During decades, public IT procurement has measured value in people and hours. A model that assumed that producing software was, almost exclusively, a matter of labor cost. Today that assumption is obsolete.
Generative AI enables more output, better quality, and faster delivery without increasing team size. Value no longer resides in the volume of deployed resources, but in real capability, software assets, and delivered outcomes.
The paradox of public procurement
Here a critical problem emerges: an AI-driven offer can be extremely competitive and, at the same time, be rejected for not meeting the minimum hour threshold defined in the tender. What the system identifies as a abnormally low bid may in fact be real, demonstrable efficiency.
Evaluation frameworks based on effort penalize innovation. Public administrations need to adapt to an environment where team capacity no longer depends on size, but on expert talent and the intelligent adoption of AI.
Algorithmic transparency: a legal obligation
Adopting AI in the public sector is not only a matter of efficiency. It is a matter of legality and citizen guarantees. Law 39/2015 requires the Administration to justify its decisions. When those decisions are made by opaque algorithms, legal defenselessness arises.
The BOSCO software case illustrates this clearly: the Spanish Supreme Court, in its Judgment 7878/2024, determined that when an algorithm generates rights, citizens may inspect its code. Public AI cannot be a black box.
The EU Regulation 2024/1689 follows the same direction: AI tools supporting public procurement processes are High-Risk Systems, with concrete obligations regarding transparency, traceability, human oversight, and prior audit.
The new paradigm of IT tenders
Tender documents must evolve. The traditional model — hours, mechanical supervision, opaque auditing — gives way to a model oriented toward assets, outcomes, and algorithmic transparency.
This implies incorporating new requirements: risk management across the entire lifecycle, representative and bias-free data, exhaustive documentation of algorithmic decisions, and human oversight with real intervention capability. Cities such as Amsterdam and Helsinki, or initiatives like Etalab in France, have already established public algorithm registers that point the way forward.
Responsible AI: speed with legal rigor
There is an approach that does not consist of providing more human resources, but of building more efficient and auditable execution models, while maintaining full legal compliance at all times. This is precisely the direction taken by Axpe Consulting in supporting public administrations through this transition.
Three levers make this possible: diamond teams composed of experts who validate and supervise AI agents; industrialization with SLMs and explainable architectures by design that ensure traceability of every decision; and full alignment with EU Regulation 2024/1689 and Spanish Administrative Law.
The new contract between technology and government
The old world bought hours, measured volume, and accepted opacity. The new world procures capability, measures outcomes, and demands transparency.
AI enables unprecedented efficiency, but it only creates real public value if accompanied by strong guarantees of explainability and human control. The future of public IT procurement is not measured in hours. It is measured in impact.
