GIGA CRM Phase 1 — PRD

AI Opportunity Platform for Construction Engineering

Status: Draft v0.1 Created: 2026-07-10 Charter Approved: 2026-06-23

GIGA CRM Phase 1 — Product Requirements Document (PRD)

Document Status: Draft v0.1 Created: 2026-07-10 Project Charter Approved: 2026-06-23 Document Owner: [TBD] Approved By: [TBD]

Revision History

Version	Date	Author	Changes
v0.1	2026-07-10	—	Initial framework, drafted from Project Charter

Project Overview
Business Background & Problem Statement
Project Objectives & Success Criteria
Scope Definition
Stakeholders & User Roles
Functional Requirements
Non-Functional Requirements
System Architecture Overview
Data Requirements
AI / Agent Requirements
Integration Requirements
Deliverable Mapping
Milestones & Timeline
Assumptions, Dependencies & Risks
Acceptance Criteria
Glossary
Appendix

1. Project Overview

1.1 Project Name

GIGA CRM — AI Opportunity Platform (Phase 1)

1.2 Project Positioning

An AI-first CRM platform for the construction engineering industry, leveraging AI Agents to automate opportunity acquisition, research, and scheduling workflows — connecting the full pipeline from opportunity discovery to project handoff.

1.3 Document Purpose

This document is based on the approved Project Charter (Approved 2026-06-23). It decomposes the Phase 1 scope defined in the charter into actionable functional and non-functional requirements, serving as the basis for subsequent solution design, development, and acceptance testing.

1.4 Target Audience

Product Managers & Business Analysts
Solution Architects & Development Teams
Testing Teams (UAT)
Project Sponsors & Business Owners

2. Business Background & Problem Statement

2.1 Industry Background

Opportunity management in the construction engineering industry relies heavily on external data sources (Dodge, ConstructConnect, etc.) and manual processes. The current business team faces the following core pain points:

2.2 Current Pain Points

ID	Pain Point	Impact
P-01	Disconnected systems, data silos	Opportunity information scattered across multiple systems, no unified view
P-02	Heavy manual processes	Researchers must manually screen opportunities, find contacts, and assess priorities
P-03	Inconsistent information	Same opportunity has different statuses across systems
P-04	Limited visibility	Management cannot view opportunity pipeline and team progress in real time
P-05	Inefficient handoff	Researcher-to-Scheduler handoff relies on email/verbal communication, high risk of information loss

2.3 Business Drivers

Increase conversion rate: AI-driven screening and prioritization lets the team focus on high-value opportunities
Reduce response time: Shorten the cycle from opportunity discovery to first outreach
Improve data quality: Unify data sources, eliminate duplication and inconsistency
Enhance traceability: Full-lifecycle status tracking, every action auditable

3. Project Objectives & Success Criteria

3.1 Phase 1 Objectives

Build an AI Opportunity Platform serving two core roles — Researchers and Schedulers — achieving end-to-end automation of opportunity acquisition, research, scheduling, and handoff.

3.2 Success Criteria

ID	Success Criterion	Measurement	Target
SC-01	Opportunity screening automation rate	AI-processed / Total opportunities	≥ 70%
SC-02	Research summary generation time	From import to summary output	≤ 5 minutes
SC-03	Scheduling recommendation adoption rate	Adopted / Total recommendations	≥ 60%
SC-04	Handoff information completeness rate	Complete packages / Total handoffs	≥ 95%
SC-05	System availability	Monthly uptime	≥ 99.5%
SC-06	Pilot user satisfaction	Pilot user survey score	≥ 4.0 / 5.0

3.3 Expected Business Benefits

Reduce researcher manual screening time by 50%+
Shorten opportunity-to-outreach cycle by 30%+
Eliminate duplicate data entry, achieve 95%+ data consistency

4. Scope Definition

4.1 Phase 1 Scope (In-Scope)

Scope Area	Includes
Opportunity Acquisition	Auto-import opportunity data from Dodge / ConstructConnect
Researcher Workflow	AI-assisted screening, enrichment, assessment, prioritization
Scheduler Workflow	AI scheduling recommendations, communication plan review, task assignment
AI Research Summaries	Auto-generate opportunity/project summaries for researchers
AI Scheduling Recommendations	Generate scheduling recommendations based on rules and models
Human Review Gates	Retain manual review at key nodes (Human-in-the-loop)
Agent Task Management	AI Agents autonomously execute tasks and track status
Status Tracking	Full-lifecycle opportunity status flow and visualization
Reporting Dashboard	Opportunity funnel, team performance, AI output quality visualization
Notifications & Reminders	Key event-triggered notifications (email/in-system)

