AI Voice Front Office Build for HIPAA-Regulated Medical Practice (NextGen EHR + RingCentral)

Overview We are an AI automation firm building a custom voice and messaging front office system for a single-location OB-GYN practice in California. We need a developer or small team to quote and build the system. We handle the client relationship, requirements, and ongoing account management. You build and support the platform. This is a real project with a live client, not a concept. We are collecting a firm quote so we can price the engagement accurately. If the quote and the developer are a fit, this becomes a paid build with a target start inside 30 days, and we have additional practices in the pipeline behind it. Current environment EHR / practice management: NextGen (also functioning as the contact system of record) Telephony: RingCentral, cloud hosted, managed by a local IT vendor. Three inbound lines with a phone tree, plus a separate billing option and an authorizations option. Patient messaging: Doctible (SMS reminders and two-way patient texting) Patient portal: Separate portal with inbound patient messages After hours: Third party answering service, billed per call, to be eliminated Staff: Five people answering phones, all of whom have other primary duties. One LVN on a newly opened triage line. Volume Roughly 300 to 400 inbound calls per business day, with Monday peaks reported above 600 Roughly 10,000 calls per month across 22 business days Phones are closed 12pm to 1pm daily and route to the answering service Answering service billed approximately 303 calls in the most recent month Hold times reported at 10 to 15 minutes at peak Triage line reported over 90 messages on a single Monday Billing line is staffed by one person who regularly misses calls and returns them later Scope, Phase 1 Quote each item separately. We need line-item pricing, not a single blended number. AI voice agent, all inbound calls, 24/7. Natural conversation quality, not IVR-style. Must fully replace the after-hours answering service. Handles overflow during business hours so no call goes unanswered. Intent routing. Classify and route calls: scheduling, billing, authorizations, clinical or triage, prescription refills, general. Route to the correct person or queue with context attached. Appointment management via NextGen API. Book, reschedule, cancel, and confirm without human involvement. Must respect provider-specific scheduling rules and booking filters. Practice currently double books manually and needs guardrails against overbooking. Waitlist and auto-rebooking. Structured waitlist replacing a manual list. Automatic outreach and fill when a slot opens. Triage handling. Detect clinical intent, capture structured intake, prioritize by urgency, notify the on-call clinical staffer immediately by SMS and email with a summary before transfer or callback. Two-way SMS. Inbound and outbound patient texting, either integrated with the existing Doctible workflow or replacing it. Quote both paths. Backend CRM / data layer. Staff should not need to learn a new system. One to two admin licenses. Include your recommended platform and its monthly cost. Reporting dashboard. Call volume, intent breakdown, disposition (handled, transferred, abandoned), containment rate, missed and returned calls, automated callback on hangups, and deep links to the existing RingCentral call recordings. Telephony integration. RingCentral SIP or API integration. Describe your approach and any carrier or number porting implications. Scope, Phase 2 (quote separately, may be deferred) Billing line automation. Retrieve patient account balance from NextGen and discuss it on the call, plus card-present-not-required payment capture. This is PHI and PCI in the same flow. Quote it as a standalone module and tell us what identity verification, encryption, and payment processor architecture you would use. If you would advise against automating this, say so and explain why. Call recording and QA scoring. Automated grading of human-handled calls against configurable criteria, with trend reporting. Patient portal message triage. Ingest and classify inbound portal messages. Explicitly out of scope Lab or test result delivery to patients Medication refill approvals or clinical decision making Prior authorization processing Any clinical advice These route to humans by design. Compliance requirements You must be willing to sign a BAA. Non-negotiable. All PHI handling must be HIPAA compliant end to end, including any third party LLM, transcription, telephony, and storage vendors in your stack. Name every vendor that touches data and confirm each will execute a BAA. California-specific requirements apply, including two-party consent for call recording. Describe your data retention, encryption at rest and in transit, and access control approach. What your proposal must include Proposals without this breakdown will not be reviewed. Build cost per numbered scope item above, Phase 1 and Phase 2 separated Monthly recurring cost to run and maintain, itemized into: Your maintenance and support fee Platform and infrastructure costs you control Pass-through usage costs (LLM tokens, speech to text, text to speech, telephony minutes, SMS) with a projected monthly figure at 10,000 calls per month and your assumed average call length Timeline to production, with milestones Your stack, named specifically: voice platform, LLM, STT and TTS providers, orchestration layer, CRM, hosting Prior HIPAA voice AI work. Include at least one healthcare deployment with call volume and containment rate achieved. Redacted client names are fine. Your realistic containment estimate for this call mix, meaning the percentage of the 10,000 monthly calls fully resolved without a human. Be honest. The practice manager believes most of her calls require a human because of refills, results, and clinical concerns. Tell us what you actually think is achievable and why. Known risks, especially anything you expect to be difficult with the NextGen API Screening questions Have you built a production voice AI system that integrates with NextGen specifically? If not, which EHRs have you integrated with? Will you sign a BAA, and will every vendor in your stack do the same? What containment rate have you achieved on a comparable inbound medical practice deployment? What is your typical time from kickoff to first production call? Are you available for ongoing maintenance and support after launch, and at what rate? Engagement notes We are the client-facing party. You work with us, not directly with the practice. White label. The system ships under our brand. We are evaluating a small number of quotes and will move quickly on the right fit. Long-term potential. This is the first of several practice deployments if the build goes well.