Generative AI has moved quickly from curiosity to boardroom priority in healthcare. 

The technology is already being tested, piloted, and deployed across clinical, operational, administrative, and patient-facing workflows. In a fourth quarter 2025 McKinsey survey of 150 U.S. healthcare stakeholders, 50% of respondents said their organizations had implemented generative AI, up from 25% in a similar 2023 survey. McKinsey also reported that all respondents had at least some plans to pursue generative AI, a sign that healthcare organizations are moving beyond whether to use it and into how to use it responsibly. (McKinsey & Company) 

That shift is exciting, but it also creates confusion. Generative AI is often discussed as if it can do everything, from answering patient questions to writing notes to coordinating care. At the same time, healthcare leaders are right to be cautious. A tool that sounds impressive in a demo still has to perform securely, accurately, and reliably inside real healthcare workflows. 

The reality is that generative AI is neither magic nor meaningless. It is a powerful technology with practical applications, real limitations, and a growing need for governance. 

For urgent care organizations, the opportunity is not to chase the most futuristic version of AI. It is to identify where generative AI can reduce burden, improve communication, and support teams without taking control away from the people responsible for care. 

What Is Generative AI in Healthcare? 

Generative AI refers to technology that can create new content based on patterns in data. In healthcare, that content might include summaries, draft messages, structured notes, patient instructions, internal knowledge responses, or other text-based outputs. 

The World Health Organization describes large multimodal models as a fast-growing form of generative AI that can accept inputs such as text, videos, and images, then generate outputs that are not limited to the same type of input. WHO also notes that these models can mimic human communication and perform tasks they were not explicitly programmed to complete. (World Health Organization) 

That ability makes generative AI different from traditional automation. Automation typically follows a defined rule: if this happens, do that. Generative AI can interpret language, summarize context, draft a response, or create an output that feels more flexible. 

In healthcare that flexibility is valuable because so much of the work depends on language. Patients describe symptoms in their own words. Providers document care in notes. Staff answer recurring questions. Teams interpret information across systems. Generative AI can help turn messy, conversational information into something more structured and usable. 

But that same flexibility is also why oversight matters. A generative AI tool can produce an answer that sounds confident without being complete or correct. In healthcare, that distinction is critical. 

Why Generative AI Is Getting So Much Attention 

Generative AI is gaining traction because healthcare is full of high-volume, language-heavy work. 

Providers document visits. Staff manage intake questions. Patients need instructions. Teams summarize information across encounters. Leaders try to make sense of operational patterns. Much of this work is necessary, but not all of it requires a human to start from a blank page every time. 

That is where generative AI can help. It can create a first draft, summarize a conversation, organize information, or suggest language that a human reviews before it is used. 

Physicians are already beginning to see the value. In the AMA’s 2026 Physician Survey on Augmented Intelligence, more than 80% of physician respondents reported using AI in a professional context, double the share reported in 2023. More than three-quarters said AI provides an advantage in the ability to care for patients, though 40% said they remain equally excited and concerned. (American Medical Association) 

That mix of optimism and caution is exactly where healthcare should be. Generative AI can be useful, but it should not be accepted uncritically. The best use cases are the ones where the technology supports people, reduces repetitive effort, and leaves final judgment where it belongs. 

What Generative AI Can Do Well 

Generative AI is strongest when it is used to support language-based tasks that are repetitive, time-consuming, and reviewable. 

In healthcare, that includes work like summarizing information, drafting communication, organizing documentation, and helping people find answers from approved sources. WHO identifies several broad applications for large multimodal models in health, including clerical and administrative tasks, patient-facing communication, medical education, research, and support for clinical care. (World Health Organization) 

For urgent care, the most practical opportunities tend to fall into a few categories. 

Summarizing information 

Generative AI can help summarize long conversations, patient histories, instructions, or internal information so people do not have to manually extract the main points. This can be useful when teams need to move quickly but still understand context. 

A summary should never replace review, especially in clinical settings. But it can reduce the time it takes to get oriented, identify what matters, and decide what needs attention next. 

Drafting content 

Generative AI can draft messages, instructions, documentation, responses, and other written outputs. This is one of its clearest strengths because the output can be reviewed, edited, and approved before it is finalized. 

In healthcare, that review step matters. A drafted patient message may save time, but it still needs to be accurate, appropriate, and aligned with the organization’s standards. 

Structuring unstructured information 

Healthcare is full of unstructured information: conversations, notes, call transcripts, patient questions, and free-text entries. Generative AI can help turn that information into a more organized format. 

This does not mean the technology understands care the way a clinician does. It means it can help organize language so humans can work with it more efficiently. 

Supporting routine interactions 

Generative AI can help answer common questions, guide patients through routine steps, and support staff by handling repeatable communication. In urgent care, that may help teams respond more consistently during high-volume periods. 

The key is to define what the tool can and cannot answer. Routine questions are very different from clinical decision-making, and the guardrails should reflect that. 

What Generative AI Should Not Do on Its Own 

The biggest mistake healthcare organizations can make is treating generative AI as a substitute for human judgment. 

Generative AI can draft, summarize, and assist. It should not independently diagnose patients, make care decisions, override clinical judgment, or operate without accountability in high-risk workflows. 

WHO warns that large multimodal models can produce false, inaccurate, biased, or incomplete statements. The organization also highlights risks related to poor-quality training data, bias, automation bias, cybersecurity, privacy, and misplaced trust in AI-generated outputs.  

