc. Introducing this type of automation removes a task that nobody, by survey, liked in the first place. Thus, enterprises can market higher satisfaction roles to the job marketplace with a better chance of landed long-term employees.

Distributed versus Centralized Redaction Compliance Workflows

Organizationally, we see workflows that have a compliance redaction functions within them falling into one of two camps.

Distributed Workflow

Redaction compliance, among other duties, is distributed to a group of individuals. Each person owns a set of cases and all the tasks required to prepare the files for case submission. The task list includes redaction compliance, and tasks like unbundling large title search documents, adding motions, adding stamps, etc. The various steps are executed serially, one after the other, by each person, case by case.


Centralized Workflow

In the other setup, certain parts of case preparation, like redaction compliance, are carried out by a single group. They redact the files to be used by others for case preparation.

Optimal Workflows Using AI/ML

Given AI/ML’s automation power, it naturally facilitates the ability to batch process, which creates the greatest efficiency. A centralized workflow more closely aligns with batch processing; thus, it is the recommended type of workflow when using AI/ML. In this fashion files can be presented to the AI/ML system by those responsible for redaction compliance or via case tools/document management systems overnight or early mornings, and then the balance of the workforce can access the redacted files as they prepare their cases for the day/week.

The centralized role for redaction compliance sees all the benefits laid out previously in this article:

Greatest time/cost savings as we have automated the task.

Better staff retention as work satisfaction increases and the fatigue factor is eliminated.

Better compliance as we have fewer people training the machine, reducing the chance of policy misinterpretation.

With a state-of-the-art technology, employees are empowered to make a more meaningful impact on the business.

If a firm has deployed a distributed workflow for redaction compliance an operational decision can be made to move toward a centralized function for this task. Change is hard, but there are very real, tangible reasons to enact this type of change. A potential blueprint for this type of change is outlined below:

Assess the current workflow.

Understand and look for points in the flow where documents could be pre-prepared for employees to access in case preparation. Aim to understand when documents would need to be prepared and the points in time where they could be batched (overnight, early AM, periodically throughout the day, etc.).

Knowing the points in the workflow to have documents ready can be facilitated by a case management or document management system.

AI/ML systems can integrate with these systems as most AI/ML solutions support an API (application programming interface) for this purpose. Blackmarker already supports several of these systems (most commercially available and some internally developed by firms).

Once the roadmap for how/when documents need to be prepared in a new workflow is established, the next step is identifying your personnel for the balance of the work.

First, identify a new, core team of people to manage and “program” the AI/ML solution. Our experience shows this will be a small team designed to cover the unique documents being prepared, and team size is related to overall volume of documents being processed and their variability.

This team is responsible for enforcing the firm’s policies for redaction compliance and interacts with the AI/ML to ensure the performance of the system adheres to these policies.

That process involves marking documents of a representative sample size for the redaction needed when a policy changes, or a new policy is created. Note, that AI/ML systems like Blackmarker come with pre-trained models specific to industries, like foreclosure and bankruptcy.

The balance of the team can be re-organized around the tasks that remain, and as you see utilization needs diminish due to the automation, you can move personnel into other areas where they can make an impact.


Redaction compliance has been a manual exercise for a long time, but AI/ML based solutions, like Blackmarker, offer real benefits. These benefits are best realized when the organization is prepared to adopt it. Identifying the core team to “program” the AI/ML and allowing others to benefit from the prepared documents is the best practice and is achievable regardless of how you are organized today.

Want to hear more about how Blackmarker
can improve your redaction workflow?

Get in Touch