AI-powered Outbound Call Bot/Platform

B2B AI Outbound Calling Platform /bot — Freelancer.com Project Brief Page 1 PROJECT PROPOSAL B2B AI Outbound Calling Platform VICIdial Integration • AI Voice Agent • Meeting Booking Project type AI voice calling platform and web dashboard Target market United States — business agencies Call language English Primary goal Qualify prospects and book meetings Dialer integration VICIdial (version to be assessed) Deployment Self-hosted backend with managed AI voice services Project objective Build a functional, tested, and deployable B2B AI outbound calling platform. The AI agent will contact eligible business leads in the United States, communicate naturally in English, qualify prospects, book meetings, and save call outcomes in a central dashboard and VICIdial. This document is intended as a project brief for posting on Freelancer.com. The final architecture and scope will be confirmed after reviewing the existing VICIdial/Asterisk environment. B2B AI Outbound Calling Platform — Freelancer.com Project Brief Page 2 1. Scope of Work A. AI Voice Agent • Natural, professional English-speaking AI voice for outbound business calls. • Configurable sales scripts, qualification questions, and conversation flows. • Meeting booking when a prospect expresses interest. • Call outcome classification, summaries, and transcripts. • Recording support only where legally permitted and properly configured. • Handling of interruptions, silence, voicemail, poor audio, and call failures. • Immediate opt-out handling and suppression of future promotional calls. B. VICIdial Integration • Assess the installed VICIdial and Asterisk versions before selecting the integration approach. • Synchronize leads, call statuses, outcomes, and external call identifiers through a verified integration method. • Implement error handling, retries, and duplicate prevention. • Evaluate two architectures: managed provider originates calls with results synced to VICIdial, or VICIdial/Asterisk originates calls and connects to the AI agent through verified SIP/media integration. • Do not assume provider compatibility; validate the chosen call path with a proof of concept. C. Backend and Database • Preferred backend: Python, FastAPI, Pydantic, SQLAlchemy, and Alembic. • PostgreSQL database for leads, campaigns, calls, appointments, consent records, suppression lists, and audit logs. • Secure REST APIs, webhooks, background jobs, structured logs, and health checks. • Idempotent event processing and safe retry behavior. D. Web Dashboard • Responsive dashboard with campaign and calling analytics. • Lead management, CSV import, search, filters, and call history. • Campaign creation, configuration, start, pause, and stop controls. • Call status, duration, outcome, summary, transcript, and permitted recording playback. • Appointment management, integration health, usage/cost tracking, and audit logs. • Do-not-call/suppression list and campaign approval controls. E. Calendar Integration • Google Calendar integration with availability checks. • Correct time-zone handling and prevention of double booking. • Book only confirmed, genuinely available meeting slots. • Save appointment details and booking status in the platform. B2B AI Outbound Calling Platform — Freelancer.com Project Brief Page 3 2. Preferred Technology Stack Layer Preferred technology Frontend React / Next.js / TypeScript Backend Python / FastAPI Database PostgreSQL AI voice Vapi or another suitable managed voice platform Telephony Compatible US telephony provider Calendar Google Calendar API Deployment Docker / Docker Compose; self-hosted backend Developers may propose alternatives with a clear explanation of technical compatibility, operating costs, and maintenance implications. 3. Suggested Milestones Milestone 1 — Technical assessment / proof of concept Inspect VICIdial/Asterisk, confirm the call architecture, identify dependencies and third-party costs, and demonstrate a basic test call. Milestone 2 — Working AI calling MVP Configure the voice agent, implement the backend/database foundation, place controlled test calls, and receive call webhooks. Milestone 3 — Dashboard and calendar Implement lead/campaign management, call history, dashboard analytics, and Google Calendar meeting booking. Milestone 4 — VICIdial integration Implement the verified integration path, synchronize call outcomes, and test retries and duplicate prevention. Milestone 5 — Testing and deployment Run end-to-end tests, fix issues, deliver deployment documentation, and demonstrate the complete workflow. 4. Required Deliverables • Complete source code and agreed intellectual-property/source-code handover. • Working frontend, backend, database schema, and migrations. • AI voice agent configuration and editable scripts. • VICIdial integration and documentation of the verified architecture. • Google Calendar integration. • API documentation, automated tests, and test results. • Docker/deployment configuration and .env.example containing placeholders only. • Setup, maintenance, troubleshooting, backup/recovery, and security documentation. B2B AI Outbound Calling Platform — Freelancer.com Project Brief Page 4 • Compliance launch checklist and a working end-to-end demonstration. 5. Security and Compliance Requirements • Secure authentication, role-based access, secret management, HTTPS, and audit logs. • Webhook signature verification where supported; protect credentials and redact secrets from logs. • No public exposure of database credentials or Asterisk management interfaces. • Build controls for lead authorization/consent records, calling windows, suppression, opt-outs, campaign limits, and auditability. • Review applicable US federal and state requirements, provider/carrier rules, and recording/transcription requirements before live campaigns. • Do not assume B2B status or a publicly listed number automatically makes every AI-generated call lawful. Human approval is required before production calling. 6. Testin