Those risks do not make generative AI unusable. They make governance essential. 

In practical terms, healthcare organizations should be cautious about using generative AI for: 

• Autonomous diagnosis 
• Unreviewed clinical recommendations 
• Patient-specific medical advice without appropriate oversight 
• Finalized documentation without provider review 
• Communication that could create confusion, fear, or inappropriate next steps 
• Workflows where no one is accountable for the output 

Generative AI works best when humans remain in control of decisions, approvals, and exceptions. 

Why Trust Is the Real Adoption Challenge 

The next phase of generative AI in healthcare will not be defined only by what the technology can produce. It will be defined by whether healthcare teams trust it enough to use it. 

Trust requires more than accuracy in a controlled demo. It requires transparency, reliability, workflow fit, privacy protections, clear accountability, and evidence that the tool improves the work it is supposed to support. 

That is why national guidance is beginning to focus less on AI as a novelty and more on governance. NIST’s AI Risk Management Framework is intended to help organizations incorporate trustworthiness into the design, development, use, and evaluation of AI systems. NIST also released a Generative AI Profile in July 2024 to help organizations identify and manage risks unique to generative AI. (NIST) 

Healthcare-specific governance is also becoming more defined. The National Academy of Medicine’s AI Code of Conduct project is focused on helping AI in health, healthcare, and biomedical science perform accurately, safely, reliably, ethically, and in service of better health. (NAM) 

For on-demand healthcare leaders, this means AI evaluation cannot be limited to features. The question is not only, “What can this tool do?” It is also, “Can we trust how it does it, where it fits, and how our team stays in control?” 

How to Evaluate Generative AI in Healthcare 

Generative AI should be evaluated like any other operational or clinical investment: based on the problem it solves, the risk it introduces, and the value it can prove. 

The strongest starting point is not the technology. It is the workflow. 

Healthcare organizations should ask where teams are losing time, where communication breaks down, where information is being repeated, and where staff or providers are starting from scratch too often. Those are the areas where generative AI may offer practical value. 

A responsible evaluation should include questions like: 

• What specific workflow problem does this solve? 
• Who reviews or approves the AI-generated output? 
• What happens when the tool is wrong or uncertain? 
• What data does the tool use, and how is that data protected? 
• How does the tool integrate with existing systems? 
• What guardrails prevent inappropriate use? 
• How will performance be measured over time? 
• Does the vendor understand the pace and complexity of urgent care? 

Healthcare organizations should also look for evidence that the tool can work in real settings, not just controlled demonstrations. ONC’s HTI-1 final rule reflects the broader direction of healthcare AI oversight by emphasizing transparency for predictive models and algorithms used to aid decision-making in certified health IT. The rule includes requirements related to source attributes, risk management, governance, and ongoing maintenance for certain predictive decision support interventions. (Federal Register) 

Generative AI tools may not all fall into the same regulatory category, but the direction is clear: healthcare AI is moving toward more transparency, not less. 

What Responsible Adoption Looks Like 

Responsible generative AI adoption does not require healthcare organizations to transform everything at once. In fact, the safest path is often narrower. 

Start with a use case where the risk is manageable, the output is reviewable, and the workflow pain is clear. Then measure whether the tool actually improves the experience for staff, providers, patients, or operators. 

For urgent care, that might mean starting with tasks that reduce repetitive language work, improve consistency, or help teams move information faster through the visit. The details will vary by organization, but the principles should remain consistent. 

Responsible adoption should: 

• Keep humans in the loop 
• Make review and correction easy 
• Protect patient data 
• Define where AI can and cannot be used 
• Train teams on appropriate use 
• Measure outcomes before expanding 
• Reassess performance over time 

The goal is not to use generative AI everywhere. The goal is to use it where it can make work easier without making care riskier. 

What Comes Next for Generative AI in Healthcare 

Generative AI is likely to become less visible as it becomes more useful. 

Instead of sitting in separate tools, it will increasingly be embedded into the systems healthcare teams already use. It may help draft, summarize, route, and coordinate work across multiple steps of the patient journey. 

McKinsey’s 2025 survey points to this next phase, noting growing interest in agentic AI, where systems move beyond generating content to coordinating more complex processes. In that survey, 19% of respondents said their organizations had reached agentic AI maturity, while another 51% were pursuing proofs of concept. (McKinsey & Company) 

That future has potential, but it also raises the stakes. The more AI moves from drafting content to taking action, the more important governance, oversight, and workflow design become. 

For urgent care, the future is not about removing people from care. It is about reducing the work around care so teams can focus more fully on the moments that require human judgment, empathy, and accountability. 

Generative AI Is Useful When It Is Grounded 

The most important thing to understand about generative AI in healthcare is that usefulness depends on context. 

A generic tool may be impressive, but healthcare requires more than impressive output. It requires accuracy, privacy, oversight, integration, and trust. It requires workflows that make sense for the people using them and safeguards that reflect the stakes of care. 

Generative AI is already changing healthcare, but the organizations that benefit most will not be the ones that adopt it fastest. They will be the ones that apply it carefully to real problems, measure whether it works, and keep people at the center of the process. 

That is where generative AI becomes practical. 

Not as a replacement for providers, staff, or care teams, but as a way to reduce the repetitive work that keeps them from doing their best work. 

See how practical AI is taking shape across urgent care. 

Step Inside Experity’s Clinic of Tomorrow