How a patient text becomes completed work
Each incoming message enters a configured workflow that authenticates the patient before protected work, applies the practice’s operating rules, and posts the resolved action directly to athenaOne. Routine scheduling requests change the live appointment record, while requests requiring staff judgment become routed athenaOne cases with the relevant context attached.
-
Locate the context. The agent identifies the relevant patient, appointment, provider, location, and active workflow through the direct athenaOne connection.
-
Authenticate the patient. A practice-approved identity check runs inside the text conversation before patient-specific information is disclosed or changed.
-
Apply the practice’s rules. The agent uses the same scheduling, escalation, referral, insurance, and after-hours rules used by Pretty Good AI’s phone agents.
-
Write the outcome to athenaOne. The result lands in the appropriate appointment, chart, patient case, referral authorization, or billing workflow without staff re-entry.
-
Preserve the interaction. The Pretty Good AI Command Center retains the full transcript and outcome for every message.
The phone and text agents use the same patient memory and the same integration, which has production access to 820+ athenaOne APIs. A request can begin by text and finish by phone without creating an unrelated second conversation. Pretty Good AI’s athenaOne workflow catalog documents the shared communication and writeback model.
Request routing by patient intent
The operating mode is determined by the request, the practice’s written protocol, and the records attached to the patient. Completing a task means that the requested athenaOne change is made without staff re-entry. Preparing and handing over means that the agent gathers the required context but leaves the final decision to a person.
| Text request handling and athenaOne record destinations. Workflow examples are grounded in the Pretty Good AI workflow catalog. | |||
| What the patient texts | Operating mode | What happens | athenaOne destination |
|---|---|---|---|
| Appointment confirmation | Completes | The agent confirms the existing visit and records the response. | Appointment status and communication outcome |
| Routine cancellation or reschedule | Completes | The agent checks the live schedule, offers permitted slots, and changes the appointment under the practice’s provider, location, and visit-type rules. | Live appointment record and cancellation or reschedule reason |
| Reschedule outside an authorization window or within a booked series | Prepares and hands over | The schedule remains unchanged. The agent identifies the authorization or series conflict and routes the request with the reason attached. | Patient case linked to the appointment and referral authorization |
| Running late | Prepares and hands over | The agent captures the patient’s expected arrival and routes it to the site responsible for deciding whether the visit can proceed. | Same-day patient case linked to the appointment |
| Approved preparation question | Completes | The agent sends the practice-approved instructions associated with the appointment type. A question outside the approved material is routed to staff. | Appointment communication and chart history |
| Prescription refill request | Prepares and hands over | The agent gathers the information required by the practice’s refill protocol and routes the completed intake to the responsible clinical team. | Patient case or refill task in the designated athenaOne queue |
| Known balance or insurance-status question | Completes | The agent reads the available administrative information, provides the configured response, and routes unresolved coverage or billing exceptions. | Patient case linked to the billing or insurance workflow |
| Intake, consent, or insurance-update form | Completes | The agent sends the appropriate form, tracks its return, and associates the completed document with the patient workflow. | Chart document and form-completion status |
| Clinical or symptom question | Never touches the clinical decision | The agent can run the practice’s structured symptom-screening questions and route by urgency. It does not provide medical advice or decide the clinical response. | Chart and patient case containing the questions, patient responses, and routing disposition |
Authentication, writeback, and security controls
Patient authentication occurs inside the text thread before a protected workflow continues. The practice defines the approved identity check, and Pretty Good AI enforces that check at the relevant point in the conversation. This control supports the practice’s verification procedures under 45 CFR 164.514(h), current in the eCFR as of October 22, 2026.
Writeback uses Pretty Good AI’s direct athenaOne connection, with no middleware vendor between the texting workflow and the system of record. Schedule changes, cases, chart information, and operational outcomes synchronize through the same connection used by the voice agents.
