Artificial Intelligence Scribes in Healthcare
Laura M. Cascella, MA, CPHRM
Artificial intelligence (AI) is the frontier in healthcare technology, and its potential seems almost boundless. One highly touted promise of AI is that healthcare organizations and providers can use it to reduce administrative workload, particularly as it relates to documentation.
Accurate and complete documentation is the basis of high-quality patient care, but it has long been recognized as a significant strain on healthcare providers. The widespread implementation of electronic health records (EHRs) did not ease the burden; rather, it created additional issues for providers to manage, such as information overload, copy-and-paste errors, alert fatigue, and less-personalized patient experiences.
A JAMA study of primary care physicians found that they spend about 36 minutes documenting in the EHR per 30-minute patient visit.1 Further, a survey from the American Medical Informatics Association revealed that about 82 percent of physicians feel that the time they spend on documentation is inappropriate, and about 83 percent said they finish work later than desired or need to work at home to complete documentation. Nurse practitioners’, physician assistants’, and advanced practice nurses’ responses to the same questions were only slightly better (73 percent and 81 percent, respectively).2
Documentation tasks have clearly created an undue burden on healthcare providers and contributed to the ongoing and pervasive issue of burnout in healthcare. As a result, the fruition of AI scribes is a welcome breakthrough. This technology employs ambient listening, speech recognition, machine learning, and natural language processing to record patient encounters in real time, convert them into text, and interpret and summarize the information to create clinical notes.3 The American Medical Association notes that AI scribes can save providers about an hour of time each day on documentation tasks.4
Although ambient scribing represents a significant advancement in healthcare and a palpable remedy to onerous documentation requirements, it does not absolve healthcare providers of their responsibility for clinical notes. They should still carefully review scribes’ output to ensure accuracy and appropriate context. AI is not infallible, and research studies and anecdotal reports have pointed to AI errors or “hallucinations” in synthesizing health-related information.
Further, AI scribes are not universally efficacious in every clinical setting or for every specialty, and they might face hurdles in relation to interpreting accents, nuances in speech, situational context, medical jargon, and so on. Healthcare organizations may need to work with their vendors to tailor and fine-tune products to produce optimal results.5
Numerous other factors also should be considered in the quest to adopt ambient scribing. For example:
- Be aware of potential bias in the technology. Bias can arise from many sources, including the training data, the people who interact with the data, and general inequities that exist in our society and healthcare systems. Carefully screen AI scribe vendors and products to understand how the technology was trained, its capabilities, and its limitations.
- As part of procurement, assess each product’s privacy and security features to ensure it meets federal and state requirements for healthcare data privacy. Determine how the technology stores data (if applicable), processes data, and transmits data. Work with technical and cybersecurity staff to evaluate products and implement safeguards at all points of risk exposure.
- Carefully review vendor contracts to determine whether they include indemnification clauses that shift any liability risks of using the technology to the healthcare organization. Work with legal counsel to ensure a thorough understanding of contracts.
- Engage in clear and transparent communication with healthcare providers about implementing ambient scribing technology. Communication should occur from the initial planning stages throughout the implementation process and beyond.
- Anticipate staff reluctance or resistance to using ambient scribing, which may occur because of lack of familiarity, distrust of AI, technical difficulties, or other factors. Develop approaches to help overcome these issues, such as having influential staff members serve as champions or creating a group of super users who can assist others.
- Determine how the new technology will integrate with the organization’s EHR system and affect workflow processes for clinical and nonclinical staff. Seek stakeholder input on aligning the capabilities of the technology with the realities of daily practice.
- Ensure organizational documentation policies are updated to include information about AI scribes, how they are used, and accountabilities for providers (e.g., reviewing all scribe-generated notes for accuracy within a specific timeframe).
- Make sure providers receive appropriate training on AI scribe technology. Training should address the functionality of the software as well as organizational documentation policies, requirements for patient consent, and so on. Consider using a pilot program to implement the technology on a small scale, identify glitches or issues, and develop best practices.
- Develop guidance related to seeking patient consent for using ambient scribing during clinical encounters. The Medical Group Management Association notes that consent requirements “can vary depending on the specific practice and jurisdiction. However, it is important to note that patient consent is generally required for the use of any technology that involves the collection, storage, or sharing of patient health information.”6
- Develop a strategy for communicating with patients/families about ambient scribing, including explaining how it works, discussing its benefits, and addressing any concerns that they may have in relation to the technology. Consider supplementing verbal communication with posters, factsheets, and information on the organization’s website/portal.
