How the AI Works

DPC Pro’s AI assistant handles routine tasks (drafting message replies, summarizing your day, and recovering failed payments) so your team can focus on patient care.

The AI assistant in DPC Pro is designed for the specific needs of DPC practices. It operates within your practice data to automate repetitive work, surface relevant information, and reduce administrative overhead. The AI does not make clinical decisions and does not communicate with patients without clinician approval.

This page provides an overview of how the AI assistant works, what data it can access, and the principles that guide its behavior. Understanding the AI’s capabilities and boundaries helps your team use it effectively and trust its output.

Every AI action is logged and reviewable. Your practice retains full control over what the AI is allowed to do.

What the AI Can Do

The AI assistant performs four main types of work:

Answer practice questions: Ask the AI about your schedule, patient panel, billing status, or membership plans. It can look up patient records, search for patients by condition or medication, and pull appointment or billing details on demand.
Draft message replies: When a patient sends a message, the AI can suggest a response for your review. You read, edit, and approve every draft before it reaches the patient. See AI-Drafted Message Replies for details.
Generate daily summaries: Each morning, the AI compiles a summary of upcoming visits, unread messages, billing alerts, and practice metrics. See Daily Practice Summaries.
Recover failed payments: When a payment fails, the system sends outreach to the patient with a link to update their payment method, and schedules retries on an escalating timeline. See Automated Payment Recovery.

AI Transparency: What You Control

Area	What the AI Does	What You Control
Message replies	Drafts a suggested response	You review, edit, and send (or discard)
Daily summaries	Compiles practice data into a morning briefing	You choose which sections to receive and when
Payment recovery	Sends outreach emails on a defined schedule	You configure timing, message tone, and can pause for specific patients
Practice questions	Retrieves data from your patient panel, schedule, and billing	You decide what to ask and whether to act on the answer

What Data the AI Accesses

When you interact with the AI assistant, it can access the following data within your practice:

Practice information

Organization name, timezone, and booking mode
Practice name, locations, and contact information
Staff roster (names, roles, specialties, availability)
Membership plans (pricing, benefits, visit limits)

Patient data (when enabled)

Patient demographics (name, date of birth, age, gender, contact information)
Medical record number and primary provider
Allergies, chronic conditions, and current medications
Membership status and billing history

Clinical records (on demand)

Recent clinical notes with SOAP content (subjective, objective, assessment, plan)
Active prescriptions (medication, strength, directions, refills)
Vitals and ICD codes from visit notes

Scheduling data

Upcoming and recent appointments
Provider, visit type, status, and reason

Billing data

Invoices, outstanding balances, and overdue amounts
Active memberships and recurring revenue
Payment history and membership status

What the AI Does Not Access

The AI assistant does not have access to:

Data from other practices in your organization (all queries are scoped to your current practice)
Other organizations’ data
Your login credentials or API keys
External medical databases or the internet
Data you have not loaded into DPC Pro

How the AI Makes Decisions

The AI assistant uses a conversational approach to help your practice. Here is how it processes a request:

You ask a question or trigger an action: For example, “Which patients have diabetes?” or “Show me today’s schedule.”
The AI reviews available context: It starts with a summary of your practice, staff, membership plans, and patient panel. For the Claude provider, patient demographics and key medical information are included in a lightweight summary, and detailed records are fetched on demand.
The AI retrieves specific data if needed: If your question requires detailed patient records, clinical notes, appointment schedules, or billing information, the AI uses built-in tools to query your practice database. Each tool call is scoped to your current practice.
The AI generates a response: Based on the context and retrieved data, the AI composes an answer. It cites sources when referencing uploaded documents.
You review the response: The AI’s answer appears in the chat interface. You decide whether to act on it, ask a follow-up question, or discard it.

Practice document search

If you enable document search for a conversation, the AI can search your practice’s uploaded documents (clinical protocols, practice policies, medical references) to inform its responses. When documents are referenced, the AI cites which sources it used. See AI Guardrails and Preferences for how to control this feature.

What the AI Does Not Do

Make diagnoses: The AI does not provide definitive diagnoses. It can surface relevant clinical information, but clinical judgment remains with your providers.
Contact patients directly: The AI never sends messages, emails, or texts to patients without a human approving the action first.
Override your decisions: If you discard a draft or ignore a suggestion, the AI does not retry or escalate.
Access data outside your practice: All queries are restricted to the practice you are currently working in. The AI cannot reach across practices, even within the same organization.

Safety and Oversight

Encryption

All conversation messages (both your questions and the AI’s responses) are encrypted at rest. Patient data referenced in AI conversations receives the same encryption as clinical records elsewhere in DPC Pro.

Audit logging

Every AI interaction generates audit log entries:

Conversation creation: Logged with the user, practice, and provider used
Patient data access: Logged each time the AI retrieves a specific patient’s records, with the patient name and data type recorded
Document access: Logged when patient-linked documents are shared with the AI provider
Tool usage: Each data retrieval the AI performs (patient lookup, appointment search, billing query, clinical notes) is individually logged

Error handling

If the AI encounters an error while processing your request, it displays a clear error message in the conversation. Errors are logged with details for troubleshooting. The AI does not silently fail or fabricate a response when something goes wrong.

Rate limiting

Your practice has a daily message limit (configurable by your organization) to prevent unexpected usage spikes. If the limit is reached, the AI will let you know and suggest trying again later.

Human-in-the-loop requirement

The AI assistant is designed as a support tool, not an autonomous agent. The guiding principle:

AI suggests, humans decide. Every patient-facing output passes through a human before it reaches a patient.
AI retrieves, humans interpret. Data lookups are presented for your review. The AI does not take action based on what it finds unless you tell it to.

Need Help?

If you have questions about the AI assistant, reach out to the DPC Pro support team at [email protected] or visit the troubleshooting guide.