A Perspective from the Field
Over the past few months at No Barrier, I’ve found myself going deeper into the world of language access and the growing role of AI in healthcare interpretation. There’s no shortage of material. High-level ideas, familiar arguments, the same promises repeated without always grounding them in what actually happens in care settings.
Then I came across “AI and Language Access in Healthcare” by Katy Haynes and it felt different. Haynes, a Bay Area–based consultant working at the intersection of digital strategy, health technology and public sector innovation, brings a perspective shaped by real-world implementation. The piece doesn’t just talk about the potential of AI. It connects it to the realities of healthcare systems in a way that’s hard to ignore.
The work was developed for the California Health Care Foundation (CHCF), an independent nonprofit dedicated to improving the healthcare system so that all Californians, especially those with low incomes and communities facing the greatest barriers to care, can access the care they need.
It’s thoughtful without being theoretical. It’s structured in a way that reflects the complexity of the space, covering not just the technology but also the realities of providers, payers, regulators and patients. It’s the kind of work that helps move the industry forward, both in terms of understanding and in terms of possibility.
It also resonates with what we’re seeing on the ground.
At No Barrier, we come to this conversation with a specific lens. Our background is rooted in over a decade of work in voice technology and AI. More importantly, we are in constant dialogue with clinicians and health systems that are already using AI-powered interpretation in real care environments.
This is not a pilot phase. It is not a future scenario. Our clients are using these tools every day, in real interactions with patients.
That creates a useful perspective.
The goal here is not to challenge the recommendations but to look at how they translate into practice. What works as expected. What becomes more complex in real-time clinical settings. Where the opportunities are clear and where more work is still needed.
Because ultimately, the conversation around AI and language access is not just about frameworks. It’s about what actually happens when a patient and a provider need to understand each other, in the moment.
If you’re looking to balance your perspective and better decide how to approach your language access strategy through the lens of technology, this piece is for you.
- Where AI delivers immediate value, especially in low- and medium-risk interactions
- How AI is currently being used in language access across real healthcare settings
- The limits of AI today, particularly in emotional contexts
- How regulatory frameworks are evolving and what they mean in practice
- Why a risk-tiered approach makes sense in theory and where it gets more complex in reality
- How clinicians actually experience AI interpretation in day-to-day care delivery
- What “human in the loop” looks like beyond policy language
- The trade-offs between cost, speed and quality and how to navigate them
- Why language nuance, dialects and cultural context remain critical challenges
- How organizations can think about governance, patient choice and responsible deployment
AI and Language Access in California Health Care
A New Chapter, Not a Replacement
California has long stood at the forefront of linguistic diversity in the United States. With more than 200 languages spoken across the state and nearly half of residents speaking a language other than English at home, language access is not a niche concern. It is central to delivering equitable care.
Organizations like the California Health Care Foundation (CHCF) have played a key role in advancing this conversation. Through research and collaboration with policymakers and innovators, they are helping the industry better understand how emerging technologies, including AI, can support more inclusive care delivery.
Today, a new wave of AI-enabled language tools is beginning to reshape how providers approach interpretation and translation. The question is no longer whether AI will play a role (human vs AI) but how to integrate AI responsibly.
From Traditional Vendors to AI-Enabled Ecosystems: A Changing Landscape
The limits of the traditional model
For years, health plans and providers have relied on a relatively small group of language access vendors, often operating remotely. While effective, this model has limitations. Not every organization can afford in-house interpreters and access can be inconsistent, especially for less common languages, remote areas, shifts and night hours.
The emergence of AI-powered tools
AI is beginning to expand what is possible. From real-time translation and interpretation to automated written communication like AI scribes, these tools offer faster turnaround times and the potential to scale language access in ways that were previously out of reach.
Early adoption across health systems has focused primarily on low-risk use cases, such as administrative communication or basic patient instructions. These applications are often offered at no cost to patients, which further supports access.
A shift toward hybrid models
Rather than replacing existing systems, AI is creating a more layered ecosystem. Traditional vendors, internal staff and AI tools are increasingly working together, each playing a role depending on the context.
Where AI Adds Value Today: Speed, Scale and Reach
Expanding capacity in constrained systems
One of the clearest benefits of AI is its ability to extend limited resources. Many health systems face shrinking budgets and a shortage of qualified interpreters. In this context, AI is often seen less as a cost-cutting tool and more as a way to increase capacity.
It can help reduce delays, support high volumes of interactions and make it easier to serve patients in languages on the spot (not talking about languages that are often overlooked).
Supporting written communication
Much of the current momentum is in written translation, where AI can assist with patient instructions, follow-up messages and educational materials. These workflows allow for review and validation, making them well suited for AI-assisted processes.
