AI Outbound Calling Agent for Real Estate Leads

Build a Cost-Efficient AI Outbound Calling System for Real Estate Leads PROJECT OVERVIEW We are seeking an experienced AI voice-agent developer to build a reliable outbound calling and lead-qualification system for a real estate professional. The system will: * Import realtor-supplied leads from an Excel spreadsheet * Call approved leads during authorized calling hours * Conduct natural and adaptive conversations * Qualify and categorize each lead * Save call outcomes in a simple lead database * Immediately alert the realtor when a lead requests direct contact * Generate a daily activity report Our highest priorities are: * Excellent voice quality * Natural, humanlike conversations * Strong conversational logic and reasoning * Accurate lead qualification * Reliable automation * Low ongoing operating costs * Simple maintenance after project completion * Complete ownership and editability of the finished system We do not need an expensive custom dashboard, proprietary CRM, or overengineered software platform. The goal is to combine a high-quality AI voice and reasoning system with inexpensive or free supporting infrastructure. We are willing to invest in the components that directly improve voice quality, response latency, speech recognition, conversational intelligence, and lead qualification. Supporting functions such as lead storage, internal notifications, reporting, and workflow automation should use free or low-cost tools whenever practical. OUTBOUND AI CALLING AGENT The AI agent must: * Call leads supplied by the realtor through an approved Excel spreadsheet * Operate Monday through Saturday * Place calls only between 8:00 a.m. and 5:00 p.m. * Respect each lead’s local time whenever location or time-zone information is available * Follow an approved real estate lead-qualification framework * Sound natural, professional, confident, and conversational * Respond appropriately to interruptions and unexpected answers * Ask relevant follow-up questions based on each lead’s responses * Avoid rigidly reading from a fixed script * Identify the lead’s needs, intentions, motivation, and timeline * Handle common questions and objections * Detect voicemail, unanswered calls, disconnected lines, and invalid contact records * Recognize requests for no further contact * Avoid calling suppressed or opted-out leads * Produce an accurate outcome and summary after every call Voice quality and conversational intelligence are more important than an elaborate user interface. The system may use Retell AI, ElevenLabs, Vapi, Bland AI, or another established voice platform. We are open to recommendations that provide better quality, reliability, reasoning, or cost efficiency. CONVERSATIONAL LOGIC AND LEAD QUALIFICATION The agent should adapt its questions and responses based on whether the lead is: * Interested in buying * Interested in selling * Considering both buying and selling * Exploring future options * Not currently ready * Already represented by another real estate professional * Requesting a later conversation * Ready to speak directly with the realtor * Not interested * Requesting no further contact When appropriate, the agent should gather: * Lead identity * Buying, selling, or combined interest * Property type * Preferred location * Expected timeline * Motivation * Financing or preapproval status * Existing realtor representation * Preferred follow-up time * Questions or concerns * Other relevant qualification details * Recommended next action The agent should recognize conversational context and avoid forcing every lead through the same sequence of questions. LEAD CLASSIFICATION After each call, the system must automatically classify the lead. Expected classifications include: * New or uncontacted * No answer * Voicemail * Invalid contact record * Disconnected line * Not interested * Follow up later * Warm lead * Interested lead * Hot lead * Realtor callback requested * Appointment requested * Already represented * Do not contact The developer may recommend additional categories when they improve the workflow without creating unnecessary complexity. LEAD IMPORT AND EXPRESS PERMISSION VERIFICATION Due to the legal requirements that apply in the region where the system will operate, every lead imported into the calling system must have provided express permission to be contacted by an AI-powered voice caller. The lead-import process will begin with an Excel spreadsheet supplied by the realtor. Each imported record must include fields that allow the realtor to confirm that: * The individual has given express permission to be contacted * The permission includes contact through an AI-powered or artificial-intelligence voice caller * The permission applies to the telephone number contained in the spreadsheet * The source of the lead is identified * The date of permission is recorded when available * Supporting notes or evidence can be referenced when applicable * The lead has not withdrawn permission * The lead is not marked as do not contact Before any calling campaign can begin, the realtor must review the imported lead list and co-sign or electronically attest that the listed leads have provided the required permission to be called by the AI voice agent. The system must not automatically call a newly imported spreadsheet immediately after upload. A required approval step must occur before campaign activation. The approval workflow should: * Import the Excel spreadsheet into a review queue * Validate that all required permission fields are present * Flag incomplete or invalid records * Prevent unapproved leads from entering the calling queue * Allow the realtor to review the imported records * Require the realtor to certify or co-sign the approved lead list * Record the identity of the person approving the list * Record the date and time of approval * Preserve the version of the spreadsheet or lead batch that was approved * Create an auditable record of the realtor’s certification * Release only approved leads into the active calling queue The realtor’s certification should clearly confirm that the realtor has reviewed the lead list and verifies that each approved lead has provided express permission to be contacted