W
hen meeting with healthcare leaders about language access, compliance, and safety, I often hear they’ve “figured out” language barriers because they have some bilingual staff or use an app. A few have even said that AI will soon “streamline” everything — including language access — by replacing most interpreters and translators.
Here’s the issue with that assumption: Language access is not a logistics problem. It is a civil rights obligation and, for the person standing in front of us, a lifeline. When a consent form is mistranslated or a plea deal is “explained” through an app, the question is not only, “Is the wording correct?” It is, “Whose rights were just put at risk?”
Recently, in a district court, a public defender asked me to help after my hearing. When I returned, she said, “No need. I already explained the plea deal using Google Translate.” The limited English proficiency (LEP) client had already left.
I was left wondering: Did he leave because he understood or because he didn’t? Technology can’t read confusion, hesitation, or the meaning behind a polite nod. In legal settings, those missed signals can mean the difference between understanding and losing one’s rights.
Advertisement
Moments like these led to a deep professional identity crisis for me. Not a small, quiet doubt, but a disorienting one.
That incident forced a question I couldn’t escape: Why is AI “good enough” when the person on the receiving end speaks another language, but it would never be considered good enough for us? Most people are not fine with the idea of an AI tool giving them a life-altering piece of news. It is a question about whose rights institutions actually value — and I’m still waiting for an honest answer.
I found part of that answer at the VAMOS JUNTOS Conference in Buenos Aires. José Palomares, senior director of global experiences at Coupa, said, “Complexity is your business model.” AI has already taken over much of the simple, repeatable language work. What remains — the work that still needs us — is the messy part: cultural ambiguity, regulatory gray areas, fragile human relationships, institutional trust.
When he talked about “complexity,” I heard “language access infrastructure.” Our work as language service providers lives at the intersection of civil rights, culture, trauma, safety, and community trust. No two encounters are the same: an undocumented mother in an emergency room, a Deaf person in Puerto Rico signing in a language that is not American Sign Language (ASL) and who is more likely to read Spanish than English, an elderly Vietnamese woman in a housing hearing whose nephew interpreted for her but is also her abuser.
Palomares put it this way: Executives do not buy linguistic quality; they buy the story of a prevented disaster or unlocked growth. When it comes to language access, we as providers need to help decision-makers see language access as the infrastructure that prevented a possibly fatal medical error and the consequential lawsuit and loss of trust.
For those of us running language services companies or departments, this is our real value proposition. We are not just selling words per minute or lines translated. We are selling managed complexity.
Another part of the answer came from Juan Santiago from Santex, who outlined four elements that could help companies stand out in the world of AI. I extrapolated these four pillars to language access to explain why human professionals remain indispensable.
The first pillar is judgment, or recognizing what a situation truly requires. When I learned Google Translate had been used to explain a plea, no system flashed a warning. A human had to say, “This is not okay. This violates language access.”
Second is cultural leadership, or bringing real community knowledge into the room. Language access is not just words; it is how meaning lives inside a culture, a migration story, and a power imbalance. It is knowing when a nod means compliance instead of comprehension, when silence means shame, or when a community’s history with institutions means they will not ask questions even when they are lost.
The third pillar is agility — not just speed, but the ability to change course in the moment. I had a situation in court when, halfway through a hearing, I realized that Spanish was not the respondent’s first language. He spoke K’iche’. The public defender hadn’t caught that detail. A human interpreter can pause, communicate, advocate, and redirect the process so the person can actually understand. A machine just keeps going.
Finally, there’s governance: deciding when and how language services are used and who is accountable when they fail. Governance is the difference between “we have some bilingual staff” and “we have a written plan, resources, and accountability for language services.” People choose it and defend it.
Language access is not a “nice-to-have.” In the United States, federal civil rights laws and court decisions point to the same reality: Meaningful language access is a civil right in healthcare, in the justice system, in education, and in access to public benefits. Not a courtesy. Not an upgrade. Not charity. A right.
The person affected often does not know those rights exist. That is why trained human professionals with a code of ethics are not optional. We are the safeguard. Language service providers design, staff, and defend that safeguard.
AI can help us translate more content, catch patterns, and speed up low-risk work. It can be a powerful tool inside a thoughtful language access plan, and it can expand access when used to draft low-risk materials that are then carefully reviewed by human linguists. But it cannot take responsibility for the life-altering consequences of misunderstandings in a hospital, a school, or an immigration office. Only humans with judgment and ethics can carry that responsibility.
As providers, we will often be the ones asked to “make AI work” for language access. That means designing models of service where AI is tightly governed and clearly limited in high-stakes settings, not simply added on top of already fragile systems.
Advertisement
The reality is that some tasks cannot be trusted to automated processes. Humans must lead when communication can affect rights, health, safety, income, education, or immigration status. They must lead when content shapes public trust: legal notices, court communications, public health messages, school communications, policy changes, benefits information. And they must lead when the communities being served define the message as high-stakes.
My identity crisis has not completely disappeared, but I have found my footing again, not in what AI cannot do, but in what I have chosen to do anyway.
We need to learn how to work alongside AI — but we must not dull our reasoning, our judgment, our ethics, or our cultural intelligence in the process. We need to sharpen them. The pillars that make us irreplaceable are not relics of a pre-AI world; they are precisely what this moment demands.
Language access is infrastructure. It is law. It is the daily practice of asking, “Who is not understanding this, and what are we going to do about it?”
Our work begins where the output of a model ends: in the human space where rights, fear, culture, and power collide. Budgets, contracts, and enforcement must reflect those values. Repairing the tears of our community depends on them, and that space still needs us.
Carol Velandia is a nationally recognized advocate for language access in the United States. She is founder and CEO of Equal Access Language Services, and she developed the award-winning program “Effective Inclusion Through Language Access” to enhance language service delivery across various sectors.
Advertisement
Related Articles
A paper recently published in the American Journal of Medical Care is shedding light on the importance of providing language access services in the healthcare…
→ Continue ReadingIn an attempt to improve its communication efforts with multilingual families and students with limited English proficiency (LEP), New York City is investing $4 million…
→ Continue ReadingThe report highlights where progress is being made, and where operational support is still needed, to help organizations and providers deliver language access reliably when…
→ Continue Reading
