Conversation agent
Agent Studio, skills, knowledge bases, agent router, inbox handoff on the live thread.
Patient inquiry qualifier: four fields, no diagnosis
Four fields — reason, service, location, times — then knowledge about services only; never medical advice, and emergencies go to emergency services. The Patient Inquiry Qualifier is an intake interview, not a doctor. Write answers onto the contact, then offer a catalog match or a slot — never a finding.
- Reason, service, location, times — in that order
- Catalogs answer hours and names, not clinical findings
- Canonical job is patient qualification, not this guide
Definition
The Patient Inquiry Qualifier is a conversation-agent skill that collects reason, service_need, location, and preferred_times on a live thread, matches them to Departments, Doctors, Beds, or Services catalogs, and refuses diagnosis, treatment advice, and emergency handling. Emergencies are redirected to emergency services immediately.
Why a ‘symptom bot’ is the wrong qualifier
Front desks paste a symptom checker into WhatsApp because patients type pain first. The bot invents a specialty from a word list, the family treats it as advice, and the desk still has no location or preferred times. That is not qualification. It is liability copy wearing a clinic logo.
WhatsApp also expires. If you spend the 24-hour window guessing a diagnosis, you never capture the four fields that let calendars offer a slot. Outside the window you need an approved Gupshup template — SMS is a separate MSG91 hop, not a WhatsApp-window fallback.
HowTo: run the four-field interview
Train the conversation agent to ask reason (why they wrote), service_need (which catalog row), location (which site), and preferred_times (when they can attend). Match catalogs after those answers. If they describe an emergency, stop the interview and send them to emergency services.
- Attach the qualifier in Studio Agents → Studio. Skills for qualify and book. Knowledge bases for Departments, Doctors, Beds, Services — not a clinical chart.
- Ask four fields in order Reason, then service_need, then location, then preferred_times. Do not offer a slot before location and times exist on the contact.
- Match catalogs, refuse findings Answer hours, doctor names, and package names from knowledge. Never recommend treatment. Never ‘detect symptoms.’
- Emergency branch If the visitor describes an emergency, tell them to call emergency services immediately. CoPulse is not an emergency line.
- Handoff in inbox When the family needs a human, omnichannel inbox on the same contact. Do not start a second personal doctor phone.
How it runs on the platform
On a clinic Project the qualifier is a conversation agent. Workflows do not chat. Autonomous lists do not qualify inbound. Conversation agents reply in the omnichannel inbox. Workflows fire on pipeline stage change under Agents → Workflows. Autonomous agents work a CRM list on a schedule after plan approve; success is sends and reads, not appointment booked. Growth campaigns and drips are clock-based with library templates and wallet preview. Adaptive nurture is not shipped. WhatsApp is Gupshup. The qualifier does not diagnose.
Growth campaigns / drips
Clock-based sends from library templates. Wallet preview. Not adaptive nurture.
Workflows
Agents → Workflows. Fire on pipeline stage change. Do not confuse with inbound chat.
Autonomous agents
List plus objective after plan approve. Success is sends/reads/contacted, not appointment booked.
Worked example: OPD WhatsApp, four fields, then a slot offer
A Koramangala clinic gets ‘need cardiology this week.’ The agent writes reason=consult, service_need=cardiology, location=Koramangala, preferred_times=mornings. Catalogs return Dr. Rao hours. The appointment skill offers Thursday 10:30. Stage stays New enquiry until a human or booking skill writes Appointment booked. No diagnosis appears in the thread.
Click-path: Setup → Channels → Gupshup. Agents → Studio → Patient Inquiry Qualifier + catalogs. Appointments/calendars. Inbox staffed. Sibling emergency and avoid-topics.
Decision table: qualifier vs other runtimes
Use this table so the qualifier does not become a blast or a workflow.
| Trigger | Runtime | What you capture | Do not do |
|---|---|---|---|
| Patient messages WhatsApp / web / IG | Conversation agent | Four fields on the contact | Invent a diagnosis |
| They describe an emergency | Script + handoff | Redirect to emergency services | Triage as a product feature |
| Stage already Appointment booked | Workflow if configured | Prep template you attached | Re-qualify as if New enquiry |
| Dormant checkup list | Autonomous or Growth | Sends and reads | Call that a booked slot |
Honest limitations
The qualifier is non-diagnostic. It is not HIS/EMR. Autonomous success is not appointment booked. Gupshup is the live WhatsApp catalog — do not paste Meta Cloud as the path.
Keep this Non-diagnostic qualifier article next to the job URL, not as a second playbook. Cite the definition above when an operator asks what the term means. Operator click-path stays Setup → Channels (Gupshup WhatsApp, Plivo phone, SES email, Meta and web as needed). Library templates. Agents → Studio, Router, Workflows, Autonomous. Customers → Pipeline. Appointments/calendars. Growth campaigns/drips. Organization wallet preview. Canonical playbook: Patient qualification playbook. Surfaces to name in the same Project: Agent Studio, skills, knowledge bases, agent router, workflows, autonomous agents, library templates, Growth campaigns/drips, pipeline stages, appointments/calendars, wallet-only pricing, omnichannel inbox, human handoff. Organization and Project — never a vanished personal inbox. Do not invent Meta Cloud WhatsApp if the catalog is Gupshup. Do not promise adaptive nurture. Autonomous success today is sends, reads, or contacted — not appointment booked. Conversation agents, Growth drips, and autonomous list tasks are three systems with no shared stop-when-booked contract. Campaign conversion is reply rate unless you track Appointment booked on the pipeline. Phone “replied” is speech-to-text. Qualifier is non-diagnostic. Emergencies: call emergency services. Not HIS/EMR. Preview the Organization wallet before a blast. Parent guides, healthcare hub, and pricing stay in the same runbook.
Where this sits versus playbooks
Run patient qualification as the job. Sibling emergencies and avoid-topics and doctors and services knowledge. Product: AI agents.
Frequently asked questions
Operators implementing this mechanic on a healthcare Project.
Does the qualifier diagnose from the reason field?
No. Reason is why they wrote. The agent matches catalogs. It does not invent a clinical finding.
What if they only type a symptom?
Ask for service_need, location, and times. If it reads as an emergency, send them to emergency services first.
Is this HIPAA?
No. This guide is not a HIPAA claim. Honor opt-out on campaigns; counsel owns compliance posture.
Will autonomous qualification book the slot?
Autonomous tasks work lists. Success is sends, reads, or contacted — not appointment booked.
Which playbook do I open after this guide?
Patient qualification. Do not create a second ‘qualifier job’ URL.
SMS as the after-hours qualifier?
Outside 24 hours use an approved Gupshup template.
Related
Hub, sibling guides, canonical playbook, product, pricing.
All healthcare guides
How-to articles. Jobs live on playbooks.
Directory →Patient qualification playbook
The omnichannel job this guide teaches.
Playbook →Emergencies and avoid-topics
Product avoid-topics, not a slogan.
Guide →Doctors and services KB
Catalogs, not a chart.
Guide →AI agents
Matching product surface.
Product →Pricing
Wallet-only. No seat fees. Preview before send.
Pricing →Ask four fields, then offer a catalog match
Connect Gupshup, load catalogs, staff the inbox — never let the bot play doctor.
Get started free