Build Roadmap

POC to production in eight phases.

The POC has already validated feasibility. The full product is a larger platform build because it connects voice AI with campaign strategy, dealer intelligence, scheme rules, order capture, Salesforce workflows, MCC/Oracle data, and auditable governance.

Build Phases

Phase 01
In Progress

Architecture & Documentation Pack

Client-approved architecture brief, module definitions, integration contracts, governance requirements, and project plan.

Phase 02
Upcoming

Campaign Manager Data Model & UI

Database schema for campaigns, dealer audiences, call queue, retry rules, and suppression. Admin UI for campaign creation and monitoring.

Phase 03
Upcoming

Scheme & Product Catalog Ingestion

Scheme metadata model, eligibility rule engine, balance lookup, and product catalog with SKU/pack-size/order-unit structure.

Phase 04
Upcoming

Sales-Specific Agent Mode & Prompts

New SALES conversation mode, campaign-objective loading, scheme pitch logic, dynamic tonality controller, and objection-handling RAG integration.

Phase 05
Upcoming

Order Capture & Validation

Structured order extraction from speech, dealer confirmation loop, order validation engine against catalog and scheme constraints, and draft management.

Phase 06
Upcoming

Salesforce & MCC/Oracle Adapters

Integration adapters for order submission to Salesforce and master data sync from MCC/Oracle. Internal staging tables for replay and failure recovery.

Phase 07
Upcoming

Dashboards, Audit & Feedback Loop

Sales outcome dashboards, full audit store, campaign conversion analytics, order funnel reporting, and structured feedback workflow.

Phase 08
Future

Future Enhancements

Salesforce case management, colloquial voice tuning, live agent transfer, and automated scheme-balance follow-up engine.

POC → Product Gap

What the POC proves vs. what the product needs.

The POC demonstrates that voice AI can be connected to Berger dealer data and can use backend tools during conversations. The full product requires dedicated sales, campaign, scheme, order, and integration capabilities.

Area POC State Product Need
Business Objective Payment follow-up and dealer account queries. Sales engagement, scheme pitching, and order capture.
Conversation Modes FAQ and COLLECTION modes. SALES mode with campaign objective, scheme pitch, objection handling, order taking.
Campaigns No full campaign manager. Audience selection, scheduling, prioritization, call attempts, outcome tracking.
Schemes No dedicated scheme engine. Scheme metadata, eligibility rules, balances, qualification criteria, recommended pitch.
Orders No order capture workflow. Structured draft, confirmation loop, validation, Salesforce submission.
Product Catalog Not modeled for sales order taking. Product/SKU/pack size/unit lookup and constraint validation.
RAG FAQ-like support exists, sales knowledge not complete. Scheme explanation, objection handling, policy-approved answers.
Tonality Basic voice prompt behavior. Standard, assertive, and persuasive tonality based on dealer behavior.
Languages English, Hindi, Gujarati, Bengali covered in POC. MVP needs 9 languages: + Marathi, Tamil, Telugu, Kannada, Malayalam.
Integrations Internal DB and call providers. Salesforce and MCC/Oracle integration contracts with full audit.
Governance Message and call logging exists. Full audit: transcripts, tool calls, order extraction, campaign decisions, integration events.
💡

Key Message for Client

The POC has already validated the feasibility of AI-assisted dealer conversations over voice. The full Berger Sales Agent is a larger platform build because it must connect voice AI with campaign strategy, dealer intelligence, scheme rules, order capture, Salesforce workflows, MCC/Oracle data, and auditable governance.