4.2 Out of Phase 1 Scope (Out-of-Scope)

Excluded Item	Rationale	Future Phase
Mobile App	Phase 1 supports Web only	Phase 2
Customer Portal	External customer self-service platform	Phase 2
Multi-language Support	Phase 1 supports English only	Phase 2
Advanced BI Analytics	Phase 1 provides basic dashboards; deep analysis in later iterations	Phase 2
Third-party Data Source Expansion	Phase 1 integrates Dodge and ConstructConnect only	Phase 2
Automated Contracting	Contract/signing process not in this phase	Phase 3
Full Historical Data Migration	Only migrate active opportunity data for pilot regions	Phase 2

4.3 Scope Boundary Diagram

┌─────────────────────────────────────────────────────────┐
│                    GIGA CRM Phase 1                     │
│                                                         │
│  ┌──────────┐   ┌──────────┐   ┌──────────┐           │
│  │Opp. Acq. │──▶│Researcher│──▶│Scheduler │           │
│  │(AI Import)│   │Workflow  │   │Workflow  │           │
│  └──────────┘   └────┬─────┘   └────┬─────┘           │
│                      │              │                    │
│                      ▼              ▼                    │
│               ┌────────────┐  ┌──────────┐              │
│               │Human Review│  │Handoff Pkg│              │
│               │    Gate    │  │ Generated │              │
│               └────────────┘  └──────────┘              │
│                      │                                   │
│                      ▼                                   │
│               ┌────────────────────┐                    │
│               │Status Tracking +   │                    │
│               │    Dashboard       │                    │
│               └────────────────────┘                    │
│                                                         │
└─────────────────────────────────────────────────────────┘
     Data Sources: Dodge | ConstructConnect

5. Stakeholders & User Roles

5.1 Stakeholders

Role	Responsibilities	Key Concerns
Project Sponsor	Project authorization & resource approval	ROI, milestone achievement
Business Owner	Business requirement definition & priorities	Process improvement, user adoption
Product Manager	Requirement management & prioritization	Requirement completeness, delivery cadence
Solution Architect	Technical solution design	Architecture soundness, scalability
Development Team	Feature development & unit testing	Code quality, technical feasibility
AI/ML Engineer	AI Agent & model development	Model accuracy, inference performance
Testing Team	Test planning & execution	Defect rate, acceptance pass rate
Pilot Users	Trial usage & feedback	Usability, feature completeness

5.2 User Personas

Persona A: Researcher

┌─────────────────────────────────────┐
│           Researcher                 │
├─────────────────────────────────────┤
│ Responsibilities:                    │
│   - Review AI-screened opportunities │
│   - Enrich and assess opportunities  │
│   - Generate or review AI summaries  │
│   - Determine opportunity priority   │
│   - Hand off qualified opportunities │
│     to Scheduler                     │
│                                     │
│ Pain Points:                         │
│   - Manual screening is time-consuming│
│   - Information scattered, multi-system switching │
│   - No unified priority standards    │
│                                     │
│ Expectations:                        │
│   - AI pre-screening reduces 70% manual work │
│   - One-stop workspace               │
│   - AI summaries for quick overview  │
└─────────────────────────────────────┘

Persona B: Scheduler

┌─────────────────────────────────────┐
│           Scheduler                  │
├─────────────────────────────────────┤
│ Responsibilities:                    │
│   - Receive handed-off opportunities │
│   - Review communication plans       │
│   - Arrange follow-ups per AI recs   │
│   - Manage opportunity status flow   │
│   - Monitor pipeline health          │
│                                     │
│ Pain Points:                         │
│   - Incomplete handoff information   │
│   - Scheduling relies on experience, lacks data │
│   - Status tracking via spreadsheets, poor real-time │
│                                     │
│ Expectations:                        │
│   - Complete, structured handoff packages │
│   - Data-driven AI scheduling recs   │
│   - Real-time dashboard monitoring   │
└─────────────────────────────────────┘

