OpenAI AI Agents Priced to Replace Engineers

OpenAI's AI Agent Pricing Threatens Engineering Jobs

The scene is cinematic: in a closed meeting with investors, OpenAI reveals its roadmap for autonomous AI agents. These programs can sit at a computer, open applications, and execute tasks from start to finish without human supervision. The price list is already on the table, and the lower figures are chilling: €2,000 per month for the basic tier and €20,000 for research. Officials warn that the price is for the service, not per professional. But the initial impact is what counts.

US-Calculated Pricing Clashes with Spain

The comparison with the Spanish labor market is devastating. With €20,000 a month, an entire research department could be paid for in Spain. “With those prices you hire four engineers and three PhDs,” summarizes one circulating estimate. The immediate conclusion from many analyses: these prices are designed for US salaries, where a computer scientist’s base pay doesn’t drop below that range. In Spain, the same money goes much further. For now, the threat deflates as soon as it crosses the Atlantic.

Smoke or the End of a Profession: Two Opposing Views

The range of opinions allows no middle ground. Some see OpenAI as another PowerPoint on steroids: the €200/month Pro plan hasn’t justified services like DeepResearch or Operator, and the first commercial agent launched with more pain than glory. The label of high-quality smoke is repeated ad nauseam. At the other extreme, a minority warns: paid agents have nothing to do with free ChatGPT. They are much larger and more expensive models, with capabilities comparable to humans, and the billions invested by Google and Microsoft aren’t just whims. When the agent produces what an entire bank department does, it will cost what that department costs.

The Real Face of AI in the Office

Practical examples deflate the headline. Asking ChatGPT to code something simple can take longer than doing it manually: it proposes flawed solutions, requires corrections, and the afternoon disappears. For many queries, official documentation or Stack Overflow remain faster. The experience of using AI as a work companion to summarize poorly organized technical documentation is positive but doesn't replace human judgment. The case of an executive who used remote workers to simulate what the AI should do alone has become the perfect metaphor for an industry selling smoke, even if it’s cheap smoke.

Automating the State and Waiting for the Big Launch

The discussion also jumps to the public sector: an agent for administrative procedures would allow dispensing with 200,000 civil servants, according to a recurring estimate. The idea has defenders, but it clashes with reality: mass automation isn’t just a technical problem, but a political decision. Meanwhile, outsourcing to countries with low wages remains cheaper than renting an OpenAI agent. The promise of replacing engineers en masse still hasn’t materialized as of 2026. The next launch, announced for Thursday of that week, will be the acid test.

The data that throws people off is precisely this: the salary of a high-end agent pays for an entire department in Spain, yet companies continue hiring humans. Universities and vocational training centers shouldn’t celebrate prematurely, nor pack their bags. The distance between the commercial brochure and the office remains huge.