February 2024

By Ronald Zurawski, Data Governance Strategist and Solution Architect

In the world of artificial intelligence (AI), the ethical implications of bias in models have gained prominence. This issue demands close attention from organizations and data governance professionals alike.

As data governance consultants, it’s vital to guide businesses on how to handle bias in AI models responsibly. This article explores the role of data governance in identifying, mitigating, and preventing bias to ensure AI systems deliver fair outcomes.

Understanding Bias in AI Models

Bias in AI models can arise from many sources — biased training data, algorithm design, or even deployment context. Recognizing bias as an ongoing challenge, data governance helps establish frameworks to oversee the AI lifecycle from data collection to deployment.

A strong data governance strategy should emphasize:

• Transparency
• Accountability
• Ethical considerations

These principles form the foundation for fair and trustworthy AI practices.

Mitigating Bias Through Robust Data Governance

Data governance must take proactive steps to reduce bias in AI models. This includes:

• Applying strict data quality controls
• Ensuring diverse and representative training datasets
• Encouraging collaboration between data scientists and domain experts

Governance frameworks should also include continuous monitoring to detect and correct bias as models evolve.

Implementing Ethical AI Principles

Data governance professionals should promote ethical AI principles within all organizational practices. This involves:

• Setting clear guidelines for responsible AI use
• Encouraging diversity in data and development teams
• Maintaining detailed documentation to ensure transparency

When organizations align governance with ethics, they build trust and show their commitment to fairness and inclusivity.

Conclusion

In the evolving AI landscape, data governance is essential to address bias and guide responsible AI use. By fostering a culture of transparency, accountability, and ethics, organizations can create AI systems that are not only powerful but also fair and reliable.

My Perspective

Does this sound familiar? Have you read something similar before? It’s a solid article — but let’s look deeper into a few areas.

“Data governance plays a crucial role in establishing frameworks…”

That’s true — but how, exactly? Let’s skip the usual “it depends” and think practically.

As data governance professionals, we need to build a basic structure. Questions to consider:

• What do current regulatory requirements say?
• Can we participate early enough in the development process to guide documentation?
• What specific items must be ready for audit reviews?

By staying aware of evolving regulations, we can provide value while balancing compliance and cost.

“Bias in AI models can arise from various sources…”

Does bias come from data, or from how humans interpret results?

AI models respond to the training data they receive. From a data quality standpoint, we can add value by treating training datasets like critical data elements (CDEs).

Working with subject matter experts (SMEs), governance teams can review data profiles and assess where bias may appear.

“Data governance should prioritize proactive measures to mitigate bias in AI models.”

How might that work? This area offers new opportunities for data governance specialists.

In the past, we often relied on SMEs to interpret profiling results. But now, governance teams can take a more active role — defining standards for what constitutes bias, documenting them, and applying those standards to training data.

By doing so, governance shifts from passive support to strategic leadership.

Final Thoughts

I’ll explore this topic further in upcoming posts, but for now, consider these questions carefully.

Ron – ron.zurawski@prolifics.com

By Chris Isayan, Associate VP, Digital Experience

The AS/400 system and RPG programs have been the backbone of many enterprises for decades. These applications are known for their reliability, security, and stability—so much so that many organizations cannot even recall the last time they restarted their AS/400 systems.

If these applications are so dependable, why modernize?
The primary reason is the shrinking AS/400 RPG talent pool. As experienced developers retire, it becomes increasingly difficult to maintain or upgrade these systems.

Other motivations include:

• High cost and complexity of changes due to the system’s monolithic design
• Limited integration with modern tools and frameworks
• User dissatisfaction with outdated “green screens” and limited UI capabilities

While modernization is essential, the timing and approach depend on each organization’s unique goals, risk tolerance, and long-term strategy.

Key Concerns in AS/400 Modernization

Many clients express concern about the potential loss of business logic during modernization. Over years of development, AS/400 programs have accumulated complex business rules and workflows deeply embedded in RPG code—with limited separation between screen and backend logic.

Another concern lies in the database model. AS/400 systems use a hierarchical data structure, which can be challenging for teams familiar with relational database models (RDBMS).

Modernization Challenges Include:

• Understanding legacy business rules and logic
• Transitioning from hierarchical to relational models
• Managing data migration and dependency refactoring
• Maintaining reporting integrity and backend performance

A deep understanding of these systems is crucial before modernizing the front end.

Modernization Approach: Where to Begin

Modernization usually starts with a discovery and assessment workshop. Each AS/400 system is highly customized to business operations, often with unique, one-off scenarios.

Our Process Includes:

1. Identifying pain points, workflows, and business nuances
2. Rationalizing existing systems — what stays, what gets rewritten, and what’s replaced
3. Establishing priorities, timelines, and budgets
4. Implementing modernization through reverse engineering, development, and testing

While automation tools exist for converting AS/400 front-end code, most are proprietary and unavailable for licensing.

Leveraging GenAI for AS/400 Modernization

At Prolifics, we are exploring how Generative AI (GenAI) can streamline the modernization journey.

We’ve broken the process into three essential steps:

1. Understanding the legacy code
2. Reverse engineering business capabilities
3. Translating code into modern languages

This structured, AI-driven approach helps accelerate transformation while maintaining code quality and business logic integrity.

