A person identifying herself as Sarah is available to speak. No appointment history or outreach reason has been disclosed.
Privacy & scope: A “yes” and the dialed number do not verify the patient. Anyone can request a human or stop outreach.
SHARED CONTEXT
AI identity disclosed · Private setting confirmed · Identity not yet verified · Recording off
03 / VERIFY BEFORE DISCLOSING Same call
A familiar check, with a callback option.
Your voice agent
Before I discuss your record, could you tell me your full name and date of birth? If you prefer, you can call the clinic using the number you already have.
Sarah · patient
That’s okay. Sarah Morgan, June 12, 1982.
Your voice agent
Thank you, Sarah.
Full name and date of birth follow the clinic’s configured verification policy; they are not a universal HIPAA formula.
Athena returns an appointment identifier. Read-back matches Sarah, annual physical, Dr. Patel, Oak Street, October 13 and 8 a.m.
Privacy & scope: A timeout, queued request or “request received” is not a booking confirmation.
SHARED CONTEXT
Athena booking verified · Appointment ID retained securely · Coverage review still open · No duplicate appointment
10 / KEEP CLINICAL QUESTIONS WITH STAFF Same call
“Do I need to fast?”
Sarah · patient
Do I need to fast? Last time they had me come back for blood work.
Your voice agent
I’ll ask Dr. Patel’s team to confirm that for this visit. I don’t want to give you the wrong instructions.
Sarah · patient
Okay. A portal message would be fine.
The AI makes no decision about fasting, lab tests or medication. It captures the question for an authorized staff member.
AI TEAM AT THIS MOMENT
Route the question with its context.
Supervisor Voice · patient conversation · Acknowledges the question Policy · configured clinic rules · Applies configured rules Human Handoff · authorized staff work · Prepares staff request
CONNECTED SYSTEMS & KNOWLEDGE
Clinic routing policyClinical staff work queue
Prepare a task for Dr. Patel’s team: “Patient asks whether fasting is needed for the booked visit.” Preferred reply: secure portal.
Privacy & scope: No general fasting advice, lab orders, medication instructions, symptom assessment or reassurance is generated.
SHARED CONTEXT
Booking remains confirmed · Fasting question unanswered · Dr. Patel’s team owns reply · Portal reply requested
11 / CHOOSE A PRIVATE CHANNEL Same call
“Please don’t text me the details.”
Your voice agent
Would you like the appointment details in your secure patient portal?
Sarah · patient
Yes, please. Don’t text me the details—my kids use my phone.
Your voice agent
I’ll use the portal for the details and note that preference for the office. Your portal may still send its usual generic new-message alert. Is that okay?
Sarah · patient
That’s fine, as long as it doesn’t say what the appointment is.
The confirmation route is separate from permission to make the original AI call.
Appointment details go inside the authenticated portal. Only a generic alert is allowed outside it, if its content and destination are verified.
Privacy & scope: If a connected portal cannot honor this preference, stop automated sending and route to staff. Do not promise control of settings the platform cannot change.
Clinic work queuesApproved portal messaging routeRestricted workflow ledger
Portal message accepted; patient reading not yet known. Preparation task and insurance task have recorded owners and clinic-defined response deadlines.
Privacy & scope: A failed message never triggers another appointment. Unreconciled writes stay pending and alert staff.
SHARED CONTEXT
Booking unchanged · Portal accepted ≠ read · Staff task found once · Open tasks have owners
14 / STAFF FOLLOW-UP Before the visit
The clinical answer comes from the clinic.
HUMAN-OWNED FOLLOW-UP
Dr. Patel’s team answers the preparation question, and billing staff complete the insurance review. Both staff-authored outcomes are delivered through the approved secure channel and acknowledged by Sarah.
The clinical task records the authorized staff answer, sender, delivery receipt and Sarah’s acknowledgment.
The billing team records its review outcome without the AI inventing eligibility or a coverage promise.
The billing outcome is delivered in the portal and Sarah’s acknowledgment closes the administrative task.
Human staff supply both answers. The workflow records authorship, delivery and acknowledgment without generating clinical advice or eligibility.
AI TEAM AT THIS MOMENT
Human decisions return with receipts.
Supervisor Domain Specialist · billing and eligibility routing · Records billing review Communications · approved channels · Verifies portal delivery Human Handoff · authorized staff work · Records staff answers Outcome · evidence-backed disposition · Closes both tasks
CONNECTED SYSTEMS & KNOWLEDGE
Clinical staff work queueBilling work queueSecure patient portal
Preparation and billing reviews are completed by authorized staff; both portal messages are delivered and acknowledged by Sarah.
Privacy & scope: Do not paraphrase clinical instructions into new advice. A task exists, staff replied and patient read are different events.
SHARED CONTEXT
Staff-authored preparation answer acknowledged · Billing review completed by staff · Portal delivery verified · No AI clinical or eligibility decision
15 / HELP HER GET THERE Before October 13
A reminder that respects the channel.
APPROVED REMINDER WORKFLOW
Before the visit, the reminder owner rechecks the appointment and communication restriction. The portal reminder is delivered and Sarah confirms receipt.
Suppress duplicate reminders across the platform and clinic systems.
Do not send if the visit was canceled or moved.
The portal records delivery and Sarah’s confirmation; no detailed SMS is sent.
The clinic selects one reminder owner. The platform must not assume it can override all Athena reminders or patient portal notification settings.
The booking remains current. The configured service delivers one portal reminder, and Sarah confirms receipt through the portal.