by the AI voice caller. The system should make it impossible, or reasonably difficult, to bypass this approval step accidentally. If a lead does not contain the required permission information, the lead must remain blocked from calling until the missing information is supplied and the lead is approved. LOW-COST LEAD DATABASE The system must automatically save lead information, permission records, approval status, call outcomes, summaries, and follow-up information in a simple database. Free or low-cost options are preferred, including: * Google Sheets * Supabase free tier * Airtable free tier * Baserow * NocoDB * Another comparable free or inexpensive option A paid CRM is not required unless there is a strong operational reason to use one. The database should store: * Lead identity * Contact record * Lead source * Express-permission status * Permission date, when available * Permission source or evidence reference * Realtor review status * Realtor certification or co-signature status * Approval date and time * Approved lead-batch identifier * Call date and time * Number of call attempts * Call outcome * Lead classification * Qualification details * Conversation summary * Follow-up date * Preferred callback time * Opt-out status * Transcript or recording reference, when available * Recommended next action * Error status, when applicable The system must: * Prevent unapproved leads from being called * Prevent unnecessary duplicate calls * Permanently suppress opted-out leads * Allow calling campaigns to be paused * Allow records to be reviewed, filtered, edited, and exported * Maintain an audit trail of permission and approval records * Remain easy to operate without an expensive CRM subscription IMMEDIATE HOT-LEAD ALERT When a lead requests direct contact from the realtor, the system should immediately send an automated internal alert through the Telegram Bot API. Telegram will be used only as a technical notification integration within the completed system. It will not be used for pre-contract communication with applicants. The alert should include: * Lead identity * Lead classification * Buying or selling interest * Brief conversation summary * Reason for requesting contact * Relevant property or timeline details * Preferred callback time, if provided * A clear instruction to contact the lead within 10 minutes The workflow should prevent duplicate alerts. This integration may be quoted as an optional component if excluding it significantly reduces the initial development cost. We are technically capable of configuring this integration ourselves if necessary. DAILY ACTIVITY REPORT The system should generate a professionally formatted daily email report containing: * Total leads processed * Total leads approved for calling * Total leads blocked because of incomplete permission records * Total calls attempted * Calls answered * Voicemails * No-answer outcomes * Invalid or disconnected contact records * Warm leads * Interested leads * Hot leads * Callback requests * Follow-up opportunities * Appointment requests * Not-interested leads * Opt-out requests * Failed calls * Failed automations * Strongest opportunities * Leads requiring realtor action * Recommended next steps The report should be concise, easy to scan, and inexpensive to deliver through a standard email provider or API. This reporting workflow may also be quoted as an optional component if excluding it meaningfully reduces development cost. We are technically capable of configuring the reporting workflow ourselves if necessary. CALLING SCHEDULE AND RETRY LOGIC The developer must configure practical rules for: * Approved calling days * Approved calling hours * Lead-local-time restrictions * Realtor approval before calling * Maximum call attempts * Time between attempts * Voicemail handling * Follow-up scheduling * Duplicate-call prevention * Opt-out suppression * Campaign pausing and restarting * Failed-call handling * Failed-automation logging * Later callback requests * Immediate realtor callback requests The realtor or system administrator must be able to pause calling without modifying the source code. COST-EFFICIENT TECHNICAL ARCHITECTURE The system should be intentionally designed to minimize recurring expenses. Development effort and operating costs should prioritize: * High-quality AI voice generation * Low conversation latency * Strong language-model reasoning * Reliable speech recognition * Natural conversational behavior * Accurate lead qualification * Accurate lead categorization The architecture should minimize spending on: * Paid CRM subscriptions * SMS services * Premium databases * Custom dashboards * Redundant automation platforms * Unnecessary hosting * Overlapping software subscriptions * Proprietary components that create avoidable vendor lock-in A possible technical stack may include: * Retell AI, ElevenLabs, Vapi, Bland AI, or another voice platform * A high-quality language model suitable for real-time conversations * Google Sheets, Supabase, Airtable, Baserow, NocoDB, or a comparable database * Telegram Bot API for internal alerts * A standard email provider or API for daily reporting * Lightweight custom code or automation * Low-cost hosting only when required This list is not mandatory. We welcome a simpler or more effective architecture that protects voice quality and reliability while reducing recurring expenses. Self-hosted or open-source components are acceptable when they meaningfully reduce costs without making the system unreliable or difficult to maintain. CONSENT, OPT-OUT, AND RECORDKEEPING CONTROLS The realtor will supply the lead lists and will be responsible for reviewing and certifying the permission status of the leads. The developer is not being hired to provide legal advice. The developer is responsible for accurately implementing the required technical controls. The system must include: * Lead-source tracking * Express-permission status tracking * Confirmation that permission includes AI voice contact * Storage of permission-related notes or evidence references * Realtor review and approval status * Realtor co-signature or electronic certification * Lead-batch version tracking * Internal do-not-contact list * Immediate recording of opt-out requests * Permanent suppression of opted-out leads * Calling-hour restrictions * Call timestamps * Call-attempt history * Script-version tracking * Exportable records * Complete call-outcome logs * An auditable history of lead imports and approvals The system should process only leads that meet the approved technical requirements established for the campaign. ERROR HANDLING AND RELIABILITY The system must account for: * Voice-platform errors * Failed calls * Failed database updates * Failed internal alerts * Failed daily reports * Duplicate records * Missing lead data * Missing permission fields * Missing realtor approval * Invalid contact records * Calls outside approved hours * Automation failures Critical failures should be logged and surfaced clearly. The system must not silently lose: * Qualified leads * Call results * Opt-out requests * Permission records * Realtor approvals * Required follow-up information REQUIRED DELIVERABLES The completed project should include: 1. Fully functioning outbound AI calling agent 2. Real estate qualification framework 3. Adaptive conversational logic 4. Excel lead-import workflow 5. Express-permission validation workflow 6. Realtor review and co-signing or electronic certification workflow 7. Automated lead classification 8. Free or low-cost database integration 9. Automatic post-call summaries 10. Immediate hot-lead alert integration, if included 11. Daily email activity report, if included 12. Calling schedule and retry logic 13. Duplicate-call prevention 14. Opt-out suppression 15. Error handling and failure logging 16. End-to-end testing 17. Written setup and operating documentation 18. Recorded system walkthrough 19. Complete transfer of the system after acceptance 20. An editable source-code project suitable for ongoing internal management EDITABLE CODEBASE AND GOOGLE ANTIGRAVITY COMPATIBILITY After completion, our team will manage, maintain, and modify the system independently. We are not seeking ongoing software management or a long-term maintenance arrangement. All custom code must be delivered in a standard, organized, editable project format that can be opened and modified through Google Antigravity IDE. The final delivery should include: * A complete project folder or Git repository export * A downloadable archive of the complete codebase * Clearly organized source files * Dependency and installation files * Configuration templates * An example environment-variable file containing no live secrets * Setup and deployment instructions * Database schema or migration files * API integration code * Prompt and conversation-logic files * Automation exports * Instructions for running and testing the system locally * Instructions for deploying or redeploying the system The project must not depend entirely on an inaccessible no-code workspace or contractor-controlled account. When no-code or low-code tools are used, all workflows must be exported in their native editable format. Any supporting scripts, functions, prompts, schemas, and configuration files must also be provided. The final codebase should be understandable and maintainable by another technically capable person using Google Antigravity IDE. TESTING REQUIREMENTS The system should be tested across representative scenarios, including: * Interested buyer * Interested seller * Immediate callback request * Later follow-up request * Uncertain timeline * Existing realtor representation * Not-interested response * Opt-out request * No answer * Voicemail * Invalid contact record * Disconnected line * Interrupted conversation * Unexpected question * Duplicate lead * Lead without permission information * Lead without realtor approval * Approved lead * Rejected lead * Duplicate spreadsheet import * Call outside approved hours * Failed database update * Failed notification * Failed report * Failed voice-agent workflow The project will be considered complete after the primary workflows have been successfully demonstrated from beginning to end. ACCOUNT OWNERSHIP AND HANDOFF The developer may use temporary development accounts while building and testing the system. After the project is accepted, the completed system must be transferred or reproduced under accounts controlled by us. The handoff must include: * Voice-agent configuration * Prompts * Conversation logic * Automation workflows * Database structure * Lead-import and approval workflow * Permission-verification fields * Realtor certification workflow * Notification configuration, if included * Reporting configuration, if included * Source code * Scripts * API setup instructions * Environment-variable documentation * Deployment instructions * Administrative access * All other materials required to operate and modify the system The completed system must not remain dependent on an account controlled exclusively by the contractor. Contractor access to production systems and confidential lead data must be removed after the transfer is complete. PROJECT APPROACH We want an experienced implementer who can complete this project efficiently without overengineering it. The business owner is technically capable of building and maintaining the system but is outsourcing the initial implementation because of limited time and the need to focus on business development. The selected freelancer should be able to: * Work independently * Make practical technical decisions * Avoid unnecessary complexity * Use existing tools effectively * Build organized and maintainable workflows * Document the system clearly * Complete a clean handoff * Keep recurring expenses low * Focus development effort on voice quality and conversational intelligence This is a focused implementation and handoff project, not a request for ongoing software management. PROPOSAL INSTRUCTIONS Please submit a fixed-price bid for the complete project. Your proposal should include: * Recommended technical stack * Fixed-price quote * Estimated completion timeline * Relevant Upwork portfolio examples or attachments * Brief description of the proposed architecture * Estimated recurring software expenses * Required paid tools or subscriptions * Optional components priced separately * Confirmation that the complete system and editable codebase can be transferred after completion