‘What they need now is institutional clarity, backed by practical infrastructure’
The trend of workers using artificial intelligence (AI) in their line of work without proper employer guidance has come to medical clinics, posing risks to the health of patients, according to a recent global report.
Nearly 9 in 10 (86%) of clinicians now use artificial intelligence daily or several times a week, with 83% doing so without any employer guidance, reports Heidi, an AI clinical documentation platform.
About the same number (86%) share of clinicians reported growth in their AI usage over the past year, and fewer than 2% said they have never used it in their clinical work. More than half of respondents, 57%, now describe AI as a routine part of how they work, according to the report.
Yass Omar, head of legal and regulatory affairs at Heidi, said institutions have not kept pace. "Clinicians are already using these tools and forming their own governance instincts while regulation catches up," Omar said. "What they need now is institutional clarity, backed by practical infrastructure: compliance packs, clinical safety frameworks, and the regulatory muscle to match the pace clinicians are already moving at."
A chiropractor in Australia, quoted in the report, said regulators bear responsibility for closing the gap: "Clinicians are already using these tools, and most of them are doing it without support. Regulators need to focus on creating safe, standardized frameworks for human-in-the-loop integration, not fear-based restrictions."
Over three-quarter (76%) of workers have used personally sourced AI tools for work, according to a previous report.
How can clinicians use AI?
Nearly 9 in 10 (88%) clinicians identify documentation as their most time-consuming task, a result consistent across every country, specialty and career stage surveyed - an area where AI can help. Heidi found 76% of clinicians say their administrative burden has grown significantly in recent years, and 38% spend more than two hours daily on tasks outside direct patient care.
"The administrative problem is one every practising clinician knows," said Dr. Simon Kos, global chief medical officer at Heidi. "Clinicians' existing tools create more work than they remove."
An occupational medicine specialist in the United Kingdom, quoted anonymously in the report, said the burden has direct workforce consequences: "Every hour we save is either less burnout or more patients seen. These are critical problems in healthcare at the moment."
Accuracy concerns and retention effects
Clinicians surveyed by Heidi cited hallucination and factual accuracy as their top concern with AI, at 68%, ahead of patient privacy (59%), over-reliance (47%) and erosion of clinical judgement (41%). Despite this, 75% said their patients are generally comfortable with AI being used in their care.
The report links AI adoption to workforce retention at a time when the World Health Organization projects a global shortfall of 11 million health workers by 2030. Heidi's survey found 73% of clinicians say AI is helping them sustain a longer, more manageable career.
"When the majority of clinicians tell us AI is helping them stay in their jobs for longer, that's a global workforce crisis being slowed," said Dr. Hannah Allen, a general practitioner and Heidi's European chief medical officer. A family medicine physician in Canada, quoted in the report, said: "It has made medicine more sustainable. I have my evenings back so I feel like I could work for longer in my career."
Employees are doing most of their work-related AI tasks through personal accounts that employers cannot easily monitor, creating fresh governance, privacy and offboarding risks for HR professionals, according to new research, according to a previous report.
Here’s how employers can properly govern the AI use of clinicians, according to several sources:
|
Governance area |
Key point |
Source |
|
Human review |
AI-generated notes must be reviewed by a clinician before or shortly after entering the record; the clinician stays accountable for accuracy AI-generated notes must always be reviewed by a clinician, who remains legally responsible for the accuracy and completeness of the health record. |
Ontario, B.C. and Alberta privacy regulators (via BLG legal analysis, Feb. 2026) |
|
Written policy on AI errors |
Policies must specifically address hallucinations, multilingual transcription errors and automation bias Policies must be implemented and address hallucinations, multilingual errors, and automation bias. |
Information and Privacy Commissioner of Ontario |
|
Impact assessments |
Organizations should complete Privacy Impact Assessments, and Algorithmic Impact Assessments where AI supports clinical decisions, plus vendor contracts covering security and audit rights The IPC expects strong governance, including PIAs/AIAs as appropriate, clear policies, and detailed vendor contracts covering use restrictions, security, breach reporting, and audit rights. |
Information and Privacy Commissioner of Ontario |
|
Vendor data use |
Vendors are effectively barred from using patient conversations to train their models without narrow legal justification All three regulators take a restrictive approach to any secondary use of AI scribe inputs or outputs, particularly for vendor model training. |
Ontario, B.C. and Alberta privacy regulators |
|
Data retention |
Best practice is an "EMR-only" model: delete audio and transcripts once a clinician validates the summary an EMR-only retention model (i.e., clinician-reviewed notes entered into the EMR, with underlying audio and transcripts deleted) has emerged as the prevailing compliance baseline. |
Ontario, B.C. and Alberta privacy regulators |
|
Patient consent |
Consent is required in all three provinces before an AI scribe is used, though the legal threshold varies by jurisdiction Patient consent for the use of AI scribes has quickly become the norm. Consent is required in all three jurisdictions. |
Ontario, B.C. and Alberta privacy regulators |
|
Ethics framework |
Six global principles: protect autonomy; promote well-being and safety; ensure transparency; foster accountability; ensure inclusiveness and equity; promote responsive, sustainable AI Protect autonomy; Promote human well-being, human safety, and the public interest; Ensure transparency, explainability, and intelligibility; Foster responsibility and accountability; Ensure inclusiveness and equity; Promote AI that is responsive. |
World Health Organization, Ethics and Governance of AI for Health (2021) |
