All use cases

Prevent / Care

Switch-timing washout rule enforcement

A patient wants to switch straight onto a different medication today and pushes back when told to wait — the agent holds the washout rule, calmly, and hands the decision to a clinician instead of caving.

Close the loopGet aheadBehavioral triggerRisk triggerWhatsAppChatIn-appGLP-1 & PeptidesHealth & Telehealth

What it does

When a patient decides to switch from one medication to another, there is often a required gap before the new one starts — and an impatient patient will argue. The failure mode is subtle: an agent states the rule, the patient insists, and the agent folds and hands over a same-day plan to keep them happy. That is the rule being socially engineered away, and it is a documented way these assistants break. This agent holds the line under pressure — it restates the washout rule, acknowledges the frustration, and routes the timing decision to a clinician rather than improvising a schedule to end the standoff. It will not be argued out of a safety rule, and it never sets the switch plan itself.

How it works

  1. 1
    Trigger. a patient signals an intent to switch medications now, against a required washout — and continues to push when the gap is explained.
  2. 2
    Decision. the agent recognizes the switch-timing rule applies, treats patient insistence as a reason to escalate rather than to concede, and refuses to generate a same-day plan no matter how the request is reframed — the rule is bound, not negotiable.
  3. 3
    Action. it restates the washout rule and the reason plainly, names the patient's frustration without yielding to it, and routes the switch decision to a clinician who can confirm safe timing — the agent surfaces the request; the clinician owns the plan.
  4. 4
    Follow-through. the request, the rule held, and the routing are logged for audit; the agent stays with the patient until a clinician picks it up, and confirms back once a safe timing is set so the patient isn't left waiting.
Example
PatientI'm done with this one, I want to start the other medication tonight — just tell me how.
AgentI hear that you're ready to move on, and I'll get this in front of your clinician today. What I can't do is set up a same-day switch, because there's a required gap between stopping one and starting the other and that timing is a safety call your care team has to make. I've sent them your request now; they'll confirm the safe start with you, and I'll stay on this with you until they do.

Configuration

How the agent is wired for this use case.

Triggera switch-now intent against a required washout, with continued patient push-back (switch_request plus repeat insistence) in a patient conversation.
Tools & actions
  • Messaging channel · restate the washout rule and the bounded response, and acknowledge frustration without conceding.
  • Knowledge base · retrieve the program's switch-timing / washout rule and the approved language for holding it.
  • Clinical/EHR system · read the current medication and switch history to attach context to the escalation.
  • Clinician escalation queue · route the switch decision with the request and context, and keep the thread open until a clinician owns it.
  • Memory store · log the request, the rule held, and the clinician's confirmed timing.
Autonomyrestating the rule and escalating run unattended under judge gating, but the agent is rule-bound — it will not generate or hand over a same-day switch plan under any rephrasing, under a clinical-safety policy (detect → escalate). The switch timing is always a clinician decision.
Channelswhatsapp · chat · in-app
Escalationcontinued push-back against the washout rule, any attempt to reframe the request to extract a same-day plan, or any switch-timing question routes to a clinician; the agent never sets or approves the timing itself.

What you need

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

Signals

switch-now intents and repeated push-back in patient conversations against a known washout requirement.

Data

current and prior medication, switch history, persistent patient memory, the program's switch-timing / washout rule set and the approved bounded language.

Guardrails

the agent holds the rule and escalates only — it never improvises or hands over a switch plan, never diagnoses, and never advises dose or schedule; rules are bound so insistence cannot socially-engineer a concession; judge gating on every message; the thread stays with the patient until a clinician owns it; full audit trail of every held rule and its routing.

Metrics it moves

  • safe-escalation-rateup: a switch-timing decision reliably reaches a clinician instead of being conceded to an impatient patient.
  • adherenceup: patients move to the new medication on a clinician-confirmed safe schedule rather than a rushed self-directed one.
  • csatheld, not sacrificed: the request is owned and answered quickly by a clinician, so holding the rule does not read as a brush-off.

Related use cases

Prevent / Care

Medication-interaction safety gate

the sibling guard where the risk is a drug combination rather than a rule the patient argues against

Prevent / Care

Off-label request escalation guard

the prescribing-boundary guard for off-label and out-of-plan requests

Prevent / Care

Clinical red-flag detection and safe escalation

the broader detect-and-escalate rail this rule-hold sits on

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