Use Case · Builders
Build Your Own AI Agent — In Under an Hour
Off-the-shelf bots break the moment your practice does anything specific. This is a real walkthrough of building a custom AI insurance verification agent for a healthcare practice — start to finish, in about 20 minutes, no code.
Why "off the shelf" bots fail for your specific need
Every healthcare practice has at least one workflow that looks weird from the outside. A chiropractic office that only takes insurance for the first visit. A med spa with a strict 48-hour cancellation rule. A functional medicine clinic with a 12-page intake form that has to be done in a specific order. A multi-location group where one location takes Medicare and the others do not.
Generic AI receptionists handle the easy 80% — appointment booking, hours, directions. They fall apart on the 20% that actually defines how your practice runs. Custom-coding around them means hiring a developer, signing up for three SaaS tools, and waiting weeks for a workflow that should take an afternoon.
Agent Studio exists to close that gap. This page walks through one concrete example, then gives you eight more starting points.
The agent we're building: AI Insurance Verification
A new patient submits an inquiry. We want an AI agent that, without a human touching it:
- Inputs: patient name, date of birth, and a photo of the front and back of their insurance card.
- Processing: OCR the card, look up the payer, extract the plan, check whether you accept it, estimate copay and deductible exposure.
- Outputs: push verification status to the chart, send a plain-English eligibility summary to the patient, and flag your front desk if anything looks unusual (out-of-network, expired card, unfamiliar plan).
This is the kind of agent you cannot buy. It is also exactly the kind of agent Agent Studio is built for.
The 5-step build
1. Trigger
Drop in a "Form Submitted" trigger node. Point it at your new-patient intake form. The agent fires the moment a card photo arrives.
2. Knowledge Base
Connect a Knowledge Base loaded with the payers and plans you accept, your fee schedule, and your front-desk escalation rules. This is the agent's brain.
3. AI Reasoning Node
Drop in an AI node and write a plain-English prompt: "Read the card photo, extract payer + plan + member ID, check against the in-network list, estimate copay, decide accept/flag."
4. Action Nodes
Three actions: update the contact record with verification status, send an SMS to the patient with their eligibility summary, and create a task for the front desk if the AI flags an exception.
5. Test + Deploy
Use the built-in tester to feed it three sample cards (clean, expired, out-of-network). Inspect the agent's decisions. Publish when it behaves correctly.
Total time: ~20 minutes
Most of the time goes into curating the knowledge base of accepted payers. The agent itself is six nodes. The first version handles roughly 80% of cases unattended.
Eight other agents worth building this week
Once you have built one, the second one takes ten minutes. Here are the agents practices ship most often:
Recall Agent
Detects patients overdue 60/90/180 days, drafts a personalized message referencing their last visit, sends and tracks reply.
Intake Triage
Reads a new patient inquiry, classifies urgency and service interest, routes to the right calendar or care team.
Review Responder
Pulls new Google reviews, drafts a compliant response per your voice guide, posts after approval (or auto-posts for 4-5 stars).
Lead Qualifier
Asks 3-5 conversational questions to inbound leads, scores them, and books only qualified prospects onto the calendar.
New-Patient Welcome
Multi-step welcome series: confirmation, parking instructions, what-to-bring, intake form reminder, day-of arrival nudge.
No-Show Recovery
Detects a missed appointment, reaches out within 15 minutes with empathy and a rebook link, escalates to staff if no reply in 24h.
Cancellation Rebooker
Triggered on cancellation. Offers the next two available slots, books directly, updates the calendar.
Post-Treatment Check-In
Day-3 and day-14 follow-up after a treatment plan visit. Captures outcomes, surfaces issues to the provider, asks for a review when appropriate.
Pricing context
Every agent in this guide runs on PatientCopilot at no extra cost. Agent Studio, the Knowledge Bases, the AI reasoning, and the action nodes are all included on paid plans. There is no separate LLM bill, no per-message charge for the AI, no "AI add-on" SKU. The same fee covers an account running one agent or twenty. See pricing for plan details.
What to build with this
If you already know what you want, open Agent Studio and start. If you want to see the building blocks first, read about Knowledge Bases (what the agent knows) and Ask AI (the in-platform assistant that helps you build). If you want external AI tools like Claude or ChatGPT to query the agents you build, read about the MCP Server.
FAQ
How long does building an agent actually take?
A focused agent: 20-60 minutes. Multi-agent systems with routing: 2-4 hours. Most practices have something useful running the same afternoon they start.
Do I need a developer?
No. Drag-and-drop nodes, plain-English prompts for the AI reasoning step, and a built-in tester. The only people we have seen need outside help are practices customizing the front-end widget styling — and that is a five-minute CSS tweak, not an integration project.
Can I start from a template?
Yes — the Template Library inside Agent Studio has starting points for most of the eight agents above. Clone, edit, ship.
Are AI agents priced separately?
No. Included on all paid plans. No LLM token billing.
Can agency users share agents across clients?
Yes. Build once, snapshot it, deploy across every client sub-account. See the agency platform page.
Ready to Get Started?
Contact us today and take the first step. Free consultations available.