Persona C: Manager

Attribute	Description
Responsibilities	Monitor team performance, approve key decisions, review summary reports
Pain Points	Lack of real-time visibility, scattered data
Expectations	Real-time dashboards, one-click report export

6. Functional Requirements

6.1 Module Overview

Module ID	Module Name	Priority	Related Deliverable
F-01	Platform Foundation & User Management	P0	D-01
F-02	Opportunity Data Import	P0	D-02
F-03	Opportunity Screening & Filtering	P0	D-03
F-04	Opportunity Enrichment & Assessment	P1	D-04
F-05	Communication Plan Review	P1	D-05
F-06	Human Review Gate	P0	D-06
F-07	Handoff Package Management	P1	D-07
F-08	Reporting & Visualization	P2	D-08
F-09	Pilot Enablement	P1	D-09
F-10	Status Tracking	P0	D-10
F-11	Notifications & Reminders	P2	D-11

6.2 F-01: Platform Foundation & User Management

Objective: Build platform infrastructure, provide user authentication, permission management, and basic configuration capabilities.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-01-01	Support user registration, login, password reset	P0	Users can complete registration-to-login flow independently
F-01-02	Support role-based access control (RBAC)	P0	Three roles: Researcher/Scheduler/Manager with mutually exclusive permissions
F-01-03	Support organizational structure management	P1	Users can be managed by team/department
F-01-04	Support system configuration management	P1	Admins can configure data source connections, AI parameters, etc.
F-01-05	Support operation audit logging	P1	All key operations are logged and queryable
F-01-06	Support session management & timeout	P0	Auto-logout after 30 minutes of inactivity

6.3 F-02: Opportunity Data Import

Objective: Automatically import opportunity data from external sources, with cleaning and deduplication support.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-02-01	Support scheduled Dodge data import	P0	Daily auto-pull, with manual trigger support
F-02-02	Support scheduled ConstructConnect data import	P0	Daily auto-pull, with manual trigger support
F-02-03	Data field mapping & transformation	P0	External fields auto-mapped to internal data model
F-02-04	Data deduplication logic	P0	Same project from multiple sources auto-identified and merged
F-02-05	Import exception handling & alerting	P0	Notify admin on import failure, support retry
F-02-06	Support manual CSV/Excel import	P1	Admins can upload files for supplementary import
F-02-07	Import history & traceability	P1	View records, counts, and status for each import

Data Import Flow:

Dodge API ──┐
             ├──▶ Field Mapping ──▶ Dedup & Merge ──▶ Data Cleaning ──▶ DB
ConstructConnect API ──┘                │
                                        ├── Error? ──▶ Alert Notification
                                        └── Success? ──▶ Log Record

6.4 F-03: Opportunity Screening & Filtering

Objective: AI auto-screens opportunities, Researcher reviews and confirms, reducing manual workload.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-03-01	AI screening rule configuration	P0	Admins can configure criteria (industry, region, scale, stage, etc.)
F-03-02	AI opportunity scoring	P0	Each opportunity has AI score (0-100) with scoring rationale
F-03-03	Opportunity list view	P0	List/card views, sortable, filterable, paginated
F-03-04	Batch operations	P1	Batch accept/reject/adjust priority
F-03-05	Screening results export	P2	Export to Excel/CSV
F-03-06	Screening rule versioning	P2	Rule changes traceable, with rollback support

Screening Flow:

Opp. Pool ──▶ AI Scoring ──▶ Auto-filter Low Scores ──▶ Researcher Review
                                        │
                                        ├── Accept ──▶ Enter Enrichment
                                        ├── Reject ──▶ Flag & Record Reason
                                        └── Adjust ──▶ Modify Priority & Accept

6.5 F-04: Opportunity Enrichment & Assessment

Objective: Enrich and deeply assess opportunities that pass screening.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-04-01	Auto information enrichment	P1	AI auto-supplements contacts, company info, history, etc.
F-04-02	AI research summary generation	P0	One-click summary (project overview, key contacts, competitive landscape)
F-04-03	Opportunity assessment checklist	P1	Structured checklist (project scale, timeline, decision-makers, etc.)
F-04-04	Priority tag management	P1	Support High/Medium/Low and custom tags
F-04-05	Assessment history	P1	Each assessment result traceable, comparable
F-04-06	AI summary manual editing	P0	Researchers can edit/correct AI-generated summaries

