AI-Driven Migration Enables Rapid Database Migration
In Brief
Secure, AI-Powered Migration Delivered Modernization in Days
A public corporation that helps provide Alaskans with access to affordable housing, faced a pressing need to modernize its database infrastructure to support future growth, enhance operational resilience, and reduce escalating licensing costs. Partnering with Resource Data, our client embarked on a groundbreaking pilot project for its first few critical databases, utilizing Meta’s open-source Llama 3.1 AI model to transition from Microsoft SQL Server to PostgreSQL.
Our Human-in-the-loop (HITL) AI, in-house approach not only accelerated the database modernization timeline from months to days and reduced costs, but also safeguarded sensitive housing data, setting the stage for broader AI-driven initiatives within the organization.
Key Takeaways
Private AI, Big Results: Faster, Cheaper, and Fully Controlled.
-
Smart AI adoption, done securely
Running Llama 3.1 locally, enabled critical databases to rapidly transition to PostgreSQL without exposing sensitive citizen data to public cloud AI models.
-
Time and cost savings that mattered
The AI-powered database migration was completed in days instead of months—avoiding large licensing fees, and reducing reliance on costly external tools and additional staffing.
-
AI-powered, human-verified
A human-in-the-loop (HITL) AI strategy preserved control, ensured accuracy, and accelerated modernization—without compromising sensitive data.
-
Igniting a future of innovation
This success story sparked interest in using AI and automation to improve other critical services—from document management to public housing program administration—all while keeping control firmly in-house.
The Challenges
Legacy Infrastructure, Skyrocketing Costs, and No Room for Risk
As our client prepared to scale its operations, its aging Microsoft SQL Server infrastructure became an obstacle. The database system required expensive licensing for high-availability features such as replication between geographically separate data centers. For a mission-driven public corporation, these costs posed a significant barrier.
In transitioning to a new database system, time-consuming and resource-heavy manual database migration was not a realistic option. Security was another paramount concern. Our client handles sensitive citizen data, and leadership made it clear: no information could be exposed to cloud-based AI services, and no external third-party could compromise data sovereignty.
The Solution
Custom AI Workflow with Human Oversight Delivered Speed and Accuracy
We began by securely deploying Llama 3.1 on internal hardware. This modest $500 local hardware setup powered the translation efforts and eliminated any risk of data leakage to public cloud services.
Database objects—tables, views, stored procedures, and triggers—were translated directly from SQL Server syntax into PostgreSQL equivalents, all behind the safety of our client’s and Resource Data’s internal networks.
Going beyond simple automation, our team carefully fine-tuned the AI’s parameters for the unique complexities of SQL and PostgreSQL and then incorporated expert human oversight at key review points. Working through a method of iterative batch processing, we built a translation system that bulk-fed exported database objects through the AI, captured the PostgreSQL outputs, and then manually reviewed and corrected any inconsistencies.
Results
Human-in-the-Loop AI Delivers High-Confidence Outcomes at Speed
The migration work was performed securely, cost-effectively, and flexibly. By integrating a human-in-the-loop approach with open-source AI in a closed environment, we achieved rapid, accurate database migration—compressing months of manual effort into days—without sacrificing data integrity or security.
In the face of a traditionally slow, expensive, and risky challenge, our client now has a tailored, secure solution that gives them full control and sets the stage for future innovation.
“We helped them save months of work and thousands of dollars—while strengthening their security posture. They’re already exploring more ways to build on this foundation.”
Cory Scheaffer, Resource Data Sr. Programmer
-
AI automation for accelerated timelines
Where manual migration would have taken months, Resource Data’s AI-powered system reduced that effort to days.
-
Reduction in manual labor for major cost savings
We avoided the need for hiring additional developers, purchasing expensive migration tools, and continuing to pay high licensing fees.
-
Closed architecture for data privacy and security
Sensitive housing and citizen data remained entirely within our client’s and Resource Data’s private networks, fully compliant with internal standards.
-
PostgreSQL for enhanced resiliency
By transitioning to PostgreSQL, our client gained built-in high availability features included in its cost.
What's Next
AI Momentum is Building
The AI database migration was only the beginning. Building on its success, our client plans to expand the AI-assisted approach to nearly 80 databases that support critical services such as energy rebate programs and housing loan administration—an effort that would be time- and labor-intensive if done manually.
The organization is also exploring AI applications in Optical Character Recognition (OCR), intelligent search, and document management. With Resource Data’s ongoing support, our client is well-positioned to further leverage AI in advancing its mission to provide affordable housing solutions to Alaskans.
