Why anesthesia care stays in human hands
Nurse anesthetists work in the few minutes where a plan meets a body that does not read the plan. They induce and maintain anesthesia, manage the airway, and change drug doses when blood pressure or oxygen levels move the wrong way. That loop is physical, fast, and legally owned by a licensed clinician. Software can watch the monitors. It cannot hold the laryngoscope, feel resistance in a jaw, or decide in three seconds that this patient needs a different approach.
The second reason is the conversation before the case. A pre-anesthesia assessment is part chart review, part interview: old reactions to anesthesia, home medications, alcohol use, fears people only admit when asked twice. Models can summarize the chart. They cannot be accountable for what the chart leaves out.
That mix is why the headline figure here sits where it does. Our Still needs a human score for nurse anesthetists is 81 out of 100 (higher is safer), and how that score is built is published in full.
What AI runs, what it assists, what the CRNA keeps
Start with the share of task time software could handle end to end today: 0%. It clusters around paperwork and pattern work. Anesthesia records now populate themselves from device feeds, and drafting case documentation or post-op summaries is a text job that language models are good at.
Assistive tasks are the bigger story: 15% of task time sits in work where the tool sharpens a clinician rather than standing in for one. Decision-support systems flag a hypotension trend before a person sees it. Dosing calculators, ultrasound guidance for regional blocks, and preoperative risk screening all shorten the thinking, not the shift. Overall coverage, the share of task time AI can handle at all, is 13; the method behind the Can AI do it score explains how that is measured.
Everything else stays with the person in the room: 85% of task time. Intubation and airway rescue, moment-to-moment management of an unstable patient, consent conversations, and handoff to the recovery team are all in that group. These are the tasks that carry the license and the liability.
The evidence, and what is missing
Nobody has run a published head-to-head test of an AI system against nurse anesthetists on their actual caseload. That is why our Is it better than a person grade here is D, and why this page gives no parity number. A D grade means not measured, not measured and failed.
What would settle it is narrow and testable: a prospective trial of closed-loop anesthesia delivery against clinician-managed cases, matched for patient risk, reporting hypotension events, awareness, airway complications, and recovery times. Add a separate test of AI-generated pre-anesthesia assessments against clinician assessments on the same patients. Until trials like that are published and replicated, parity stays ungraded. You can read how we weigh evidence strength on our methodology page.
When the picture could change
Most likely after 2042 (8 in 10 of our scenarios). The replacement-year method sets out what that window does and does not mean.
Two things could pull it earlier. First, routine, low-risk cases: short endoscopy and ambulatory procedures are the most protocol-driven work in the specialty, and automated sedation has been attempted there before. Second, staffing economics. Demand is rising, with the Bureau of Labor Statistics projecting 9.7% employment growth for nurse anesthetists between 2025 and 2035 (BLS, 2025), and tight supply pushes hospitals toward anything that stretches a clinician across more rooms.
Two things hold it back. Hardware is the first: the physical share of this job needs dexterous humanoid capability to touch a patient at all, which is the hardest robotics tier we track and nowhere near routine hospital use. Regulation and liability are the second. Anesthesia is state-licensed and insured on the assumption that a named clinician is responsible for the case. Changing that takes rule changes, not model upgrades.
Good to know: with about 51,840 people employed and median pay of $236,590 (BLS, 2025), the pressure on this job shows up first as scope-of-practice and staffing-model fights, not as software doing cases alone.
How to stay needed as a CRNA
Lean into the parts of the job that sit in the human column. Airway management on difficult cases is the clearest one: fiberoptic and emergency airway skills travel everywhere and are nobody’s software feature. Managing the unstable patient is the second, including trauma, obstetric, and high-ASA cases where the protocol runs out. The third is the pre-anesthesia conversation, where you surface the history a chart review misses.
Two skills are worth adding. One is reading machine output critically: knowing when a decision-support alert is noise, and being able to say why in a chart note. The other is teaching and supervision, because models of care that spread one clinician over more rooms turn senior CRNAs into the people who train, oversee, and set standards.
If you are weighing adjacent paths, compare the work before the title. The closest jobs are anesthesiologists, anesthesiologist assistants, and nurse practitioners. You can put any two of them side by side on our job comparison tool, see how the wider diagnosing and treating practitioners family scores, or look at the healthcare sector view. For context on where hands-on clinical work lands overall, see the list of jobs that mostly need a person or search the full job rankings.