- Monitor the technology to establish benchmarks, measure performance over time, and collect data for quality improvement. Additionally, stay current on evolving legal implications related to the technology, and share pertinent information with stakeholders.7
Healthcare often faces abundant challenges but a lack of viable solutions. AI is poised to help level the playing field with applications that address a range of areas, from diagnostics to logistics. AI scribes stand ready to tackle the long-standing burden of documentation in clinical care. The technology offers many potential benefits, such as saving time, improving efficiency, reducing burnout, and improving provider–patient encounters.
Although ambient scribing may seem to have many advantages and few drawbacks, organizations should take a calculated and systematic approach to implementing it. Carefully assessing vendors and products, developing detailed policies, educating providers and patients, and clearly communicating with all stakeholders will help facilitate successful change management and reduce the risk unintended consequences.
For a wide range of information and topics related to AI, see MedPro’s Risk Resources: Artificial Intelligence.
Endnotes
1 Berg, S. (2024, January 23). Primary care visits run a half hour. Time on the EHR? 36 minutes. American Medical Association. Retrieved from www.ama-assn.org/practice-management/digital/primary-care-visits-run-half-hour-time-ehr-36-minutes; Rotenstein, L. S., Holmgren, A. J., Horn, D. M., Lipsitz, S., Phillips, R., Gitomer, R., & Bates, D. W. (2023). System-level factors and time spent on electronic health records by primary care physicians. JAMA Network Open, 6(11), e2344713. doi: https://doi.org/10.1001/jamanetworkopen.2023.44713
2 Sloss, E. A., Owoyemi, A., Mishra, A. K., Sidwell, K., Abdul, S., Dawood, G., . . . Rossetti, S. (2024). Development of the TrendBurden survey: Capturing the pulse of the nation related to excessive documentation burden for health professionals. American Medical Informatics Association. Retrieved from https://brand.amia.org/m/2dcad47c0da0c01d/original/PowerPoint-Presentation.pdf
3 Kaushik, G., & Reifsnyder, C. (2024, November 1). Transforming medical documentation with ambient scribes. Veradigm. Retrieved from https://veradigm.com/veradigm-news/ambient-ai-medical-scribe/
4 Robeznieks, A. (2024, March 18). AI scribe saves doctors an hour at the keyboard every day. American Medical Association. Retrieved from www.ama-assn.org/practice-management/digital/ai-scribe-saves-doctors-hour-keyboard-every-day
5 Coiera, E., & Liu, S. (2022). Evidence synthesis, digital scribes, and translational challenges for artificial intelligence in healthcare. Cell Reports Medicine, 3(12), 100860. doi: https://doi.org/10.1016/j.xcrm.2022.100860; American Medical Association. (2025, January 6). AI scribes for clinicians: How ambient listening in medicine works and future AI use cases. Retrieved from www.ama-assn.org/practice-management/digital/ai-scribes-clinicians-how-ambient-listening-medicine-works-and-future; Athreon. (2024, July 15). AI scribes explained: What they can and cannot do. Retrieved from www.athreon.com/ai-scribes-explained-what-they-can-and-cannot-do/
6 Good, C. (2024, April 2). Sample patient consent form for using artificial intelligence for dictation, transcription. Medical Group Management Association. Retrieved from www.mgma.com/member-tools/sample-patient-consent-form-for-using-ai
7 Robeznieks, AI scribe saves doctors an hour at the keyboard every day; Coherent Solutions. (n.d.). Benefits and pitfalls of AI medical scribe and transcription solutions. Retrieved from www.coherentsolutions.com/insights/benefits-and-pitfalls-of-ai-medical-scribe-and-transcription-solutions; Kaushik, et al., Transforming medical documentation with ambient scribes; Tierney, A. A., Gayre, G., Hoberman, B., Mattern, B., Ballesca, M., Kipnis, P., . . . Lee, K. (2024). Ambient artificial intelligence scribes to alleviate the burden of clinical documentation. NEJM Catalyst, 5(3). doi: 10.1056/CAT.23.0404
