Operational Mortgage Automation

What it is, how it works, and how to use it in mortgage operations
Vova Pylypchatin
Founder and CTO @ MortgageFlow

Hey there,

Welcome to the new subscribers who joined our mortgage tech newsletter last week.

Today, let’s talk about operational mortgage automation.

Everyone has some sort of understanding of what automation is.

But automation is such a loaded term that it is often too abstract to be useful.

In this issue, I’d love to share a more practical overview of operational mortgage automation, how it works, and its applications.

I hope this will help you identify more opportunities to apply it in your mortgage operations.

Below, you can find my analysis of:

  • What operational mortgage automation is and is not
  • How it relates to operational data, analytics, and apps
  • How operational automation is different from the traditional
  • How operational mortgage automation works
  • How it can be applied to improve mortgage operations

What’s operational mortgage automation

Let's start by unpacking this loaded term.

Automation refers to using software to perform tasks that would otherwise require human intervention.

Operational automation refers to using software to automate routine and repetitive tasks in business operations.

Operational mortgage automation refers to applying automation specifically to the operational aspects of mortgage lending.

It focuses on operational processes like loan processing, underwriting, servicing, and default management.

The goal is to increase efficiency, reduce human error, save time, and lower costs while ensuring compliance with regulatory standards.

What isn’t operational mortgage automation

There are many tech and software systems used in mortgage operations that might seem like operational automation but aren't.

For clarity, let's define what operational automation is not.

Operational automation is about performing actions.

Thus, if software doesn't actively reduce the actions a human needs to take, it's not considered operational automation.

For example, a document storage system like Dropbox or an analytics dashboard isn't operational automation.

I even lean towards categorizing AI products that help extract data from documents into operational data engineering rather than operational automation. It definitely can be a step within automation. But until the software system takes any actions based on this data, it is more data transformation software.

How operational mortgage automation relates to Data, Analytics, and Apps

Operational Mortgage Automation is closely intertwined with operational data, analytics, and apps. Understanding how they relate to each other might help you in your mortgage automation efforts.

Automation and Operational Data

Data is the foundation of automation. Without operational data, there is no operational automation.

Thus, most automation projects start by first collecting the required operational data.

Usually, changes in operational data trigger automation processes. Automation then utilizes this operational data to perform actions.

Automation and Operational Analytics

Like automation, operational analytics relies on the same operational data. But unlike automation, the primary function of analytics is to answer questions based on the data. While for automation, it is to perform action based on the data.

Analytics and automation go hand in hand in the improvement of processes.

Here are 2 use cases of operational analytics in the context of automation:

  1. Identify processes ripe for automation. Operational data often informs automation. By examining operational data, you can pinpoint inefficiencies or errors within the processes.
  2. Evaluate the impact of automation. After implementing automation, operational analytics can be used to understand its effect on the performance of operations.

Automation and Operational Apps

The terms automation and software applications are often used interchangeably. While automation is part of most applications, they aren’t the same.

Applications like automation and analytics rely on the same operational data to function.

But the primary function of the applications is to:

  • Manage operational data used by automation → Create/view/update loan applications.
  • Trigger automation → Click the button “Run AUS.”
  • See the output of the automation → See AUS findings.
  • Configure automation → Provide settings to the AUS

Most automations are invisible to us. They work under the hood, triggered by changes in operational data, take actions via APIs, and store output in applications.

Applications serve as a user interface to interact with automation and operational data.

How operational mortgage automation is different from traditional mortgage automation

A few traits in operational mortgage automation make it highly practical in mortgage operations.

While I don't think there is a rigid line between operational and traditional automation, contrasting them can highlight these traits more clearly.

Here's my analysis of the differences between the two.

Focus and application

Operational mortgage automation focuses explicitly on the operational aspects of mortgage lending. It deals with automating actions in processes like loan origination, underwriting, servicing, and default management.

Traditional mortgage automation focuses on broader aspects of mortgage lending, including marketing, sales, and customer support.

Scope of automation

Operational mortgage automation is process-oriented, focusing on automating multiple actions that compose an entire process, allowing data to flow from one step to another.

Traditional mortgage automation, however, tends to automate narrow, isolated tasks within a process.

Data integration

Operational mortgage automation relies on the operational data automatically pulled from various systems like LOS, CRM, data vendors, etc.

Traditional mortgage automation relies on manually inputting data into the system by humans.

Automation triggers

Operational Mortgage Automation is triggered in response to real-time changes in operational data, operating without manual intervention.

Traditional Mortgage Automation relies more on a human initiating the automation. Automation may carry out tasks autonomously, but it only starts when a person activates it.

How operational mortgage automation works

An automation system consists of a set of automated workflows.

These automated workflows can be distilled down to four key concepts:

  • Triggers
  • Steps
  • Actions
  • Runs

A workflow consists of a trigger and one or more multiple steps. Each Step represents an action that needs to be taken. Events in operational data trigger workflows. Workflows generate a Run for every time it is triggered. A Run is a single execution of the Workflow.

Below is a deeper overview of each concept.

Triggers

Triggers define WHEN a workflow should be run.

