Preparing for the AI-Enabled Practice

Clinical Report: Preparing for the AI-Enabled Practice

Overview

At the AAO 2026 Meeting course, consultant Jeffery Daigrepont advises ophthalmology practices to assess AI tools carefully and avoid committing to products before understanding their function and contractual terms. He highlights data use, intellectual property, performance obligations, termination rights, and the possibility that some AI capabilities may become embedded in existing software.

Background

AI is increasingly incorporated into technologies used in ophthalmology practices, while the source describes a market characterized by substantial hype and uncertainty about which products will persist. Daigrepont compares the current market to the early electronic health record market, which he says once included more than 600 vendors and later contracted to a handful. The source also raises the possibility that AI may be delivered as features within existing tools rather than as standalone products. Related articles describe potential uses in practice productivity, documentation, and administrative work.

Data Highlights

Source detail	Reported information
Historical EHR vendor market	More than 600 vendors at its highest point; described as having contracted to a handful
FDA AI-enabled medical devices	More than 1,600 authorized as of September 2026, according to the supplied web research context
Tanzania randomized controlled trial	Among participants with true referable diabetic retinopathy, attendance at a central ophthalmology clinic within 8 weeks was 35.9% with AI-supported screening and point-of-screening results plus in-person counseling, versus 18.2% in the comparison group; absolute difference 17.7 percentage points; OR 2.52; p<0.0001 (n=2,378)

Key Findings

Daigrepont recommends distinguishing AI capabilities from marketing claims and considering whether a proposed tool is a standalone product or a feature likely to be incorporated into existing systems.
He cautions practices against purchasing or financially committing to tools before determining whether similar capabilities are already included in core software.
Before contracting, he identifies key issues to clarify: ownership of intellectual property, how practice data will be used, vendor obligations if the product fails to perform as promised, and termination rights.
He notes that technology vendors and products may not remain in the market over the long term, citing the contraction of the EHR vendor market as a comparison.
The supplied web research context reports that the FDA had authorized more than 1,600 AI-enabled medical devices by September 2026 and issued 2025 guidance on predetermined change control plans for AI-enabled device software functions.
A randomized trial in Tanzania reported higher 8-week ophthalmology-clinic attendance among participants with true referable diabetic retinopathy when AI-supported screening provided point-of-screening results and in-person counseling (35.9% vs 18.2%; OR 2.52; p<0.0001).

Clinical Implications

For practice leaders evaluating AI, the source supports reviewing vendor contracts for data-use provisions, intellectual-property ownership, performance commitments, and termination terms before making long-term commitments. The supplied research context also identifies FDA-authorized AI algorithms as an appropriate strategy for diabetic retinopathy screening under the ADA Standards of Care in Diabetes—2026.

Conclusion

The source presents AI adoption as an area for active exploration paired with careful evaluation of product claims and contractual obligations. It emphasizes that practices should understand how a tool uses data and what commitments they accept before contracting.