6.6 F-05: Communication Plan Review

Objective: AI generates communication plan recommendations; Scheduler reviews and confirms before execution.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-05-01	AI communication plan generation	P1	Generate recommended communication method, frequency, talking points
F-05-02	Communication plan template management	P1	Pre-built templates, support customization
F-05-03	Communication plan approval flow	P0	Scheduler review → Manager confirmation (high priority)
F-05-04	Communication record tracking	P1	Record each communication outcome, linked to opportunity
F-05-05	Communication effectiveness analysis	P2	Track reply rate, conversion rate, etc.

6.7 F-06: Human Review Gate (Human-in-the-Loop)

Objective: Retain manual review at key nodes in the AI automation flow to ensure quality and compliance.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-06-01	Review gate configuration	P0	Admins can configure which nodes require manual review
F-06-02	Review task assignment	P0	Review tasks auto-assigned to appropriate roles
F-06-03	Review interface	P0	Display AI decision context, support approve/reject/return
F-06-04	Review SLA management	P1	Set review time limits, auto-escalate on timeout
F-06-05	Review feedback loop	P1	Review results fed back to AI model for continuous optimization

Review Gate Nodes:

AI Screening ──▶ [Gate 1] Researcher confirms opportunity
                                        │
AI Enrichment ──▶ [Gate 2] Researcher confirms summary
                                        │
AI Comm. Plan ──▶ [Gate 3] Scheduler confirms plan
                                        │
AI Handoff Pkg ──▶ [Gate 4] Scheduler confirms receipt

6.8 F-07: Handoff Package Management

Objective: Structure handoff information from Researcher to Scheduler, ensuring completeness.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-07-01	Auto handoff package generation	P1	AI compiles opp info, summary, assessment into handoff package
F-07-02	Handoff package template	P0	Standardized template: project overview, key contacts, assessment conclusion, recommended actions
F-07-03	Handoff confirmation flow	P0	Send → Recipient confirms → Complete/Return
F-07-04	Handoff package versioning	P1	Multi-version support, new version on return-and-modify
F-07-05	Handoff statistics report	P2	Track handoff count, return rate, avg processing time

Handoff Package Structure:

┌─────────────────────────────────┐
│      Handoff Package             │
├─────────────────────────────────┤
│ 1. Project Basic Info            │
│    - Project name, address, type │
│    - Owner/GC information        │
│ 2. Opportunity Assessment Summary│
│    - AI summary + manual edits   │
│    - Score & priority            │
│ 3. Key Contacts                  │
│    - Name, role, contact info    │
│    - Decision chain analysis     │
│ 4. Recommended Actions           │
│    - AI recommendations +        │
│      Researcher additions        │
│ 5. Attachments                   │
│    - Related documents, images   │
└─────────────────────────────────┘

6.9 F-08: Reporting & Visualization

Objective: Provide visual dashboards for opportunity funnel, team performance, and AI output quality.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-08-01	Opportunity funnel dashboard	P1	Show opportunity count and conversion rate by stage
F-08-02	Team performance dashboard	P1	Display processing volume, efficiency metrics by user/team
F-08-03	AI output quality report	P1	AI scoring accuracy, adoption rate, summary quality score
F-08-04	Custom reports	P2	Generate custom reports with selectable dimensions/metrics
F-08-05	Report export	P2	Export to PDF/Excel
F-08-06	Scheduled report emails	P2	Auto-send daily/weekly reports

6.10 F-09: Pilot Enablement

Objective: Provide training, documentation, and support for pilot users to ensure smooth transition.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-09-01	User operation manual	P1	Cover operation guides for all functional modules
F-09-02	Online training courses	P1	Role-based training videos and exercises
F-09-03	In-system guided tips	P2	Guided tutorial for first-time users
F-09-04	Pilot support channel	P0	Provide ticket/instant messaging support
F-09-05	Pilot feedback collection	P1	In-system feedback entry, regular user opinion collection

