AI Scribing and Telephone Management

“The days of ‘press one for scheduling, press two for billing’ are over.” Or soon may be, according to an AI avatar created by Robert T Chang MD.

Using technology already available, phone menus could be replaced by generative AI call agents—human-sounding programs that respond appropriately and in context to natural language questions, giving better service to callers while reducing staffing needs.

“They do what is needed, like making an appointment, scheduling a refill, whatever,” Dr Chang said. Agents can be programmed with different personas, languages, and behaviours, such as casual, formal, or direct. General guidelines, such as coding the agent as a receptionist for a physician’s office, can be established. Also included are detailed workflows, such as which questions to ask in response to specific scenarios and prompts and when to alert a nurse or other provider.

“You can program your agent to do whatever you want, whether [describing] premium lenses when you are scheduling cataract surgery or immediately directing you to a nurse if you have new flashes or floaters—all this is going to be worked into the AI,” Dr Chang said.

Speaking the patient’s language

And taking over the phones is just one way generative AI agents can improve practice efficiency, Dr Chang said. “AI is kind of a superpower. There are some things you can do with the click of a button that could take hours to do without AI.”

Virtual scribing is another example. Software already exists capable of handling many tasks that currently draw physicians’ time and attention away from patients, simply by listening to the doctor–patient conversation. These include creating summary notes, suggesting diagnosis and procedure codes, pre-charting test orders and prescription refills, creating after-visit summaries, and pre-writing answers to patient messages.

In Dr Chang’s clinic, the virtual scribe looks like a cell phone. It connects to the practice’s medical records system via an AI program. It prepares the clinical summary and notifies the physician for review. Once approved, the scribe sends the records and orders on their way. The whole process takes about 30 seconds, he said.

The system can also translate questions and answers into several languages in real time, Dr Chang noted. Templates can be customised to write notes in whatever format individual physicians prefer. The systems can handle a great deal of detail, which subspecialists such as neuro-ophthalmologists love, he said.

Some systems also automatically generate patient instruction notes written at a young adolescent reading level, which a clinician can review and edit if needed. “They’re getting better all the time,” Dr Chang added.

“You can see where we’re going here, with the variability of language, patient instructions, foreign languages—how can you make that visit better, faster, with fewer call-backs, better instructions? It’s AI,” he said. “AI is going to be doing a lot of those tasks for us.”

Dr Chang spoke in an AI Symposium at the 2025 ASCRS annual meeting in Los Angeles.

Robert T Chang MD is associate professor of ophthalmology at Stanford University, Stanford, California, US. viroptic@gmail.com