Practical Application Example

• Point the AI to a supply chain ERP codebase
• Ask it to describe the procurement workflow
• Generate code for a “New Purchase Order” screen using Angular (front-end) and Node.js (middle-tier)

By creating a scaffold code with architectural decisions, design patterns, and best practices, GenAI-generated code can be efficiently integrated—balancing automation with manual control.

Overcoming GenAI Challenges

While GenAI adds value, there are important considerations:

1. Outdated Training Data

Models like ChatGPT may rely on outdated frameworks. To address this, we provisioned a custom LLM on the cloud, fine-tuned for current libraries and best practices.

2. Inconsistent Code Outputs

Each generation can vary slightly. To achieve consistency, we use:

• Temperature control parameters
• Predefined templates
• Manual code review and unit testing

Even with these precautions, manual validation remains essential for production readiness.

It’s worth noting that IBM watsonx Code Assistant for Z follows a similar hybrid approach, combining rules-based translation with GenAI—although it currently supports only Mainframe COBOL on Z systems.

Coexistence of Legacy and Modern Applications

Modernization doesn’t happen overnight. Many organizations must run legacy and modern systems simultaneously during transition phases.

Best Practices for Coexistence:

• Roll out modernization in phases (e.g., procurement → inventory → logistics)
• Develop a legacy decommissioning plan
• Optimize hardware and resource utilization for both environments
• Train existing staff to manage modern applications

Proper planning minimizes operational disruption and ensures continuity throughout the transformation journey.

Deployment & Change Management

End-user adoption can make or break modernization success. Many long-time AS/400 users love their keyboard-driven green screens for speed and familiarity.

Key Success Factors:

• Maintain critical functionality during modernization
• Deliver complete business workflows to avoid switching between systems
• Enable single sign-on (SSO) for seamless navigation

Change Management Essentials:

• Keep a log of user concerns and changes
• Develop comprehensive Run Books
• Communicate updates transparently
• Provide pilot user programs to champion adoption

The goal is zero disruption and maximum efficiency in daily operations.

Conclusion

More organizations are now transitioning from AS/400 RPG to modern programming environments. The key is to start while RPG expertise is still available in-house, ensuring continuity and smoother knowledge transfer.

Partnering with experienced digital engineering experts ensures:

• Accurate code reverse-engineering
• Automated quality assurance
• Effective training and knowledge handover

At Prolifics, we combine GenAI-driven prototypes, accelerators, and automation testing tools to fast-track modernization with quality and precision. Our integrated approach empowers your teams to take ownership and maintain the new systems confidently post-transition.

Special Thanks

Special appreciation to Konrad Konarski, Gregory Hodgkinson, Honda Bhyat, Rajeev Sharma, Ramnath Nalluri, and Sheryl Oleksa for their insightful contributions to this blog.

About the Author

Chris Isayan, Associate VP, Digital Experience at Prolifics, specializes in global-scale digital transformation and modern application design. Known for his “learn the industry first” approach, Chris leverages Cloud, AI/ML, NLP, and computer vision technologies to accelerate enterprise innovation.

By Ronald Zurawski, Prolifics Data Governance Strategist and Solution Architect 

One of the new terms spreading into Data Governance vernacular is “Drift.” How do we, as Data Governance Specialists, begin to adjust our Data Governance programs to begin to accommodate drift and the possible regulatory and compliance issues that we may face? 

First, let’s take a look at what “Drift” is. Let’s say the developers have put together an AI data model and have trained it to do a specific task. As time passes, new input may need to be added for the model to evaluate. For example, in the Health Insurance industry, claims are evaluated against IDC-10 codes. These codes have been known to change over time. If developers simply add the new codes to the training set and place the model back into learning mode, how will the future output of the model change? It will change. This change is called drift. The business will need to evaluate if this drift is good or bad for the intended purpose. 

From a Data Governance perspective, we might profile the output data from the AI Model and then work with the SMEs (subject matter experts) and data owners to determine if the results are acceptable. Going back to our Health Insurance industry example, some of the IDC-10 codes may be added. Some may be deprecated. We know that smaller providers may not update their system immediately, yet some claims may need to be forwarded to entities for reimbursement, like the government.  

How do we as Data Governance specialists interact with these AI Models? Do we mandate that AI Model Code and Data Training Sets be secured and archived before the model is placed into production? Will this become a regulatory requirement? Do we laboriously go through multiple iterations of large output results from the AI Model and confirm Drift has not caused the model to stray from its initial accuracy or purpose? 

In the past our software, OS and any underlying components of the infrastructure have been built to always produce the same result. We were assured that an If-Then statement would return the same result every time and could build test systems that confirmed this. Now we are working with interactions between numerical numbers. If we were to change the order the underlying AI Model evaluates its neural network, can we guarantee the same results? 

AI Drift is opening up a whole new set of challenges for Data Governance. We will need to take those lessons learned from Glossary builds, Data Quality assessments and Data Lineage pulls, applying them to the whole new world of AI. This is going to be fun! 

More from Ron Data Governance And AI Model Bias, Part 1 | AI Governance (prolifics.com)

Ronald Zurawski

Ron is Data Governance Strategist and Solution Architect at Prolifics. His experience includes more than 10 years of working policy-driven data governance and more than 20 years in enterprise database and data warehousing systems. His industry expertise includes finance, health care and consumer product goods. Ron’s expertise is in strategic planning, systems architecture, program and project management. His tactical experience includes analytics development, architecture, ETL and database administration. Ron has experience in both Big 4 and boutique professional services organizations. Ron holds an MBA from the University of California and an MSCS from the University of
Colorado.