All use cases

Prevent / Care

Doctor copilot — in-tool conversational draft revision

The clinician has a draft reply but wants it softer, or with one more line of context. Instead of rewriting it by hand, they just say so — and the agent regenerates.

Next best thingBehavioral triggerIn-appGLP-1 & PeptidesHealth & Telehealth

What it does

Once a clinical reply is pre-drafted, the clinician rarely accepts it untouched — they want a gentler tone, a caveat added, a paragraph cut. Doing that by hand means re-typing inside the tool, which gives back much of the time the pre-draft saved. Inside the clinician's own tool, the agent lets the doctor refine the draft by plain-language instruction — "soften this," "add a line about hydration," "shorten it" — and regenerates the reply, with the knowledge-base passages and patient context surfaced alongside so the clinician can check each change against the source. The medical decision and the send stay entirely with the clinician; only the editing turns into a conversation.

How it works

  1. 1
    Trigger. the clinician gives a natural-language revision instruction on an existing draft inside the tool ("change the wording," "add context on X," "make it shorter").
  2. 2
    Decision. the agent interprets the instruction against the patient context and the knowledge base, regenerates the reply, keeps a source behind each clinical claim, and surfaces the relevant passages and patient history alongside so the change is checkable; uncertain points are posed as questions rather than asserted.
  3. 3
    Action. the revised draft replaces the previous one in-tool with the citations and context panel updated, ready for another instruction or for approval.
  4. 4
    Follow-through. each revision instruction is logged; on approval, the final reply and its sources write back to the patient's memory, and the instruction patterns feed back as a quality signal for future drafts.
Example
ClinicianThis reads a bit clinical for her — warm it up and add a line that the nausea usually eases with time.
AgentRevised: the tone is gentler and a line now notes that nausea commonly eases over the following weeks, cited to the program's GLP-1 guidance. Patient context and source shown on the right. Want another change, or approve?
ClinicianPerfect — approve.

Configuration

How the agent is wired for this use case.

Triggera natural-language revision instruction from the clinician on an existing draft inside the tool.
Tools & actions
  • Clinical queue · accept the clinician's revision instruction and regenerate the draft in place, in-tool
  • Knowledge base · re-retrieve the relevant passages and keep a source behind each clinical claim after each revision
  • Clinical/EHR system · surface the patient context (current dose, program stage, side-effect log) alongside the draft so each change is checkable
  • Memory store · log each revision instruction and, on approval, write the final reply and its sources back to the patient's memory
Autonomyinstruction interpretation, regeneration, and re-citation run unattended in-tool, but only as a draft the clinician is editing; nothing reaches a patient without explicit clinician sign-off, and anything uncertain is posed as a question rather than asserted. The agent never sets a dose, diagnoses, or sends without the clinician.
Channelsin-app
Escalationthe clinician owns every send and can keep revising by instruction or approve; an uncertain regeneration is posed as a question for the clinician rather than asserted to the patient.

What you need

The inputs this use case runs on. Your channels stay yours; the agent supplies the judgment.

Signals

in-tool revision instructions on an existing draft from the clinician.

Data

the working draft and its citations; the patient's context and memory (current dose, program stage, side-effect log); GLP-1 clinical protocols and knowledge sources with citable passages.

Guardrails

nothing patient-bound without explicit clinician approval; a source kept behind every clinical claim through each revision; a full audit log of revision instructions, drafts, and the approved send; the agent never advises a dose or diagnoses.

Metrics it moves

  • clinician-hours-savedup, because revising by instruction replaces manual re-typing inside the tool, protecting the time the pre-draft saved.
  • time-to-resolutiondown, as the draft converges on a send-ready reply in a few spoken instructions rather than a manual rewrite.
  • csatup, since the clinician can tune tone and add context quickly, so the patient gets a reply that reads as personal and lands sooner.

Related use cases

Prevent / Care

Doctor copilot — pre-drafted clinical reply with citations

the initial pre-draft this revision loop edits

Prevent / Care

Clinician copilot — approve instead of writing

the cross-industry approve-queue these GLP-1 copilots specialise

Prevent / Care

Pre-visit intake summary and SOAP-note copilot

the visit-documentation copilot in the same clinician-facing family

See it on your own customer journey

Bring one drop-off, one churn cliff, or one silent segment. We will show you what a proactive agent with memory and judgment does with it.

Book a demo