6.11 F-10: Status Tracking

Objective: Full-lifecycle opportunity status flow and visualization, with audit traceability.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-10-01	Opportunity state machine definition	P0	Clearly define states and transition rules
F-10-02	Auto status transition	P0	AI Agent advances status based on workflow
F-10-03	Manual status intervention	P0	Users can manually adjust status (requires permission)
F-10-04	Status change history	P0	Complete record of status changes (who, when, why)
F-10-05	Status board view	P1	Kanban-style display of opportunities by status
F-10-06	Timeout alerts	P1	Alert when opportunity stays in a status beyond threshold

Opportunity State Machine:

┌──────────┐     ┌──────────┐     ┌──────────┐     ┌──────────┐
│ Imported │────▶│AI Screen │────▶│Pending   │───▶│Enriching │
│          │     │          │     │Review    │    │          │
└──────────┘     └──────────┘     └──────────┘    └────┬─────┘
                                                     │
┌──────────┐     ┌──────────┐     ┌──────────┐       │
│  Closed  │◀────│Completed │◀────│Scheduling│◀──────┘
│          │     │          │     │          │
└──────────┘     └──────────┘     └──────────┘
     ▲                                    │
     │          ┌──────────┐              │
     └──────────│ Rejected │◀─────────────┘
                │          │
                └──────────┘

6.12 F-11: Notifications & Reminders

Objective: Key event-triggered notifications to ensure timely response from relevant personnel.

Req ID	Requirement Description	Priority	Acceptance Criteria
F-11-01	Review task notification	P0	Notify assigned personnel when new review task is created
F-11-02	Handoff notification	P0	Notify both parties on handoff package send/receive
F-11-03	Timeout reminder	P1	Remind 1 hour before task timeout
F-11-04	Import exception notification	P0	Notify admin on data import failure
F-11-05	Daily summary email	P2	Send daily work summary to users
F-11-06	Notification preferences	P2	Users can configure receiving method and frequency

Notification Channels: - In-system messages - Email notifications - (Future phase) Feishu/WeCom push

7. Non-Functional Requirements

7.1 Performance Requirements

ID	Requirement	Metric
NFR-P-01	Page load time	First screen ≤ 3s (P95)
NFR-P-02	API response time	Standard API ≤ 500ms (P95), complex query ≤ 2s (P95)
NFR-P-03	AI summary generation time	≤ 30s (P95)
NFR-P-04	Data import throughput	≥ 5,000 opportunities per import
NFR-P-05	Concurrent user support	≥ 100 concurrent users

7.2 Availability Requirements

ID	Requirement	Metric
NFR-A-01	System availability	≥ 99.5% (monthly)
NFR-A-02	Planned maintenance window	Once per month, 48-hour advance notice
NFR-A-03	Recovery Time Objective (RTO)	≤ 4 hours
NFR-A-04	Recovery Point Objective (RPO)	≤ 1 hour

7.3 Security Requirements

ID	Requirement	Description
NFR-S-01	Data transmission encryption	Site-wide HTTPS / TLS 1.2+
NFR-S-02	Data storage encryption	Sensitive fields encrypted at rest
NFR-S-03	Authentication mechanism	Support SSO (OAuth 2.0 / SAML)
NFR-S-04	Least privilege	Role-based fine-grained access control
NFR-S-05	Audit logs	Retained ≥ 1 year, tamper-proof
NFR-S-06	Data backup	Daily full + real-time incremental backup

7.4 Scalability Requirements

ID	Requirement	Description
NFR-E-01	Horizontal scaling	Core services support horizontal scaling, stateless design
NFR-E-02	Data source extensibility	Data ingestion layer supports plug-in style new data sources
NFR-E-03	AI model extensibility	AI inference layer supports hot model replacement
NFR-E-04	Multi-tenancy readiness	Architecture pre-reserves multi-tenancy capability (Phase 2)

7.5 Compatibility Requirements

ID	Requirement	Description
NFR-C-01	Browser compatibility	Chrome 100+, Edge 100+, Safari 15+
NFR-C-02	Resolution support	Minimum 1280×720, recommended 1920×1080
NFR-C-03	Responsive design	Support tablet landscape access (basic adaptation)

