AI won't cut civil servants: incentives favor growth, not savings

AI won't cut civil servants: incentives favor growth, not savings

Will artificial intelligence empty government offices as it does industrial plants? The short answer is no, and the reason has nothing to do with technology. In the private sector, those who don't automate disappear; in public administration, those who manage more personnel and budget accumulate more power. The incentive structure is inverted. A department with fifty employees and a five-million-euro budget is more influential than one with ten staff and one million. Efficiency, in this context, is literally an enemy.

The shield protecting public jobs from automation

The core argument is simple: the administration doesn't compete. There are no rival companies arriving with AI agents to steal market share. This structural protection explains why AI adoption in the public sector won't lead to staff cuts, but rather to other outcomes. Less future replacement, perhaps. Task reallocation, likely. Layoffs, almost never.

Some go further, arguing that AI will increase bureaucracy. They cite the parallel of computing and ERP systems: when spreadsheets arrived, it was expected that one accountant could do the work of several. Instead, administrative staffing needs grew. Technology freed up capacity, and the system absorbed it by creating more control layers.

Which public tasks AI can handle and which it cannot

Not everything is immune. There are tasks AI already performs better than humans: data tracking, information cross-referencing, and detecting tax irregularities. Here, the potential for improvement is huge. The problem arises when those affected by AI have status, lawyers, and the ability to appeal. An agent that breaches data protection or accesses information without a judicial order can trigger political fallout. This hinders ambitious implementation.

In social services and employment agencies, some pilots have been rolled back due to system rigidity. The most striking example cited involves automated processes summoning patients with serious diagnoses for review without human filtering. AI doesn't distinguish between a file and a person.

Implemented AI: tenders, consultants, and mediocre results

When the administration buys technology, the result rarely matches the PowerPoint promises. Tenders are poorly drafted, awarded to the cheapest bid, and final designs are signed off by unsuitable profiles. The outcome is a useless tool for the actual task, but perfect for justifying expenditure.

AI for legal consultations is the most repeated example. A system can competently explain the difference between two legal concepts and then invent a BOE article with a fake link. When pointed out, it denies the error. It took four attempts for it to admit it had made it up. In law, this isn't a minor bug: it's a liability generator.

The real cost: who pays for administrative expansion

The public sector doesn't compete, but it isn't free either. Staff and budget expansion is funded through more regulation for the private sector, higher taxes, and debt when taxes aren't enough. The cycle always closes the same way: more procedures justify more officials to oversee them, and higher public payroll demands more revenue. The host endures, but weakens.

Added to this is a pending debt absent from headlines: pensions. Medium and long-term equations force higher collection—already limited—or cuts elsewhere. AI doesn't change this arithmetic. At best, it disguises it.

What happens to vacancies: less replacement, not layoffs

The most likely scenario isn't an administration without civil servants, but one that replaces fewer people. Retirements aren't covered, positions are gradually eliminated, and AI becomes an excuse not to call competitive exams (*oposiciones*). Those already inside have little reason to antiestéticar. Aspirants have much more.

Meanwhile, national productivity still depends on the competing sector. And that sector has no shield. The paradox is that AI will arrive sooner and stronger where it can destroy jobs, not where it could save public money. The State has the technology, but lacks the incentive. And ultimately, the incentive always wins.