AI Tools
Areal Copilot Agent and Areal CD Balancer are now patent pending. The application covers the execution architecture the two products share: the way a mortgage task is broken into an ordered, guardrailed plan, carried out step by step against the loan file and the lender's system of record, and recorded as it happens.
It also covers the mortgage Document AI underneath — six years of R&D across 1,500+ document types and 4,000+ data points per loan, running at 99% accuracy on critical fields.
This post is the longer version of the announcement: why a company files at this stage, what the filing actually protects, and why it shows up most clearly in CD balancing.
It is a fair question in a market where the technology moves every quarter.
Because the most valuable intellectual property is not an AI model or a prompt. It is the knowledge accumulated through years of solving the same difficult problem at scale.
The model layer is not where our years went. Models improve whether we do anything or not. The years went into the unglamorous part: working out what "right" looks like on a mortgage file, across millions of pages and four generations of production software, and then building something that applies that judgment the same way every time.
Agentic AI has made it remarkably easy to build an impressive demonstration. Building a system that performs reliably across tens or hundreds of thousands of loans is an entirely different problem. That gap is what we filed on.
Four properties decide whether agentic AI survives contact with production. They are the principle behind Areal Copilot Agent.
There is a pattern in this category worth naming, because buyers are being asked to underwrite it.
A number of agentic platforms rely on large, free-form prompts — sometimes pages of instructions describing what an agent should do. That can produce impressive results in a demonstration. It is much harder to guarantee the same process executes consistently across one loan, let alone hundreds of thousands. When it is wrong there is nothing to inspect, and when a compliance officer asks which rule was applied to a particular fee, there is no answer to give.
Copilot Agent works the other way. Its guardrailed agentic architecture provides structure around how agents plan, execute, validate and complete mortgage workflows. The plan is what runs; the model does the reading and the judgment inside a step, but it does not decide what the steps are.
And it does not operate on documents alone. Copilot Agent can bring together Areal's Document AI with data from the LOS, APIs, emails and external systems — whatever a lender wants to validate against. It runs natively inside ICE Encompass, Byte LOS and MeridianLink, so there is no migration and no parallel system of record.
The second patent-pending technology sits behind Areal CD Balancer, the industry's #1 CD balancing solution. It did not arrive that way.
The journey began in 2023, with the first generation. Since then the product has been through four generations of development. Generation 4 goes well beyond comparing numbers between two documents: it adds balancing intelligence, workflow automation and TRID tolerance categorisation, with every fee sorted into the tolerance bucket that governs it and logged against that classification at the moment the decision is made.
Every button, recommendation, validation and automated action in the product builds on years of observing how mortgage professionals actually balance Closing Disclosures.
That accumulated research is what lets Areal CD Balancer complete the balancing workflow in approximately one to two minutes — a process that traditionally requires extensive manual review, at 25 to 35 minutes per round for a typical lender and 35 to 45 minutes at a bank closing operation, across three or four rounds.
That is not the result of a single prompt. It is years of product development, mortgage expertise, customer feedback and thousands of engineering decisions about how the process should work. If you are new to the step itself, start with what CD balancing is.
It means a patent application has been filed and is under examination. It does not mean a patent has been granted. Areal Copilot Agent and Areal CD Balancer are patent pending as of September 2026.
The guardrailed agentic execution architecture shared by both products — how agents plan, execute, validate and complete mortgage workflows within defined structure — and the mortgage Document AI the two of them run on.
Four properties: accountable, consistent, reliable and repeatable. Accountable means every step is logged against the rule that governed it. Consistent means the same file produces the same result. Reliable means the agent writes back to the system of record instead of handing a human something to re-key. Repeatable means it behaves the same on the hundred-thousandth loan as on the first. Platforms built on large free-form prompts demo well but struggle to guarantee all four.
A long free-form prompt hands the whole task to a model and takes back whatever comes out. A guardrailed agent runs an ordered plan of discrete steps with defined inputs, defined actions and recorded outputs. The model works inside a step rather than in place of the plan, which is what makes the behaviour repeatable and the record inspectable.
The balancing workflow completes in approximately one to two minutes. Manually, a single round takes 25 to 35 minutes at a typical lender and 35 to 45 minutes at a bank closing operation, and most files need three or four rounds.
ICE Encompass, Byte LOS and MeridianLink, natively. The agents read from and write back to the LOS, so no export or rekeying step is required and there is no migration.
No. Both products are available today and continue to work exactly as they do now.
We'll show you which workflows to automate first — and the exact ROI for your volume.
Book a Demo →