8. System Architecture Overview

8.1 Architecture Overview

┌─────────────────────────────────────────────────────────────┐
│                    GIGA CRM Architecture                     │
│                                                             │
│  ┌─────────────────────────────────────────────────────┐   │
│  │           AI Agent Runtime (Sandboxed)               │   │
│  │  ┌─────────┐  ┌─────────┐  ┌─────────┐            │   │
│  │  │Screen   │  │Summary  │  │Schedule │            │   │
│  │  │Agent    │  │Agent    │  │Agent    │            │   │
│  │  └─────────┘  └─────────┘  └─────────┘            │   │
│  └───────────────────────┬─────────────────────────────┘   │
│                           │                                 │
│  ┌───────────────────────▼─────────────────────────────┐   │
│  │         CRM Core Operational Database                │   │
│  │  Opp. DB | Contacts | Status | Audit | Config       │   │
│  └───────────────────────┬─────────────────────────────┘   │
│                           │                                 │
│  ┌───────────────────────▼─────────────────────────────┐   │
│  │            Operational Modules                       │   │
│  │  ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐     │   │
│  │  │Import│ │Screen│ │Enrich│ │Handoff│ │Report│     │   │
│  │  └──────┘ └──────┘ └──────┘ └──────┘ └──────┘     │   │
│  └───────────────────────┬─────────────────────────────┘   │
│                           │                                 │
│  ┌───────────────────────▼─────────────────────────────┐   │
│  │                  Data Layer                          │   │
│  │  Dodge API | CC API | File Storage | Cache           │   │
│  └─────────────────────────────────────────────────────┘   │
└─────────────────────────────────────────────────────────────┘

8.2 Architecture Layer Description

Layer	Name	Responsibility	Key Technology (Suggested)
L1	AI Agent Runtime	AI Agent execution environment, sandboxed	Python / LangChain / Containerized
L2	CRM Core Database	Core business data storage	PostgreSQL / Redis (cache)
L3	Operational Modules	Business function modules	Microservices / Modular Monolith
L4	Data Layer	External data integration & file storage	API Gateway / S3-compatible storage

8.3 Key Architecture Decisions (Pending Solution Design)

Decision Point	Option A	Option B	Recommendation
Deployment Architecture	Modular Monolith	Microservices	Phase 1: Modular Monolith to reduce complexity
AI Inference	Cloud API	On-premise	Hybrid: Cloud API for summaries, local model for screening
Frontend Framework	React	Vue	Based on team tech stack
Message Queue	RabbitMQ	Kafka	Phase 1: RabbitMQ, sufficient for needs

9. Data Requirements

9.1 Core Data Entities

Entity	Description	Key Fields	Source
Opportunity	Core business object	ID, project name, address, type, stage, score, status	Dodge / CC
Contact	Opportunity-linked contact	ID, name, role, phone, email, company	Enrichment / Manual
Company	Owner/GC company	ID, name, address, type, industry	Enrichment / Manual
Assessment	Opportunity assessment record	ID, opp ID, assessor, score, summary, date	System generated
HandoffPackage	Handoff information package	ID, opp ID, sender, receiver, content, status	System generated
AgentTask	AI Agent executed task	ID, type, input, output, status, timestamp	System generated
AuditLog	Operation audit	ID, user, action, object, time, details	System recorded

9.2 Data Retention Policy

Data Type	Retention Period	Archive Strategy
Active opportunity data	Permanent	—
Closed opportunities	3 years	Archive to cold storage after 3 years
Audit logs	1 year	Compress and archive after 1 year
Agent task records	6 months	Aggregate stats then delete details after 6 months
AI summary history	2 years	Retain only final version after 2 years

10. AI / Agent Requirements

10.1 AI Agent Definitions

Agent Name	Responsibility	Input	Output	Trigger
Screening Agent	Opportunity scoring & filtering	Raw opportunity data	Score + screening result	Scheduled / Event
Summary Agent	Generate research summaries	Opportunity + enrichment info	Structured summary	User request
Scheduling Agent	Generate scheduling recommendations	Opportunity + history + rules	Scheduling recommendation	Status change
Enrichment Agent	Information supplementation	Basic opportunity info	Supplemented contacts/company info	Event triggered

