Reply to patients in the clinician’s voice

Replies to patient messages, written from the record

Patient Reply DraftingBy Corti
Meet with an AI expert

Book a call to scope it against your data and tools.

Overview

Draft replies to patients’ written questions from the relevant parts of the record, in the clinician’s own writing style. Guardrails keep answers safe, and the clinician edits and sends.

How it works

Takes the patient’s question and the relevant parts of the record and drafts a reply in the clinician’s style, learned from examples the clinician has approved. Guardrails apply: no new diagnosis, no medication changes, red flags are escalated and not answered. The source passage for each fact is shown. The clinician edits and sends.

What Corti handles

  • Finding the relevant record data for a question
  • Drafting in a style learned from approved examples
  • Guardrails and red-flag escalation

What you build

  • The inbox and draft view in your product
  • Access to the record through tools the agent can call
  • Storing each clinician’s approved examples

Make it yours

  • Write the instructions the agent follows
  • Start from a ready-made template, or write your own
  • Pick the model the agent runs on
  • Decide the structure and writing style of each document
  • Connect the tools and data sources it can use
  • Produce several documents from one visit, such as a note and a patient summary
  • Decide who can see and use the agent
  • …and much more

Frequently asked questions

What is patient reply drafting?

A draft answer to each written patient question, such as an e-consultation or portal message, built from the relevant parts of the record. It is written in the clinician’s own style, and the clinician edits and sends it.

How does it learn to sound like the clinician?

From replies the clinician has already approved. Those examples set the style, and the draft shows the source passage for every fact it states. Personal Note Style does the same kind of learning for visit notes, from past notes and edits.

What stops it from saying something unsafe?

Guardrails. It never proposes a new diagnosis or a medication change, and red flags are escalated to the clinician instead of answered.

Who is this for?

Clinicians whose portal inbox fills faster than they can answer it, and who find themselves writing similar replies again and again.

What does Corti do, and what do we build?

Corti finds the record data for a question, drafts in the learned style and applies the guardrails and escalation. You build the inbox and draft view, the tools through which the agent reads the record, and storage for each clinician’s approved examples.

How do we start?

Start with the inbox view and the record tools, then collect a first set of approved replies from your clinicians. To plan it with a Corti expert, press the “Meet with an AI expert” button.

Add patient reply drafting to your product

Book a call to see Patient Reply Drafting running on your own data and workflows.