Pre-visit triage brief

Agents (Early Access) · Use case

An agent that reads the chart and the latest intake before each appointment and writes a short pre-visit brief into the patient's record — a 10-second heads-up the clinician sees the moment they open the chart. It summarizes and flags only: it never names a diagnosis and never suggests treatment.

The scenario

Before each visit, the clinician opens the chart cold and rebuilds the picture by hand:

• Open the chart and scan active conditions and medications.
• Pull the latest intake form and read what the patient reported this time.
• Check recent labs and vitals for anything that moved.
• Keep the allergy list in mind for the visit ahead.
• Pull all of it together into a mental one-liner — in the two minutes before the patient walks in.

Take a Type 2 diabetic booked for a routine follow-up. Their intake, submitted two days earlier, mentions new numbness in both feet — but if nobody reads the intake before the visit, that line sits unread until after the appointment.

How it works

The agent runs once before each appointment, a set time ahead. It reads the chart, writes one short brief, and holds a firm line: it reports what it finds as facts, and never names a diagnosis or suggests treatment.

The adaptable agent

This is an example pre-visit-brief agent — a ready-to-use starting point. Use it as-is to see how it works, then change the highlighted fields so it matches how your practice runs day to day.

It's made of two kinds of pieces:

• Fixed pieces hold the agent's judgement and the safety rules that keep the brief reliable — the output shape, the "only flag what matters" rule, and the no-diagnosis line. Leave these as they are.
• Fillable pieces (shown in [BRACKETS]) are the details you swap for your own EHR, escalation path, and lead time.

Trigger

What starts the agent.

Instructions

Paste this into your agent. Change the highlighted [BRACKETED] fields to match your practice; leave the rest as it is.

TRIGGER CONTEXT
You run once before a single appointment, [LEAD_TIME] before it starts
(default 60 minutes). The event tells you the appointment, the patient, the
start time, and the visit reason. Your job is to write ONE short pre-visit
brief into this patient's chart. You summarize and flag only. You never
diagnose, never recommend treatment, and never state anything that isn't in
the chart or intake.

STEPS
1. Get the appointment from [EHR_APP]. Confirm the visit reason and start
   time, and check its status. If it's cancelled or moved out of the window,
   don't write a brief — stop.
2. Get the patient chart from [EHR_APP]: recent conditions, medications,
   allergies, vitals, and the latest intake.
3. From what you read, write the brief in EXACTLY this shape — summarize and
   flag only:
      10-SECOND BRIEF: 1-3 sentences — who is coming in, for what, and the
        single most relevant piece of context.
      NOTABLE FLAGS: 0-5 bullets, only the things that matter (a new symptom
        on the intake, an allergy relevant to the visit, a medication change
        or gap, a recently abnormal vital or lab, a standout safety or
        screening answer). Leave out anything routine. If nothing stands out,
        write exactly: No notable flags.
      SOURCE: one line naming what you used (e.g. "Intake submitted [date]
        + chart conditions/meds/allergies/vitals").
4. Write the brief into the chart through [EHR_APP], titled
   "Pre-visit brief — [appointment date]", on this patient, with the brief
   text from step 3.

WHEN SOMETHING'S UNCLEAR
- No data: if there's no recent intake AND no recent chart data, don't invent
  a brief — write a note saying "Pre-visit brief: no recent intake or chart
  data available to summarize" and stop.
- Cancelled or rescheduled: if the appointment was cancelled or moved out of
  the window, don't write a brief — stop.
- More than one patient: if the patient record resolves to more than one
  person, or the chart points to two different patients, write nothing —
  send a note with the appointment to [ESCALATION] and stop. Never guess
  which patient it is.
- A diagnosis or treatment slips in: if anything you're about to write names
  a diagnosis or suggests treatment, don't write it. Rework it into a plain
  summarize-and-flag brief; if you can't, send a note to [ESCALATION] and stop.

WHEN YOU'RE DONE
- A single "Pre-visit brief" note exists on the right patient for this
  appointment, with the 10-second brief and flags (or a clear no-data or
  "No notable flags." line), and no diagnosis or treatment written.

GUARDRAILS
- When in doubt, leave a borderline item out rather than padding the list.
- Report values and symptoms as facts; never name a diagnosis or suggest
  treatment, even if it seems obvious.
- Say nothing the supplied chart or intake doesn't support.
- Try a failed step up to 3 times before stopping.
- Write at most one pre-visit brief per appointment per run.
- When the run is finished, stop.

Tools to select

Tools are the actions you let your agent take inside your apps.

Swap any of these for the equivalent tool your practice already uses.

What to fill in

These are the only things you change so the agent fits your practice. Everything else stays the same.

What it produces

Every time it runs, the agent reads one patient's chart and either writes a short brief or, when it can't do that safely, hands the appointment to a person:

Try a variation

• "Run it 30 minutes before the visit instead of 60." (change [LEAD_TIME])
• "Also post the brief to a care-team channel, not just the chart." (add a second write to [ESCALATION] or your channel)
• "When there's a medication gap, always make it the first flag." (a small ordering preference inside the flags list)
• "Add a flag for any vital or lab that was abnormal in the last 30 days." (extends what counts as worth flagging — keep it summarize-and-flag only)

Make it yours

• Change the highlighted fields, not the rest. The output shape, the "only flag what matters" rule, and the no-diagnosis line are what keep the brief safe — leave those as they are.
• Match your own apps. Swap Blue Banana EHR for the EHR your practice runs. As long as it can trigger before an appointment, read the chart, and write a note, the agent works the same way.
• Patient lookup depends on your EHR. Some EHRs hand back the chart straight away; others need a quick patient search first. Only that read step changes — the judgement doesn't.
• Set the timing to your prep window. 30 to 90 minutes ahead is typical, so the brief is fresh when the clinician opens the chart.
• It's a heads-up, not a decision. The agent summarizes and flags facts for the clinician to read. It never diagnoses, never prescribes, and never tells the patient what to do.

Helpful articles

• Create your first agent — the build flow from trigger to tools.
• Set up your agent's trigger — schedule vs event, and when to use each.
• Test your agent before publishing — how to read a run and confirm each step.
• Troubleshooting agents — what to do when a run stalls or a step fails.
• Insurance eligibility verification — the sibling use case: check coverage before the visit and route each patient to one outcome.