Quick answer: AI is already in many clinical workflows, mostly as assistance: ambient documentation that drafts notes for review, decision support like sepsis and deterioration alerts, diagnostic flagging in radiology and pathology, and patient flow tools. It is reducing documentation burden, while clinical judgment, presence with patients, and relationships stay with clinicians. The most useful response is informed, curious engagement.
I have been using AI in our EPIC system for a few months to help summarize patient info, it is not always 100% perfect in mentioning all important events, but it is very close. And it has saved me particularly in spending so much time reviewing charts of patients who has been in hospital for a long time.
The ambient documentation tool went live in my hospital earlier this year.
Some of my colleagues were relieved. Some were suspicious. Most were somewhere in between, doing what busy clinicians do with a new system: learning just enough to get through the shift, then moving on to the next thing.
Nobody sat us down and gave us an honest framework for thinking about what AI actually is in our clinical environments right now, what it is not, and what a thoughtful working clinician should be doing about it.
That is the conversation I want to have in this post.
What AI Is Actually Doing in Healthcare Right Now
The category of “AI in healthcare” in 2026 covers a wide range of tools at very different levels of maturity and clinical impact. Being specific about which tools and what they actually do matters more than discussing AI as one monolithic thing.
Ambient documentation tools like Nuance DAX, Suki, and similar products are the most visible right now in clinical settings. These systems listen to the patient encounter and generate a draft note in real time. The physician or nurse reviews and approves. The early evidence suggests significant reduction in documentation time for the clinicians who use them well.
Clinical decision support tools have existed for years but are becoming more sophisticated. Sepsis prediction algorithms, deterioration alerts, medication interaction warnings integrated into the EHR workflow. These are already embedded in most major EHR systems, and most clinicians are interacting with them daily whether or not they think of it as AI.
Diagnostic support in radiology, pathology, and some imaging specialties is developing quickly. AI tools that flag potential findings for radiologist review are in active use at several major health systems. The model here is augmentation: the AI surfaces what to look at, and the clinician decides what it means.
Triage and patient flow tools are being adopted in emergency departments and larger hospital systems to help allocate resources and predict admission volumes.
What connects all of these is that they are assistive. The clinical judgment, the conversation with the patient, the decision about what to do next, those remain with the clinician.
What AI Is Not Doing
The question I hear most often from clinicians who are anxious about AI is some version of: will this replace me?
The honest answer, based on what the technology is actually doing in 2026, is no. The reason matters.
Clinical medicine depends on things AI cannot replicate. Presence in the room with a patient who is frightened. The ability to read what is not being said. The judgment that synthesizes years of training, pattern recognition, and the specific facts of the case in front of you. The relationship with a patient who has been coming to you for ten years.
AI tools work on data. The most meaningful parts of clinical care are not reducible to data.
What AI does replace, gradually, is documentation burden and some forms of pattern recognition at the level of information processing. For most clinicians, that is a feature rather than a threat.
The anxiety is understandable. Major institutional change almost always generates anxiety, and healthcare has had more than its share of change in recent years. But the fear that AI will make clinicians obsolete is not supported by what the technology is actually doing in clinical environments right now.
The Three Responses and Why One of Them Works
When a significant technological change lands in a professional environment, most people respond in one of three ways.
Panic looks like: reading every alarming article, assuming the worst, spending significant mental energy on a threat that is not yet materialized, and feeling powerless. This response is understandable and almost never useful.
Ignoring looks like: deciding that AI is someone else’s problem, learning the minimum required, and continuing to practice exactly as before. This response feels safer than panic but carries its own risk. Clinicians who disengage from the conversation about how AI is implemented in their specialty lose the ability to influence those decisions.
Integration looks like: getting curious rather than anxious, learning enough about the tools that are actually in your environment to use them well, forming a view about what AI should and should not do in your clinical context, and contributing to the conversation about how it gets implemented in your specialty.
Integration is harder than the other two and requires more intention. It is also the only response that positions you as someone who is shaping what healthcare looks like rather than someone it is happening to.
A Practical AI Toolkit for Working Clinicians
You do not need to become an AI expert. You need to know enough to engage usefully.
Learn the AI tools that are already in your EHR. Most clinicians are interacting with AI-assisted tools already without realizing it. Clinical decision support alerts, sepsis prediction flags, medication interaction warnings. Knowing which ones are there, how they were trained, and what their error rates look like puts you in a much better position than most of your colleagues.
Get specific about what you would want AI to do and not do in your clinical context. The clearer you are about this, the more usefully you can participate in institutional decisions about implementation. What would genuinely help you? What would get in the way? Where would you want a human in the loop, and where would automation serve your patients better?
Seek out one trusted source on healthcare AI specifically. Not general technology journalism, which tends toward both hype and panic. The AMA has published increasingly grounded material on AI in clinical practice. Academic medical centers with active implementation programs publish implementation findings that are more honest than vendor materials.
Try one ambient documentation tool if you have access to one. The learning curve is real. So is the documentation relief for most clinicians who work through it. You cannot form a grounded view about a technology you have not actually used.
The Leadership Question
There is a question underneath the practical toolkit that I think matters more than any specific tool.
Who is shaping how AI enters your specialty?
In most health systems right now, that answer is: administrators, vendors, and technology teams, with varying degrees of input from clinical staff. The less clinical input there is in those conversations, the more likely the implementation serves institutional cost and efficiency goals rather than patient care and clinician wellbeing.
Clinicians who are engaged, informed, and willing to say clearly what they will and will not accept in their practice environments are the people who shift those decisions. That engagement takes time and energy that most clinicians feel they do not have. But the alternative is that the decisions get made without you.
This is one of the most concrete applications of the kind of moral courage I wrote about last week. Showing up to the AI implementation conversation in your institution, with a specific and informed view, is exactly that.
What to Do Next
If you want to keep doing this work with other healthcare professionals, Rounds Society is the monthly membership I lead: a live coaching session every month, a private community that is open day or night, and short tools for the hardest moments of a shift.
If you are ready for a structured path, The Purposeful Doctor Bootcamp takes you through four recorded sessions with a workbook for each one, and you can start today.
Frequently Asked Questions
Will AI replace doctors, nurses, and other clinicians?
Not based on what the technology currently does in clinical settings. AI tools assist with documentation, alerts, and information processing, but they cannot replicate presence with a frightened patient, reading unspoken cues, judgment built from experience, or long-term patient relationships. What AI is gradually taking over is part of the documentation burden and some low-level information work.
How can clinicians get started with AI in their workplace?
Start with four practical steps. Learn which AI tools are already in your EHR, including how they were trained and what their error rates look like. Define where you want a human in the loop. Pick one grounded source on healthcare AI, such as AMA publications or academic medical center implementation findings. If you have access, try an ambient documentation tool so your opinion is based on experience.
Why should clinicians be involved in AI decisions at their hospital?
AI adoption is often shaped by administrators, vendors, and technology teams with limited clinician input. When clinicians stay out, implementation is more likely to favor cost and efficiency over patient care and clinician wellbeing. Clinicians who engage, learn the tools, and state clearly what they will and will not accept can influence how AI is used in their specialty.
