AI makes it easier for people with disabilities to find work, but almost no one uses it to help them keep it, the ILO warns

By Titularísimos · September 14, 2026.

The International Labour Organization (ILO), the United Nations agency that sets global labor standards, has issued a specific warning about the digitalization of employment: automating employment services does not, on its own, guarantee greater inclusion for people with disabilities. It can even create new barriers if accessibility is not built in from the design of the tool. The conclusion comes out of a European meeting held in Berlin on September 1 and 2, with the participation of public employment services, organizations of people with disabilities and specialized professionals from several countries, as reported by the ILO itself.

There are already concrete examples on the ground. Greece is testing a job-matching system based on each person's actual skills, to connect capabilities with companies' needs. Germany is experimenting with smart glasses and digital work instructions designed to adapt the job to someone who has already been hired. Added to that are AI systems for matching candidates with vacancies, support tools and automatic captioning, according to the catalogue of technologies the organization handles.

The ILO itself detects, however, an underlying imbalance: almost all technological development is concentrated in the access phase —job search, training, matching openings and candidates— and very little on what comes afterwards. The ILO calls for more tools for reasonable accommodations (the workplace adaptations the law requires when they are reasonable and not disproportionate), job retention and career advancement. There is plenty of AI to open the door and almost none to accompany those who have already come in.

There is an additional nuance that often goes unnoticed: not all disabilities are visible, not even to the person who has one. Brussels' public employment service, Actiris, cited as an example young people with hearing loss who are unaware of their own condition, something that shapes their relationship with employment services long before any algorithm gets involved. The European Federation of Hard of Hearing People, for its part, argued for low-tech-cost measures —automatic captioning, flexible hours, accessible onboarding processes— against the temptation to solve everything with more sophisticated systems.

The ILO frames this work within PROSPECTS Digital Opportunity Accelerator, its own project that over the course of 2026 has gathered experiences from public employment services in different countries: accessible platforms, remote support with sign language, remote work and digital follow-up after hiring. For now it is a mapping of practices, not a settled standard.

This analysis connects with something we have been repeating as we analyze the impact of AI on employment sector by sector —technology, healthcare, banking, education, tourism, law, public administration—: the effect is never uniform, it depends on who designs the tool and with whom in mind. What the ILO adds here is an axis that cuts across all those sectors. If recruitment software, vacancy matching or performance evaluation systems do not incorporate functional diversity from the first draft, the result is not neutral: it is automated exclusion, with no discriminatory intent behind it, just an incomplete design that nobody audits until someone is left out.

In the short term we have to be honest about the risk: automated screening, video interviews analyzed by AI or platforms that assume a standard interaction with keyboard, screen and voice can leave out anyone who does not fit that mold, precisely when technology is most needed to widen access to employment rather than narrow it. The ILO is not exaggerating: it is a problem of engineering and investment priorities, not a distant hypothesis.

In the long term, however, the direction the organization points to —automatic captioning, digital work instructions, smart glasses— is exactly the kind of AI application we are most interested in highlighting at Zendoric: technology that does not replace anyone, but expands who can participate fully in the labor market. It is a concrete example, already verifiable and not speculative, of the abundance we defend as this technology's horizon: when cheap, intelligent assistance covers what previously required constant and costly human support, capacity is freed up for people to work on what they want and for organizations to stop managing nothing but paperwork.

There is also a side effect worth underlining because it is rarely told: solutions designed to remove accessibility barriers almost always end up improving the service for the entire workforce, not just for people with disabilities. Captioning helps in any noisy environment; simplified digital instructions help anyone on their first day at work. Designing for the margin, once again, ends up improving the center. That detail, more than the warning itself, is the one that should stick in the conversation about how the employment services of the future are built.