The $1 Billion Blueprint: How Socotra Capital Is Using AI to Double Its Assets Without Doubling Its Headcount
Client: Socotra Capital
Project: Streamlining processes with a custom AI-powered agent built on Microsoft Copilot Studio
Every morning at 8 AM, Roy Malone—president of Socotra Capital, a Sacramento-based private money lender—meets with his loan committee. For an hour each day, a team of senior executives gather to review a queue of anywhere from five to fifteen deals—a mountain of documents, folders, and browser tabs standing between them and a lending decision.
Socotra Capital raises funds from individual investors and deploys that capital as loans backed by real estate, helping developers, investors, and business owners who can’t get traditional bank financing to acquire, rehabilitate, and reposition properties. It’s a high-stakes, detail-intensive business. And for years, the process of reviewing those deals was almost entirely manual.
That changed when Socotra Capital partnered with Convverge to build what they’re calling “Agent PAT”—a custom-built AI-powered agent that’s now the backbone of their operation.
The Challenge: Information-Intensive Decisions Slowing Scale
For each deal that came through the pipeline, Socotra Capital’s loan committee needed to review three distinct streams of information:
- Documents provided by the borrower (tax returns, bank statements, purchase contracts, personal financial statements)
- Publicly available web data (news about the borrower or business, neighborhood and market trends)
- Gated subscription data (property details like zoning, current title holder, tax status, and square footage)
Pulling all of that together into something usable fell largely on the loan officers and underwriters—manually opening folder after folder, document by document, sometimes navigating files that were 100+ pages long.
Then came the committee meeting. Depending on its complexity, the team might spend 45 minutes on a single deal—pulling up Google Maps to view the property, hunting through SharePoint for a specific document, trying to cross-reference financial data scattered across a dozen sources. The last deals of the morning were routinely the ones that got deferred or rushed because the committee was running out of time and patience.
We were finding it was taking us more than an hour every morning to get through our opportunities. That’s a lot of time with highly paid executives sitting through data searches. We figured there had to be a better way.
Roy Malone, President, Socotra Capital
Beyond the time cost, there was an accuracy problem. Without a consistent, structured process for surfacing information, things could slip through—a red flag in a borrower’s public record, a discrepancy in the property data, or a legal issue that wouldn’t show up until well into the transaction. Discovering a problem late meant wasting valuable hours and resources.
The Solution: Enter Agent PAT
With a vision for a more streamlined process in mind, Socotra Capital started looking for an AI development partner who could help them build something that didn’t really exist yet—a custom AI agent. They found Convverge. What followed was a collaborative, iterative, nine-month build that pushed the boundaries of what AI-powered agents could do in a regulated, document-heavy financial environment.
The agent was named PAT—an acronym for Paul, Adham, and Tony (the three founding members of Socotra Capital’s original loan committee) and a fitting name for a tool that would sit at the heart of their decision-making process.
Three Jobs in One
Agent PAT wasn’t designed to serve one moment in the loan lifecycle—it was designed to run three times per deal, each time adding value at a different stage:
1. Pre-Committee Review
Before a deal ever reaches the loan committee, loan officers can run Agent PAT to check for gaps or red flags. If the borrower’s public record turns up a fraud conviction or legal issue, the deal gets killed before anyone has spent time on it, saving hours of prep work upstream.
2. Loan Committee Meeting
Agent PAT generates a comprehensive credit memo—an eight-page document that aggregates all three data sources into a single narrative. Committee members can click a hyperlink to pull up the property on Google Maps. They can access any source document directly without hunting through folder structures. Everything they need is in one place.
3. Final Underwriter Approval
Once all due diligence is complete, Agent PAT runs a final time with the full dataset. The output becomes the official approval document and doubles as a summary sheet that can be shared with potential buyers if the loan is later sold.
What’s neat is that with one agent, we’re saving a ton of manpower time, and we’re creating a much more robust work product. An underwriter would have taken three to six hours to draft that document manually.
Roy Malone, President, Socotra Capital
Built for the Real World
One of the early technical breakthroughs during Agent PATs development came when the team upgraded to a newer generation AI model capable of reading scanned documents and photographs of paperwork, not just natively digital PDFs. In the world of private lending, borrowers often send images of documents. That capability was a game-changer.
The Convverge team also built a web research function directly into Agent PAT. The agent actively searches the internet for information about the borrower, their business, and the collateral property, and then surfaces that information in a structured “Adverse Findings” section, complete with source links so the committee can verify anything it flags. In one early test, the agent surfaced serious legal issues related to an alleged fraudulent scheme connected to a prospective borrower—exactly the kind of information that could have derailed a deal far too late in the process.
Colour-coded risk indicators—green, yellow, and red—were added to help the committee quickly scan for issues without needing to read every paragraph in detail. The final credit memo links directly to the Google Maps listing for the property, and every claim is backed by a cited source.
The Convverge team thought, let’s do some colour coding—a stoplight system—to help the reader quickly assess whether there’s a potential issue instead of reading through every paragraph. There’s practical business application built into how they developed the product.
Roy Malone, President, Socotra Capital
Security, Non-Negotiable
Because the firm handles sensitive information including federal tax returns, bank statements, social security numbers, and credit profiles, security was never optional. Socotra Capital needed a solution built on enterprise-grade infrastructure that could guarantee client data would never be used to train external models or shared outside the organization. The solution was built on Microsoft Copilot Studio, leveraging the enterprise privacy architecture that Socotra Capital already relied on.
The Results: Faster, Smarter, and Built to Scale
Time Savings Across the Entire Lifecycle
Socotra Capital’s committee no longer spends 45 minutes on a single deal. The property is a click away, documents are linked, and risk flags are colour-coded. Now, their goal is to double their deal intake and get through 20 deals in under an hour.
Beyond the committee meeting itself, the savings compound across the loan lifecycle. Appraisal review alone—previously a manual process of hunting through a 182-page document to verify a checklist of items—can now be automated through an Agent PAT appraisal addendum. That’s an estimated hour saved per transaction, just on one task.
And the upstream impact is just as significant: deals with fatal flaws can be identified and killed in minutes rather than after days of preparation.
Accuracy and Consistency That Scales
Perhaps more valuable than speed is consistency. Every deal now goes through the same structured review process, asking the same questions, and checking the same data sources. When you’re processing the kind of volume Socotra Capital handles, that consistency is enormously valuable and difficult to maintain manually.
Roy describes the quality of the output as something that simply couldn’t be replicated by pulling the same prompt into a generic AI tool.
I could not get this document out of standard Copilot. This is what we wanted.
Roy Malone, President, Socotra Capital
A Foundation for Growth
Socotra Capital currently manages close to half a billion dollars in assets. Their ambitious goal is to double that, reaching a billion dollars under management. Getting there means processing more deals, faster, without proportionally expanding headcount. Agent PAT isn’t just a productivity tool, but an infrastructure investment that makes that growth possible.
There are already plans for Phase 2: a direct integration between Agent PAT and Salesforce, so that the agent can pull deal information from their CRM and further reduce the manual steps in the workflow.
Any company serious about growing, whatever vertical they’re in—if you don’t make this investment [into AI], somebody else is. And the company that does is going to take away market share by being faster, more efficient, and more accurate. It does cost on the front end. But the savings on the back end are going to more than pay for it.
Roy Malone, President, Socotra Capital
About Convverge
Convverge is an IT solutions provider specializing in AI-powered automation, data and analytics, and digital transformation. From proof-of-concept to full deployment, Convverge partners with organizations to build practical, enterprise-grade solutions that create real business impact.
Explore our AI Consulting services today >
