État de Vaud
Modernizing legacy systems with AI without compromising data sovereignty: a Swiss public sector case study.
When an organization can't expose its data to the outside world, AI-accelerated ways of working are still available.
Most public sector organizations have not been able to modernize with AI while keeping their data on their own infrastructure. AI-accelerated development almost always runs through public cloud platforms or third-party AI services, which is exactly what sovereignty-bound organizations need to avoid.
The État de Vaud's Directorate General for Digital Affairs and Information Systems (DGNSI) faced that problem. A business-critical application used daily by public employees and residents across the canton needed to move off an aging AngularJS codebase before technical debt turned into operational risk. The mandate was clear: modernize the application, and do it faster than a traditional migration would allow. The constraint was that no application or project data could touch a public cloud service or third-party AI platform.
SQLI provided a way for DGNSI to do both. The solution was the bespoke architecture SQLI put in place.
The results: 30% more development velocity, a 25% reduction in delivery time for new features and 99.9% service availability.
Why AI-accelerated modernization often isn’t possible in the public sector.
The issue at DGNSI was typical: AngularJS reached end-of-life years ago, and every additional feature built on it added maintenance cost and risk. A migration to modern Angular was necessary to secure the application, improve maintainability, and bring delivery costs down over the long term.
AI-assisted code analysis and transformation is genuinely useful for this kind of migration: identifying outdated patterns, mapping what needs to change, and automating repetitive rewrites. But most of the tools that do this well run in the cloud, which is usually a non-starter in the public sector.
The approach: AI in the infrastructure you control, with humans in every decision.
SQLI deployed its AI for Delivery approach entirely inside an environment controlled by DGNSI using open-source AI models running on a dedicated, secure infrastructure.
What the AI produced was a starting point, not a finished result. SQLI's engineers reviewed every recommendation the AI agents generated, validating, refining, and correcting it against the project's quality, security, and performance requirements. No AI-generated change reached production without a person deciding it belonged there.
We were convinced by SQLI’s approach of leveraging AI to enhance human expertise within a controlled environment supervised by their team. This enabled us to effectively address the challenges of adaptively maintaining our software solution in the face of rapidly evolving technologies.
The division of labor between AI for pattern recognition and humans for judgment calls is what allowed the project to move fast without asking DGNSI to loosen its data controls.
Automating the repetitive parts of code analysis and transformation freed the team to save time and money by spending more time on the harder, higher-value parts of the migration.
This project demonstrates that innovation and sovereignty are not mutually exclusive. By combining the DGNSI team's expertise with our AI for Delivery approach, we were able to accelerate the modernization of a business-critical application while ensuring full control over data and technology choices.
A model for organizations that can't compromise on data control.
Any organization operating under strict data residency rules, regulatory oversight, or sovereignty requirements faces the same challenge: slow down the modernization, or find a way to bring AI acceleration inside an environment they fully control.
The same approach applies to:
- Various legacy application modernizations
- Technical debt reduction in existing codebases
- Automated testing
- Software delivery process optimization
In each case, the requirement is the same the DGNSI had: running AI where the organization controls the infrastructure, with every output reviewed by people who own the outcome.
Modernizing a business-critical system is hard enough without the risk of exposing sensitive data. If your organization is weighing AI-accelerated modernization against data sovereignty, regulatory, or security requirements, that's a conversation worth having with SQLI.
Results achieved
Facing similar challenges?
Whether you're modernizing business-critical applications, reducing technical debt, addressing regulatory constraints, meeting data sovereignty requirements, or accelerating software delivery cycles, every organization faces its own unique challenges.
Let's discuss your priorities and explore how an AI for Delivery approach can help accelerate your transformation initiatives while meeting the specific requirements of your environment.