AACE 2026: Using AI in clinical practice, best practices

Objective:

To discuss the integration of AI tools in clinical practice, emphasizing best practices for effective use and patient safety.

Key Findings:

• 80% of LLM training data comes from general Internet sources, with only 2% being high-quality curated data.
• AI scribes can reduce documentation time by about 10% depending on clinician interaction and engagement.
• The P.U.L.S.E. framework helps in creating precise prompts for AI tools, emphasizing the need for structured prompts.
• The V.E.T. framework aids in risk mitigation by emphasizing verification, evaluation, and the importance of clinician judgment.

Interpretation:

Clinicians must be cautious when using AI tools, ensuring they understand the limitations of the data and the importance of structured prompts and clinician engagement to achieve accurate outputs.

Limitations:

• AI tools may not provide context-specific medical advice due to the nature of their training data, necessitating clinician verification.
• The effectiveness of AI tools is heavily dependent on clinician engagement, prompt structuring, and the need for clinical judgment.

Conclusion:

AI can enhance clinical practice efficiency, but its integration requires careful consideration of data quality, structured usage, and clinician engagement to ensure patient safety and effective outcomes.