SIGNAL: AI-Assisted Credit Monitoring

SIGNAL (Strategic Intelligence Gateway for Next-Gen Analytics & Lending) is a proof-of-concept, AI-assisted decision-support system for commercial credit monitoring, built as our SAIT Integrated AI capstone in consultation with a major Alberta-based financial institution. It addresses the operational challenges of manual credit risk tracking with an automated, explainable, and responsible risk-monitoring framework.

I was Project Manager, leading delivery across a four-person team.

Commercial Portfolio Managers and Credit Analysts are responsible for continuously monitoring borrower financial health. However, manually reviewing quarterly financial statements across growing portfolios is time-consuming, difficult to scale, and inherently inconsistent under tight deadlines. Catching warning signs late limits remediation options for both institutions and borrowers while creating regulatory audit challenges. Credit teams need a scalable, consistent, and explainable way to catch early warning signs without losing human judgment.

SIGNAL automates the end-to-end credit monitoring workflow through a continuous, six-step process:

• Ingest: Accepts borrower financial packages in mixed formats, including PDFs, Excel spreadsheets, CSVs, scanned images, and emails.
• Extract: Pulls critical data points such as revenue, EBITDA, debt, working capital, covenant terms, and key dates.
• Validate: Verifies extracted data for completeness, formatting, and plausibility before it is trusted.
• Analyze: Calculates standard credit-monitoring ratios (leverage, liquidity, coverage, cash conversion) and checks them against Early Warning Indicator (EWI) rules.
• Alert: Flags metrics that breach thresholds as rated (Low / Medium / High) alerts, complete with source evidence and specific rule triggers attached.
• Present: Displays findings in role-tailored dashboard views so analysts can review, validate, correct, or act.

To support operational workflows, system access and actions are tailored across three primary roles:

• Administrator: Complete operational and governance access, including scheduling, system logs, rule configuration, and user management.
• Credit Manager: Oversight of documents, analyst views, audit logs, rule modifications, and team workflows.
• Credit Analyst: Dedicated view to inspect extracted data, review analysis outputs, and record review decisions.

SIGNAL is strictly built as a decision-support platform, ensuring every final decision remains with qualified human professionals. By design, SIGNAL:

• Does not approve or decline credit applications.
• Does not assign official credit ratings, risk scores, or default predictions.
• Does not recommend specific restructuring or remediation actions.
• Does not connect to live core banking systems or production client data (operates entirely on synthetic and public reference data).

As part of the proof of concept we built a business case for the workflow SIGNAL replaces. Reviewing a single borrower's quarterly financials by hand takes a credit analyst seven to eight hours. Against that baseline the team modelled:

• 60% to 70% less time per review, from seven to eight hours down to two and a half to three.
• 50% to 60% more review volume from the same team, meaning more borrowers monitored and more proactive engagement.
• $400K to $550K per year in productivity improvement across the portfolio.

These are modelled projections built for the proof of concept's business case, not measured results from a production deployment. SIGNAL was never connected to live banking systems.

Developed as an auditable proof of concept, SIGNAL establishes a repeatable, defensible monitoring framework. Every step is logged and every alert traces back to a source document, an extracted value, and a named rule version, so a finding can be explained and defended after the fact rather than simply trusted.

About the Course: PROJ 407

SIGNAL was delivered for PROJ 407, the capstone of the SAIT Integrated Artificial Intelligence (IAI) program. It is a semester-long project delivered in phases, running from May 2026 to August 2026, and serves as the cumulative capstone designed to synthesize learnings across all core IAI disciplines:

• AI Management & Maintenance
• Human-Centred Design
• Ethics & Governance
• Predictive Analytics
• Natural Language Processing
• Computer Vision
• Web Design & Cloud Computing

From the PROJ 407 Course Description:

This course is designed to provide students with the opportunity to apply the knowledge and skills they have gained throughout the program to real-world AI projects and applications in a “safe to fail” environment. This course emphasizes hands-on learning and provides students with the chance to work on projects that address real-world AI challenges and problems using the skills they are developing in their other courses.

In addition, the course covers important computer vision applications, such as autonomous vehicles, medical imaging, and security systems. Students will learn how to build and deploy computer vision systems that can be used in real-world applications, such as robotics, surveillance, and manufacturing.