A Trigger acts like a sensor, monitoring changes in operational data and initiating an automated workflow when specific conditions are met.

Triggers can be categorized into:

  • Date-based → Observes current date and checks if it matches conditions
  • Event-based → Observes events and initiates a workflow if a particular event occurs
  • Entity-based → Observes changes in entities and triggers a workflow upon a specified change.

Here are a few examples of the triggers:

  • (Event) When an application is taken.
  • (Event) When an application is submitted to underwriting for the 4th time.
  • (Date) When the rate lock expires in 5 days.
  • (Date) On the 2nd Friday of the month.
  • (Entity) When the application status changes from 'Pending' to 'Approved.'
  • (Entity) When the loan type is specified as 'Fixed Rate' instead of 'Adjustable Rate.'

Triggers ensure that workflows are executed when they are supposed to, including in response to real-time changes in operational data.

Steps

Steps define WHAT actions should be taken when a Trigger initiates a workflow.

Think of steps as the blueprint for the workflow. A workflow usually consists of multiple steps, although it can sometimes be a single step.

Each step is defined by:

  • Action type → What action to perform (e.g., send an email).
  • Action sequence → In what sequence (e.g., after the 3rd step).
  • Action inputs → What data to use (/e.g., email recipient and subject line).

Here’s an example of the steps in a workflow:

  1. (Send email) to [borrower]
  2. (Send email) to [loan officer]
  3. (Send email) to [ops manager]
  4. (Update application status in LOS) on [Application taken]
  5. (Create calendar event) in [5 days]

“1.” defines the order of the action, “( )” is an action, and “[ ]” is a parameter of the action.

A workflow can consist of multiple steps, executing the same action with different parameters, like the first 3 steps in the example above.

Actions

Actions define what staff your automation CAN DO. It is like the skills of your automation system.

Typically, actions involve running computations on the supplied data or interacting with other software systems via APIs.

Here are some examples of actions an automation system might perform:

  • Sending an email.
  • Validating a document.
  • Moving money between accounts.
  • Performing a calculation.
  • Generating a PDF document.
  • Creating a loan application in a Loan Origination System (LOS).
  • Placing an order for an appraisal.
  • Executing tasks within LOS.
  • Etc.

An action is the component that generates the core value of an automated workflow.

The more actions your automated systems can do, the more processes you can automate.

Each automated action is one less action that needs to be taken by a human.

Runs

A Run is a single execution of the workflow.

Anytime your workflow runs, the software system executes each workflow step in order.

The workflow run ends when all the tasks have been executed, and any final outputs have been produced.

Single Run is usually defined by the following:

  • Status (Waiting, Executing, Canceled)
  • Workflow
  • Completed steps and their output
  • Current step

How to use operational automation to improve mortgage operations

The primary function of automation in mortgage operations is to perform actions that would otherwise require human involvement.

So, what defines how operational automation can be applied to mortgage operations is WHAT ACTIONS automation can get done.

Below is a list of common actions that can be automated in the mortgage process. The list is not exhaustive but should give you an idea of what's possible.

As technology advances, the range of actions that can be automated expands, especially with the latest developments in AI technology. More and more tasks that traditionally required human intervention can now be effectively handled by software.

Data entry and actions within applications

Given enough data, most routine actions taken within software applications can be automated through API or RPA.

Here are some examples of actions within this use case:

  • Import application data from POS into LOS
  • Complete forms in the LOS using data from documents
  • Create/assign tasks based on the specific events
  • Update CRM based on the changes in the LOS

Alerts and notifications

Operational automation can generate real-time or near-real-time alerts based on operational data. These alerts can trigger the sending of SMS, emails, or messages through various messaging platforms (such as Slack, Teams, WhatsApp, etc.).

Here are some examples:

  • Send SMS alert when rate lock expires in 5 days
  • Send MS Teams message when team member performance falls below the threshold
  • Send Email notification when the loan is completed, particular milestone

Typically, alerts fall into one of the following categories:

  • Team Performance Alerts: Notifications related to the productivity or efficiency of team members.
  • Company Performance Alerts: Alerts regarding overall company metrics and performance indicators.
  • Loan Lifecycle Alerts: Notifications tied to specific stages or events in the loan processing cycle.
  • Due Date Alerts: Reminders about approaching deadlines or critical dates.

Placing orders and interacting with vendors

In the same way actions and data entry are automated within software services, actions and data entry can be automated with the vendors through API or RPA.

Here is a list of orders that can be automated:

  • Credit reports
  • Property appraisals
  • Flood certifications
  • Title search and insurance
  • Verification of Employment (VOE)
  • Tax transcripts
  • Homeowners Insurance
  • Property survey

Document generation

A significant portion of the documents traditionally created manually during the loan lifecycle can be automatically generated. This process involves using document templates and populating them with the necessary data.

Here are some examples of documents that can be automated:

  • Pre-approval letters
  • Closing disclosure documents
  • Mortgage amortization schedules
  • Deeds of trust
  • Monthly mortgage payment invoices

Fraud detection

AI and ML technologies are extensively utilized for fraud detection automation in the mortgage industry. These technologies enable software products to identify fraud that might be invisible to the human eye.

