Long development times and high costs have so far made custom solutions attractive mainly for larger projects. A project by Glue itself shows that this is changing fundamentally: Urs Nacht, Senior Engineer and member of the executive management, developed a comprehensive solution for managing and billing solar power in a self-consumption community (ZEV) in just two weeks.
AI Software Engineering: From precise specifications to deterministic code
AI can generate code today – but for production software, that alone isn’t enough. At Glue, we combine AI Software Engineering with precise, unambiguous specifications and experienced senior engineers. This produces code that remains traceable and controllable within clearly defined rules.
Our approach: Spec-driven development instead of free-form prompting and vibe coding. This creates software that is precisely tailored to a specific process while remaining traceable, maintainable, and controllable.
More on AI Software Engineering at Glue →
Case study: A ZEV process as a custom software solution
A self-consumption community (ZEV) distributes locally produced solar power among multiple parties in a property. Production and consumption data from different meters previously had to be compiled manually — time-consuming and error-prone.
As the owner of his own solar installation, Urs Nacht faced exactly this challenge. He wanted full control and transparency over production, consumption, and billing based on the existing Swisseldex data — and no longer wanted to work with Excel or standard subscriptions.
Custom software that truly performs in production
The result isn’t an AI demo, but software used daily in production for distributing and billing solar power. The dashboard covers:
Paid standard solutions exist on the market for this use case. Since Urs Nacht owns the apartments himself, a custom solution proved better and more cost-effective — saving around CHF 2’000 per year.
AI-supported processes instead of manual data entry – Different files and formats are automatically recognized, read, and assigned to the corresponding units. AI handles the laborious groundwork, while the software reliably integrates the process into the existing workflow.
When the feed-in tariff is high, the system feeds power into the grid. If the tariff is low or even negative, the battery is charged instead. This lets us feed solar power into the grid exactly when it’s economically worthwhile – while relieving strain on the power grid during production peaks.
Why the project worked
“Real software comes from engineering work, not from free-form prompting. The ZEV solution proved that we can build production software in two weeks – and we’re now passing that same edge directly on to our first customer projects.”Urs Nacht — Member of Executive Management & Senior Engineer, Glue AG
Custom software is becoming attractive again
The ZEV use case shows what becomes possible with AI Software Engineering: solutions precisely tailored to a specific process – built in weeks instead of months.
Do you have a process or an idea you’d like to turn into software? We’ll advise you with no obligation.