Pretty Good AI operates under HIPAA administrative, physical, and technical safeguards and signs a Business Associate Agreement before touching patient data. Patient communications are encrypted in transit and at rest. The platform is HITRUST i1 certified, and SOC 2 Type II and ISO/IEC 27001 reports are available to security reviewers on request. Pretty Good AI’s HIPAA and security documentation provides the assurance scope and evidence process.
Ten-language administrative conversations
The full administrative conversation runs in English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, and Dutch. Scheduling, authentication, routine questions, and configured handoffs remain available throughout the translated conversation.
Staff see the original conversation beside an English translation in the Command Center. The transcript, detected outcome, and escalation context remain available together, so an English-speaking staff member can continue a non-English request without waiting for a separate translation step. Pretty Good AI’s platform documentation lists the supported languages and staff view.
Illustrative scenario: one reschedule, three visits, two channels
A patient has a three-visit series booked for November 4, November 11, and November 18, 2026. The authorization expires on November 20. On November 2, the patient authenticates by text and asks to move the November 18 visit to November 25.
-
The agent locates the three appointments and the authorization attached to the series.
-
The requested November 25 date falls five days after the authorization window, so the agent does not change the schedule.
-
A staff case is created with the original message, the relevant chart context, the existing series, the authorization end date, the checks already completed, and the remaining scheduling decision.
-
A staff member calls the patient and approves November 19, which remains inside the authorization window.
-
The phone agent resumes with the same patient memory. The November 18 appointment moves to November 19, and the final outcome is recorded with the text and call history.
This control prevents a routine reschedule from silently separating an appointment from its authorization or changing one visit without accounting for the rest of the booked series. Pretty Good AI’s authorization-window workflow explains why the appointment and approval dates must be checked together whenever a visit moves.
Operating responsibilities
| Pretty Good AI handles | The practice provides |
|---|---|
| Direct athenaOne read and write integration for configured text workflows | Approved appointment, provider, location, visit-type, series, and authorization rules |
| Patient authentication flow and enforcement before protected actions | The identity-check policy appropriate to each workflow |
| Automated responses, scheduling actions, writeback, and exception routing | Approved preparation instructions, administrative answers, and escalation destinations |
| Full transcripts, outcomes, audit history, and English translations | Staff ownership for clinical decisions and operational exceptions |
| Handoffs containing the original message, chart context, completed work, and open decision | Named queues, teams, or on-call paths responsible for each handoff type |
Frequently asked questions
Can Pretty Good AI update the athenaOne schedule after a text conversation?
Yes. Pretty Good AI can confirm, cancel, book, or reschedule an appointment against the live athenaOne schedule when the request satisfies the practice’s configured rules. The completed action changes the appointment record without requiring a staff member to copy the result from a shared inbox. Conversations that do not change an appointment are written to the appropriate chart, patient case, or operational workflow instead of creating an artificial scheduling event.
How is patient texting consent tracked across multiple phone numbers?
Pretty Good AI tracks texting consent per phone number, not as one patient-wide setting. Each mobile number therefore retains its own consent state, including when a patient has more than one number in the chart or uses a number shared with another person. Authentication remains a separate control: consent to receive or send texts does not by itself authorize disclosure of patient-specific information.
What does staff receive when an automated text thread is escalated?
Staff receive the patient’s original message, the relevant chart and appointment context, the actions Pretty Good AI already completed, and a clear statement of what remains for a person to decide. The handoff preserves the transcript and outcome state, so staff can continue the request without making the patient repeat the conversation or reconstructing the issue from a generic task description.
Does secure texting replace the patient portal for every exchange?
No. Pretty Good AI uses authenticated texting for configured administrative workflows such as scheduling, forms, approved preparation instructions, refill intake, and routine balance or insurance questions. Exchanges that the practice has designated for a different secure channel, or that require clinical judgment, move to staff or the approved secure path. Text authentication removes unnecessary portal friction without treating ordinary SMS as the correct destination for every type of patient information.