Below are a few types of fraud detection that can be automated:

  • Document Verification: This process automates the checking of documents for authenticity and data consistency, ensuring that all submitted documents are valid and accurate.
  • Predictive Analysis: Utilizes historical data to identify patterns and forecast potential fraud in new mortgage applications.
  • Behavioral Analysis: Monitors applicant behaviors for fraud indicators, such as atypical login activities, which might suggest unauthorized access or false identities.
  • Data Cross-Verification: Cross-checks applicant information with external databases to verify the accuracy and consistency of the information provided.
  • Synthetic Identity Detection: Detects fabricated identities by analyzing credit histories and personal data to uncover discrepancies that suggest a synthetic identity.
  • Payment Transaction Monitoring: Observe payment transactions for anomalies or inconsistencies, such as unusual payment sizes, frequencies, or sources, which can indicate fraudulent activities in mortgage transactions.

Risk assessment and decision-making

Risk assessment and decision-making tasks in mortgage operations can also be automated.

Here are some applications of automation in the risk assessment and decision-making process:

  1. Knock-Out Rules Automation: Automatically disqualifies applications that fail to meet basic criteria like minimum credit scores or recent bankruptcies.
  2. Risk Score Calculation: Uses algorithms to analyze financial data and calculate a risk score, assessing the likelihood of default.
  3. Automated Approval/Rejection: Quickly decides on loan applications based on risk scores and alignment with the lender's criteria.

Document analysis and validation

Thanks to advances in the OCR and AI, a great deal of the document analysis and validation can be automated.

Here are a few examples of how document analysis and validation can be automated:

  • Alerts for missed or mismatched data
  • Validation against business rules
  • Closing Disclosure (CD) balancing
  • Signature verification
  • Validation against public data sources

Document indexing and storage

Using AI to classify documents and APIs of the file storage software, you can automate document indexing and storage tasks.

Here are a few examples of how these tasks can be automated:

  • Split loan packages into individual documents
  • Rename documents following a naming convention
  • Sort and organize documents within file storage or LOS

Calculations and logic

Automated systems can do automatic calculations relying on the operational data.

Here are a few common calculations in the mortgage process that can be automated:

  • Borrower income
  • Risk-based pricing
  • Closing costs
  • Late payment fee
  • Foreclosure cost
  • Capital gains and losses

Payment processing and money movement

Automated systems have made it possible to programmatically move money between bank accounts.

Here are some use cases within mortgage operations:

  • Payment Processing and Direct Debit Setup: Automated systems process monthly mortgage payments from borrowers, ensuring timely and accurate recording of payments.
  • Escrow Account Management: Automate handling payments for property taxes and homeowner's insurance through escrow accounts, ensuring these are paid on time and accurately from the borrower's funds.
  • Refinance Disbursements: Automates the disbursement of funds in refinancing scenarios, ensuring that the new mortgage funds are appropriately allocated to pay off the existing mortgage and other related costs.
  • Payment Allocation: Automatically allocates payments to interest, principal, taxes, and insurance in the correct proportions per the loan agreement.
  • Overpayment and Underpayment Adjustments: Automatically adjusts future payments or account balances in case of overpayments or underpayments by the borrower.
  • Payouts for Insurance Claims: Manages the disbursement of funds for insurance claims related to the property, such as damage or loss.
  • Investor Remittances: In cases where loans are sold to investors, automates the process of remitting payments from borrowers to these investors.
  • Interbank Transfers for Loan Funding: Facilitates rapid and secure funds transfer between banks during the loan origination.

How to build operational mortgage automation

The high-level process of developing operational mortgage automation involves these steps:

  1. Understand what workflows you want to automate
  2. For each workflow, define its trigger, steps, and actions
  3. Define what data is needed for triggers, steps, and actions
  4. Collect the data required for triggers, steps, and actions
  5. Develop triggers based on the data you collected
  6. Develop actions required for workflow
  7. Develop automated workflows

Here are some key tech used in operational automation:

  • Database for the entity storage (e.g., MongoDB, PostgreSQL, etc.)
  • Database for the events storage (e.g., ClickHouse, Swnoflake, etc.)
  • Event streaming technology (e.g., Kafka, Redpanda, etc.)
  • Event processing technology (e.g., Decodable)
  • Automation platform (e.g., Pipedream, n8n, Make, Zapier, etc.)
  • Action-specific tech (e.g., SendGrid, Twilio, Taktile)

An in-depth article on building operational mortgage automation is coming soon.

What’s next

I hope this post gave you insight into Operational Mortgage Automation and how it can be used in mortgage operations.

If you’d like to stay on top of the latest mortgage tech and how it can be applied to mortgage operations, consider joining our mortgage technology newsletter.

MORTGAGE TECH NEWSLETTER

Discover how technology can assist your mortgage company in reaching its strategic goals

A weekly newsletter about leveraging data, custom software, and modern technology to drive efficiency in mortgage operations.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Written by
Vova Pylypchatin
Founder and CTO @ MortgageFlow

I’m a software consultant with background in software engineering. Currently, I run a mortgage software consulting and development company that builds custom tools and automation solutions for mortgage lenders.