Modernizing critical infrastructure with AI-assisted engineering

For one federal enrollment program serving millions of people each year, an issue with a foundational software dependency had evolved into a significant issue over time. After a period of slower development followed by a fast-paced period of updating, adopting the latest version required far more than a routine upgrade. It meant adapting to new conventions while preserving capability with the existing platform.

Doing so would strengthen the platform’s security posture, reduce long-term maintenance risk, and create a stronger foundation for future development. While what needed to be done was understood, the challenge came down to dedicating sustained engineering time to a technically demanding modernization effort while also continuing to deliver new capabilities and support day-to-day operations.

Guided by experienced engineers and rigorous validation, Ad Hoc used AI-assisted engineering to complete the upgrade in a single day.

The challenge

A key open-source dependency used to generate database models had been pinned to a pre-release version. Since then, the software that uses this open-source dependency had undergone a significant redesign, making the upgrade much more complex than a routine version update. Existing build scripts, development workflows, and templates across three services depended on the older interface, and a direct upgrade would have broken database model generation.

Pinning dependencies to a specific version is a very common approach in software development. It’s also one of the most common contributors to technical debt because it encourages a “set it and forget it” mindset. While the temptation to simply leave outdated pinned dependency versions alone can be tempting, doing so can also carry significant risk.

Continuing to rely on the older dependency would also have meant carrying forward older libraries and components that were no longer receiving security updates, increasing long-term maintenance and creating security risk.

TThe engineering team recognized the need for the upgrade and made meaningful progress while continuing to prioritize feature delivery and operational support. Over time, the team developed a deep understanding of the problem and identified a path forward. But completing the work required a sustained engineering effort that was difficult to dedicate alongside customer priorities.

Our approach

Rather than rewriting every workflow that depended on the older software dependency, Ad Hoc used GitHub Copilot to develop a compatibility approach that allowed existing processes to continue working while adopting the latest supported version.

Engineers provided the technical context and oversight needed to guide the AI while ensuring the solution met established engineering standards.

Every change followed the team's standard engineering practices. We:

• Reviewed the AI-generated implementation.

• Validated the generated database models across all three services.

• Confirmed the existing test suite passed successfully.

• Verified the solution before it was merged into production.

By combining AI-assisted engineering with experienced human oversight, we accelerated a challenging modernization effort while maintaining the quality and rigor required for a production system.

Outcomes

In a single 11-hour engineering session, Ad Hoc completed a technically complex infrastructure upgrade that had remained on the team's roadmap for nearly two years.

Key outcomes included:

• Reduced long-term maintenance and security risk while establishing a stronger foundation for future development without disrupting ongoing customer delivery.

• Preserved existing functionality across all three services through a compatibility-first approach rather than a large-scale rewrite.

• Verified production readiness through engineering review, testing, and validation, with no corrections required to the AI-generated implementation before deployment.

This work demonstrates how AI-assisted engineering can help teams complete complex modernization efforts that are often difficult to prioritize alongside feature development. By combining AI with experienced engineers, disciplined delivery practices, and rigorous human review, Ad Hoc strengthened the long-term reliability and maintainability of a mission-critical government platform while allowing the team to remain focused on customer priorities.