Funding Infrastructure, Built to Scale
How Fox Funding replaced manual operations with automated ACH, underwriting, and syndication, inside a live production stack.
- Financial Services
- AI-Driven Development
- Brownfield
- Industries
- Financial Services, B2B Funding
- Platform
- Funding platform: ACH, underwriting, syndication
- Delivery scope
- Full-cycle infrastructure, AI-Driven Development
ABOUT THE PROJECT
Project overview
- Engagement
- 2025 — ongoing, full-cycle delivery
- Team size
- 2 AI-Augmented Software Engineers
- Core technologies
- .NET/C#, React, Azure
Fox Funding provides financial products for businesses. Growing deal volume required infrastructure that scales under load, speeds up underwriting, and gives syndicators transparent access to their positions.
Limestone delivered several infrastructure projects on one scalable architecture, with AI-Driven Development applied at every stage, inside systems already running in production.
The Challenge
Manual operations, fragmented tools, and a live system underneath.
-
Funds movement didn't scale
Disbursement and collection ran on manual steps. Every new deal added operational load, and growing volume made it worse.
-
Underwriting had no single home
No one environment for assessing applications and making decisions. Applicant data, risk rules, and workflows sat in fragmented tools.
-
Syndicators had no visibility
No transparent access to positions, deal participation, or returns. Reporting was manual and always behind.
-
Brownfield, zero downtime
Every change had to respect the systems already running in production and land without disrupting live operations.
WHY LIMESTONE
Discipline, not just speed.
Fox Funding needed a shared architecture, predictable code quality, and a fast delivery pace. Speed without discipline creates technical debt, and the risk is higher in a brownfield codebase where a regression touches live money movement. Limestone paired the speed of working with AI with the engineering discipline the financial industry demands.
WHAT WE BUILT
Solution
Three infrastructure projects on one shared architectural foundation. Shared principles and components cut duplication, simplify maintenance, and let functionality grow without rewriting the core.
-
ACH Automation
Automated funds movement for disbursement and collection. Manual operations replaced with reliable transaction processing that holds up as volume grows.
-
Underwriting Portal
One environment for assessing applications and making decisions. Applicant data, risk rules, and team workflows in a single interface.
-
Syndicator Portal
Real-time, self-service access to positions, deal participation, and return metrics. Manual reporting replaced for good.
-
Spec-Driven Development
Work starts from a specification, the single source of truth for people and agents. It removes ambiguity and makes outcomes predictable.
-
TDD + Verification Harness
Tests written before implementation, plus static analysis, type checking, and custom validators. Every change passes verification before merge.
-
Automated PR Review
A PR agent and CodeRabbit review every change against the spec and standards, holding the quality bar at high velocity.
AI-DRIVEN, GOVERNED
Generated code, verified before merge.
Pointing an agent at a brownfield repository doesn't work. Limestone ran an AI adoption audit, then built the harness the audit called for: repository and module context files, a skills library encoding the team's recurring procedures, deterministic verification, and automated review. Every change passes tests, static analysis, type checking, and custom validators before it reaches review. AI-generated code ships governed, not trusted blindly.
HARNESS LAYER
From audit findings to a governed pipeline.
An AI adoption audit surfaced what blocked reliable agent work: missing context, undiscoverable build and test commands, inconsistent patterns, no reusable procedures. Each gap got a dedicated fix.
-
Step 01
AI Adoption Audit
Gaps and bottlenecks, named
-
Step 02
Context Files
Architecture, conventions, and commands, documented at repository and module level
-
Step 03
Skills Library
Recurring procedures, encoded once and reused
-
Step 04
Verification Harness
Tests, static analysis, and custom validators wired around the agent
-
Step 05
Automated Review
PR Agent plus CodeRabbit on every change
HOW QUALITY HELD AT SPEED
Fast, and safe for financial infrastructure.
-
SPEC-DRIVEN
The spec is the source of truth.
Every piece of work starts from a specification shared by people and agents. It removes ambiguity and makes outcomes predictable before code is written.
-
TEST-DRIVEN
Behavior locked before code.
Tests written ahead of implementation lock in expected behavior and give AI-generated changes a safety net where regressions cost real money.
-
GOVERNED MERGES
Nothing lands unverified.
Agents run inside a deterministic harness. Tests, static analysis, type checking, and custom validators pass before any change reaches review.
-
Infrastructure projects on one shared architecture
3
-
Automated review passes on every pull request
2
-
Changes merged without automatic verification
0
Impact & Results
What shipped, and what it changed.
- Automated ACH replaced manual disbursement and collection, and holds up as deal volume grows.
- Underwriting moved into one environment: applicant data, risk rules, and team workflows in a single interface.
- Syndicators see positions, deal participation, and returns in real time, self-service instead of manual reports.
- New projects launch on a ready foundation. Functionality grows without rewriting the core.
- AI-Driven Development ran reliably in a brownfield environment, with no disruption to live operations.
-
Infrastructure projects delivered
3
-
Shared architecture underneath them
1
-
Disruption to live operations
0