Privacy & scope: Recheck permission and confidential communication preferences at each send, including across automated clinic systems.
SHARED CONTEXT
Appointment still current · Single reminder owner · Portal reminder delivered · Patient receipt confirmed
16 / MEASURE THE ACTUAL RESULT Visit day · later status event
Booked is a milestone. Coming in is the result.
CLINIC-RECORDED COMPLETION
The clinic records Sarah as arrived and later marks the appointment completed. Every scheduling, billing, clinical-message, reminder and attendance commitment now has final evidence.
The clinic records the appointment completed. Booking, staff answers, billing review, portal delivery, reminder receipt and attendance are all reconciled and closed.
Privacy & scope: If no status feed is available, report attendance unknown. No-show and canceled visits use staff-approved follow-up rules.
SHARED CONTEXT
Visit completed · clinic recorded · Campaign outreach closed · All messages delivered and acknowledged · No unresolved tasks or exceptions
WHEN THE CALL TAKES ANOTHER PATH
The safeguards change what happens next.
The successful call above is one path. These alternatives are part of the same workflow.
Someone else answers / wrong number
Caller: “She’s not here.” AI: “Thank you. I’ll try another time.”
Do not disclose the physical, clinician or patient record. A reported wrong number stops outreach to that number and creates a contact-review task. A relative is not automatically authorized; a proxy requires the clinic’s identity and authority checks. Requests involving minors or uncertain guardianship route to staff.
Voicemail / no answer
“This is Oak Street Primary Care’s AI assistant. Please call our office.”
Use a neutral message only if contact instructions permit it, and provide the clinic’s verified callback number. Disclose no visit reason, date of birth or overdue status. Otherwise leave nothing. Apply bounded retries and stop conditions.
Identity mismatch / scam concern
Patient: “I don’t give my birthday on calls.” AI: “That’s fine. Please call the clinic using the number you already have.”
Stop patient-specific discussion. No hints, repeated guessing, DOB read-back or unsolicited verification link. A clinic-approved staff route handles additional verification or accessible alternatives.
Stop calling / not a good time
Patient: “Please stop these calls.” AI: “Of course. I’ll stop this outreach.”
Apply the stop immediately without requiring identity verification or persuasion. Persist the request, cancel queued attempts and propagate to relevant outreach systems. If the person asks for a later call, record that preference within approved hours; do not treat it as broader consent.
Already completed / visit type unclear
Patient: “I had my physical somewhere else.” AI: “Thanks for letting me know. I’ll ask the office to update the outreach list.”
Pause the campaign and assign staff reconciliation. Do not mark a clinical care gap closed on a verbal report alone. Annual physical, Medicare Annual Wellness Visit and Welcome to Medicare visit need distinct clinic-approved mappings.
Clinical question / human requested
Patient: “Can I speak to someone at the office?” AI: “Of course. I’ll connect you.”
Human Transfer sends only authorized context. Symptoms go to the clinical team without AI assessment, triage or reassurance. A caller-stated emergency triggers the clinic’s fixed emergency instruction; no symptom scoring. If transfer fails, say so and use the approved fallback without suggesting it is safe to wait.
No slots / uncertain booking / dropped call
AI: “I haven’t been able to confirm that appointment. I can connect you with scheduling.”
No invented slot, unauthorized provider switch or success claim. A dropped call before approval causes no booking. After a submitted transaction, reconcile its result before retrying. A new session repeats the required identity check; the ledger preserves task history.
Message failure / cancellation / no-show
Keep booking, delivery, attendance and open staff tasks separate. A failed portal message routes to staff; it does not authorize detailed SMS. Cancellation or a no-show triggers only the clinic’s approved follow-up process. One system owns reminders, and completed visits suppress the original recall campaign.
FROM AN INVITATION TO A VISIT
Sarah gets a time that works. The clinic can see what happened.
A confirmed appointment
The patient approves the details, and Athena confirms the saved booking before the agent announces success.
Human expertise stays human
Clinic staff answer the preparation question and billing staff record the coverage review. Sarah acknowledges both in the portal.
Follow-through is measurable
Private messages, reminders, staff tasks and clinic-recorded visit completion each carry their own evidence before the workflow closes.
Integrations, permissions and compliance controls are configured for each healthcare environment. No clinical judgment is performed.
01 / 16
Who We Are
The Team Behind rundesk360
Complementary perspectives — voice at scale, healthcare operations, and customer impact — building one platform together.
The Voice
Vinay Krishna
Vinay led speech products for the Indian market at Microsoft, including the first ASR and TTS models for Indian languages and Azure Speech deployment for Jio, India’s leading telecom provider. That foundation gives him firsthand expertise in the realities that determine whether voice AI works at scale. At Rajneethi, a reputed premium political consulting brand operating in multiple countries, he took a services firm to 90% software revenue and 80× growth in five years, with recent work focused on agents that act—not simply models that talk.
The Mind
Sawad Thotathil
Sawad Thotathil, MD, MS, MHM, is a physician, healthcare technology founder, and experienced operator who transforms complex healthcare problems into scalable products and businesses. His background spans clinical operations, AI-enabled care, enterprise technology, and healthcare entrepreneurship, including leadership of an international joint venture and a focus on driving innovation and scaling programs across technology implementations and clinical operations.
The Muscle
Anit Parikh
Anit is a global tech executive with over 30 years of experience scaling high-growth businesses. His leadership background includes serving as Managing Director at Slalom, founding PwC’s Salesforce Pharma and Life Sciences practice, and overseeing professional services as a Regional Director at Salesforce.
