Referrals connect patients with specialized expertise, but the process involves more than sending an order. Physicians must communicate the clinical question, provide relevant records, track the consultation, and incorporate recommendations into ongoing care. AI can assist with these tasks by organizing information and preparing documentation for review.
In this article, artificial intelligence (AI) refers to AI-enabled documentation and communication tools, including features within an EHR/EMR as well as separate platforms that connect with a practice’s clinical systems. Available functions depend on the tool and its integration capabilities. Effective use also depends on accurate source information, physician review, and clearly assigned follow-up responsibilities.
AI can turn clinical notes into a structured referral letter that clearly states the reason for referral, the clinical question, relevant findings, treatments already attempted, and the requested response timeframe.
It can also compile pertinent medications, allergies, laboratory results, and imaging reports, while flagging missing attachments, conflicting information, or an unclear consultation request. However, it is crucial that the physician verify AI-generated information for accuracy and correct any inaccurate information prior to sending the referral letter.
AI can draft messages for secure delivery to request a consultation, respond to questions from the receiving practice, or follow up on a missing consultation report.
It can also organize dictated notes from telephone conversations into a structured record identifying who participated, what was discussed, the recommendations made, and who is responsible for follow-up. However, drafting a message does not mean that a referral has been sent, an appointment has been scheduled, or a specialist has accepted responsibility for care. Staff must verify and document these steps. A standalone chatbot cannot transmit records or confirm appointment status unless it is connected to systems that support those functions.
Patient communication is another practical application. AI can prepare plain-language instructions explaining:
The AI-generated instructions should be carefully reviewed for accuracy, and the patient's ability to access the instructions should be verified.
When integrated with referral-management systems, AI can help identify referrals awaiting acknowledgment, scheduling, consultation reports, or physician review. It can generate staff worklists and draft follow-up communications.
AI can also summarize consultation findings and highlight medication changes, recommended testing, and follow-up needs. However, there must be documented evidence in the medical record that:
AI-generated material should be reviewed before signing or sending because AI may introduce unsupported or false statements or omit vital details. Verification should include:
Human oversight remains essential to help ensure the referral accurately reflects the patient's clinical situation and intended plan of care.
Practices should be transparent with patients about their use of AI and protect patient privacy. Patients should receive a plain-language explanation of how AI affects their care, communication, and personal information. Only approved systems with appropriate access controls and security safeguards should be used.
Combined with reliable tracking and assigned follow-up responsibilities, AI can support clearer communication and more complete referral documentation.
AI can prepare a draft referral letter using clinical information, but the physician should review the AI-generated content for accuracy and correct inaccurate information before the referral is sent.
No. Drafting a referral or communication does not mean that it has been sent, an appointment has been scheduled, or the receiving specialist has accepted responsibility for care. These steps should be verified and documented.
When appropriately integrated with a referral-management system, AI can help identify referrals awaiting acknowledgment, scheduling, consultation reports, or physician review. The practice still needs clearly assigned responsibility for follow-up.
Yes. AI-generated material should be reviewed before it is signed or sent because AI may introduce unsupported or false statements or omit important details.
Practices should use only approved systems with appropriate access controls and security safeguards to protect patient information.
Looking for more guidance? Explore our risk management catalog and practice resources. If you're not currently insured with PICA, fill out our online form to receive a free, no-obligation quote.
References
1. Institute for Healthcare Improvement and National Patient Safety Foundation. Closing the Loop: A Guide to Safer Ambulatory Referrals in the EHR Era. Institute for Healthcare Improvement; 2017.
https://www.ihi.org/library/publications/closing-loop-guide-safer-ambulatory-referrals-ehr-era
2. Agency for Healthcare Research and Quality. Make referrals easy: Tool 21. In: Health Literacy Universal Precautions Toolkit. 3rd ed.
https://www.ahrq.gov/health-literacy/improve/precautions/toolkit.html
3. Office of the National Coordinator for Health Information Technology. 2025 SAFER Guide: Clinical Communication. 2025.
https://healthit.gov/wp-content/uploads/2025/06/SAFER-Guide-1.-Clinical-Communication-Final.pdf
4. Van Veen D, Van Uden C, Blankemeier L, et al. Adapted large language models can outperform medical experts in clinical text summarization. Nature Medicine. 2024;30:1134–1142.
https://pubmed.ncbi.nlm.nih.gov/38413730/
5. American Medical Association. Augmented Intelligence Development, Deployment, and Use in Health Care. November 2024.
https://www.ama-assn.org/system/files/ama-ai-principles.pdf
6. U.S. Department of Health and Human Services, Office for Civil Rights. Guidance on HIPAA & Cloud Computing. Accessed October 24, 2026.
https://www.hhs.gov/hipaa/for-professionals/special-topics/health-information-technology/cloud-computing/index.html
Disclaimer: The information contained on the PICA Blog does not establish a standard of care, nor does it constitute legal advice. The information is for general informational purposes only. We encourage all blog visitors to consult with their personal attorneys for legal advice, as specific legal requirements may vary from state to state. Links or references to organizations, websites, or other information is for reference use only and do not constitute the rendering of legal, financial, or other professional advice or recommendations. In the event any of the information presented conflicts with the terms and conditions of any policy of insurance offered by ProAssurance Insurance Company of America, the terms and conditions of the actual policy will apply. All information contained on the blog is subject to change.