AAFMAA: Advancing Cyber Security and Compliance with AI and Adabas & Natural 

We’ve relied on Adabas & Natural since 1974.  Recently, we’ve deployed mobile applications and cloud-based systems. Now, we’re using Artificial Intelligence (AI) to enhance data security, regulatory compliance, and operational efficiency. That’s progress.

Amarish Pathak, CTO, AAFMAA

Data integrity: a modern-day must  

As a nonprofit financial services provider serving military families since 1879, AAFMAA handles highly sensitive data across its insurance and wealth management divisions. With increasing regulatory scrutiny—such as the California Consumer Privacy Act (CCPA), the EU’s GDPR, and the Gramm-Leach-Bliley Act—it is paramount for the organization to ensure complete data integrity and security. The mortgage division in particular faces rigorous audits from state regulators. Every loan must be completely traceable, secure, and compliant, from origination to closure.  

So, when a cyberattack hit, it became clear that traditional systems were no longer sufficient. Third-party vendors were eager to plug the gaps. But AAFMAA wanted to take a more intelligent approach to anomaly detection and fraud prevention. It needed a solution that could analyze real-time data, and detect and avoid threats, while being adaptable to evolving compliance requirements: Ideally, something home grown. 

---

Adabas & Natural meets AI 

“We decided to build a system which could rely on a database as the key data source and then use Large Language Models (LLMs) in our own hosted environment with the most modern data integration approaches to detect vulnerabilities,” explains Amarish Pathak, CTO at AAFMAA. “Under my direction we launched a strategic initiative to integrate AI into our Adabas and Natural environment. The goal was to build an in-house anomaly detection engine that could leverage Adabas data logs and employ modern Machine Learning techniques to identify anomalies and prevent fraud in real time.” 

Tangible results surfaced soon after work to create the engine began. So far, five use cases are already showing huge potential: 

1. Forensic analysis with Adabas CLogs and PLogs: AAFMAA’s engine uses Adabas command logs (CLogs) and protection logs (PLogs) to capture real-time snapshots of database activity. CLogs record user sessions, command types, execution times, and transaction statuses. PLogs provide before-and-after images of data changes to enable rollback and, if necessary, forensic analysis. 

2. Machine learning for anomaly detection: The custom AI engine uses Python and Natural and integrates LLMs such as Mistral and LLaMA. These models can analyze patterns in logs to separate normal behavior from anomalies that could indicate fraudulent activity. 

3. Threat modeling and penetration testing: To ensure the database is resilient, AAFMAA’s internal penetration testers use AI both offensively and defensively, realistically simulating attacks to test the system’s security. 

4. Real-time alerting and case management: The engine includes a rule-based detection layer and a policy enforcement module that triggers alerts for suspicious transactions. The web-based interface allows compliance officers to monitor activity, adjust rules, and manage flagged cases independently and remotely. 

5. Seamless integration across the business architecture The AI engine has been integrated with AAFMAA’s CRM systems and verification APIs. This means it can support checks for credit history, appraisals, titles, and OFAC compliance. 

---

Unlocking value in the database

Although AAFMAA is just at the start of its journey, the team has already learned several important lessons: “Using machine learning isn’t just about automation,” Pathak explains. “It’s made our employees much smarter at detecting fraud and anomalies. We’re building a culture of deep thinkers.” 

But the benefits aren’t just felt in continuous learning and improvement. Since deploying the engine, operational costs have dropped significantly as external vendors have been replaced by reliable, internal AI solutions. This gives AAFMAA more independence, flexibility, and frees up budget for further investment in innovation.  

Recently the AI engine was integrated into Encompass— AAFMAA’s Loan Origination Platform—as well as its underwriting processes. This has put real-time fraud detection capabilities into the hands of core business units. And has significantly boosted the trust in data integrity across departments.  

Each state in America has different compliance rules. Prior to this undertaking this project, this degree of regulatory variation would have been a minefield. Now, rules are now customizable on a state-by-state basis, which has improved audit readiness and reduced the risk of fines.   

Most important, however, are the benefits that flow to the members that entrust their data and livelihoods to the  Association. By adopting AI technology in a thoughtful way—and setting it to work within a reliable, powerful database—AAFMAA has transformed how its customers’ data is secured, analyzed, and used to drive smarter decisions. Faster development cycles with AI- assisted coding and testing means innovation is easier than ever. 

---

Future-ready to better serve those that served 

AAFMAA’s integration of AI into its Adabas & Natural environment marks a significant milestone in its modernization journey. By combining decades of experience with cutting-edge technology, the organization has built a future-ready system that protects its data and that of its members, empowers its teams, and ensures compliance in an increasingly complex regulatory landscape. 
 
For other organizations navigating similar challenges, AAFMAA’s story offers a powerful example of how continuous modernization, in combination with the right technology and like-minded partner, can deliver game-changing transformative results. 

“I don’t look at Software AG as a vendor. I look at it as family.” 
— Amarish Pathak, CTO, AAFMAA