Reaching underserved languages
AI also opens the door to broader language coverage. However, performance is not uniform. A system that performs well in widely spoken languages like Spanish may struggle with more nuanced or less-resourced languages such as Hmong or Marshallese.
At No Barrier, this is where a more granular approach becomes essential. Language is not just about translation. It is contextual, not linear. It includes dialects, cultural context and nuance. For example, Spanish alone can vary significantly across regions, which is why more tailored approaches are needed to truly support patient understanding. For instance, No Barriers offers 5 nuances of Spanish.
The Guardrails That Matter: Human Oversight and Legal Boundaries
AI is not a replacement for human interpreters
Current federal and California regulations are clear. AI cannot serve as a standalone substitute for qualified medical interpreters in clinical, legal or consent-related situations.
It can support workflows but human involvement remains essential in contexts where accuracy and accountability are critical.
The importance of “human in the loop”
Most emerging frameworks emphasize a human-in-the-loop model, where AI outputs are reviewed or supported by trained professionals. This approach helps balance efficiency with safety.
It also reflects a broader understanding that both humans and machines can make errors. What matters most is the context in which those errors occur.
Rethinking what is “critical”
As care increasingly moves to digital channels like patient portals and messaging, regulators are being challenged to redefine what counts as “vital” versus “non-vital” communication.
This distinction will shape how and where AI can be safely used.
Balancing Cost and Quality: Efficiency Without Compromise
The promise of lower costs
AI has the potential to significantly reduce the cost per interaction, particularly in lower-risk scenarios. In fact, most interactions revolve around scheduling and patient history. This idea also comes up in a conversation with Dr. Yosef Berlyand on Care Culture Talks. This could make language access more sustainable for many providers.
The risk of over-prioritizing efficiency
At the same time, there is a growing awareness that efficiency should not come at the expense of quality. In healthcare, even small misunderstandings can have meaningful consequences depending on the situation.
Supporting overstretched staff
Another important dimension is the burden placed on bilingual staff, who are often asked to step in informally as interpreters. AI can help alleviate this pressure by providing support tools, allowing staff to focus on their primary roles.
Moving Toward Risk-Based Frameworks: Beyond Binary Thinking
From “AI vs human” to contextual decision-making
One of the most promising ideas emerging from current discussions is the move toward risk-tiered frameworks. Instead of treating language access as a binary choice, these models adapt the level of human involvement based on the situation:
- High-risk interactions rely fully on human interpreters
- Lower-risk interactions may incorporate AI, often with review mechanisms
Where this works well
This approach is particularly well suited for workflows that are not time-sensitive, such as document translation or follow-up communication, where there is time for validation.
Where reality is more complex
In practice, however, risk and timing do not always align.
Clinicians often need immediate communication, especially in urgent or high-pressure environments.
From a No Barrier perspective, this highlights an important nuance. Instant access to communication can be just as critical as the level of risk. In some cases, waiting for a human interpreter may not be feasible.
Two practical principles can help guide decisions:
- When possible, give patients a choice in how they receive language support
- Prioritize human interaction for emotionally sensitive conversations
Building Responsible AI Governance: Structure, Accountability and Trust
Formalizing evaluation and oversight
Health systems are beginning to take a more structured approach to AI adoption. Some are creating governance frameworks to evaluate vendors, define acceptable use cases and establish quality assurance standards. This includes clear protocols for escalation when issues arise.
Tracking usage and outcomes
There is also a growing emphasis on transparency. Tracking how AI is used, across which languages and in what volume, helps organizations better understand its impact and identify areas for improvement.
Equity and the Road Ahead: Investing Where It Matters Most
Supporting less common languages
Ensuring equitable access means going beyond the most widely spoken languages (Spanish for instance). There is a clear need to invest in both human interpreter pathways and AI systems that can support less common and indigenous languages.
The challenge of scarcity
At the same time, the industry faces a structural challenge. The same scarcity that affects human interpreters also impacts AI development, since high-quality models require training, validation and cultural expertise.
A shared responsibility
Addressing this gap will require collaboration between funders, policymakers, technology providers and communities.
Conclusion: Responsible Innovation in Language Access
AI is opening new possibilities for language access in healthcare. It offers speed, scalability and the potential to reach patients who have historically been underserved.
At the same time, its role must be carefully defined. In practice, risk and timing do not always align and clinicians often need immediate communication, especially in urgent or high-pressure environments. This makes reliability, responsiveness and clarity critical in real-world use.
Human interpreters remain essential, particularly in high-stakes and emotionally complex situations.
The path forward is not about choosing between human and machine. It is about designing systems where both can work together effectively.
With the right guardrails, thoughtful governance and a continued focus on patient experience, AI can help extend access without compromising quality.
That is where the real opportunity lies.
.svg){kind=small}
%201.png)