10.2 Agent Operation Modes

┌─────────────────────────────────────────┐
│         Agent-assisted Mode              │
│  AI provides recommendations/drafts →    │
│  Human reviews and confirms              │
│  Applies to: Screening, Summary,         │
│  Communication Plan                      │
├─────────────────────────────────────────┤
│         Agent-operated Mode              │
│  AI executes autonomously →              │
│  Notifies human of results               │
│  Applies to: Data Import, Enrichment,    │
│  Status Transitions                      │
├─────────────────────────────────────────┤
│         Human-gated Mode                 │
│  AI prepares →                           │
│  Must be approved by human to proceed    │
│  Applies to: Handoff Package Send,       │
│  High-priority Opportunity Processing    │
└─────────────────────────────────────────┘

10.3 AI Quality Requirements

Metric	Target	Measurement
Screening accuracy	≥ 80%	Manual sampling verification
Summary completeness	≥ 90%	Key information coverage rate
Scheduling recommendation reasonableness	≥ 60% adoption rate	User adoption ratio
Enrichment information accuracy	≥ 85%	Manual sampling verification
AI response time	≤ 30s	P95 latency

10.4 AI Feedback Loop

AI Output ──▶ Human Review ──▶ Review Result
                                │
                         ┌──────┴──────┐
                         │ Accept       │ Reject/Modify
                         │              │
                         ▼              ▼
                   Record Positive  Record Negative + Correction
                         │              │
                         └──────┬───────┘
                                │
                         ▼      ▼
                   Periodic Model Retraining / Prompt Optimization

11. Integration Requirements

11.1 External System Integrations

System	Integration Method	Data Flow	Frequency	Priority
Dodge	REST API	Pull opportunity data	Daily	P0
ConstructConnect	REST API	Pull opportunity data	Daily	P0
Email System (SMTP)	SMTP	Send notifications	Event-triggered	P0
SSO (OAuth 2.0)	OAuth	User authentication	Real-time	P0
File Storage (S3)	SDK	Store attachments	Real-time	P1

11.2 Integration Interface Specifications

Dodge API Integration
├── Endpoint: [TBD]
├── Authentication: API Key
├── Frequency: Daily 02:00 UTC
├── Data Format: JSON
├── Error Handling: Retry 3 times, 5-min interval
└── Alerting: Notify admin after 2 consecutive failures

ConstructConnect API Integration
├── Endpoint: [TBD]
├── Authentication: API Key
├── Frequency: Daily 03:00 UTC
├── Data Format: JSON
├── Error Handling: Retry 3 times, 5-min interval
└── Alerting: Notify admin after 2 consecutive failures

12. Deliverable Mapping

12.1 Charter Deliverable to PRD Requirement Mapping

Deliverable ID	Deliverable Name	Corresponding Module	Related Requirements	Acceptance Status
D-01	Platform Foundation	F-01	F-01-01 ~ F-01-06	Pending
D-02	Data Import	F-02	F-02-01 ~ F-02-07	Pending
D-03	Opportunity Screening	F-03	F-03-01 ~ F-03-06	Pending
D-04	Enrichment & Assessment	F-04	F-04-01 ~ F-04-06	Pending
D-05	Comm. Plan Review	F-05	F-05-01 ~ F-05-05	Pending
D-06	Human Review Gate	F-06	F-06-01 ~ F-06-05	Pending
D-07	Handoff Package	F-07	F-07-01 ~ F-07-05	Pending
D-08	Reporting & Visualization	F-08	F-08-01 ~ F-08-06	Pending
D-09	Pilot Enablement	F-09	F-09-01 ~ F-09-05	Pending
D-10	Status Tracking	F-10	F-10-01 ~ F-10-06	Pending
D-11	Notifications & Reminders	F-11	F-11-01 ~ F-11-06	Pending

13. Milestones & Timeline

13.1 Milestone Overview

Milestone	Target Date	Key Deliverables	Status
M1: Solution Design	June 2026	Technical solution design document	✅ Completed
M2: Core Platform	July 2026	F-01, F-02, F-10 foundation features	🔄 In Progress
M3: AI Workflows	August 2026	F-03, F-04, F-05, F-06, F-07	⏳ Pending
M4: UAT	September 2026	Full-function testing + defect resolution	⏳ Pending
M5: Pilot Go-Live	September 2026	Pilot user launch	⏳ Pending
M6: Stabilized	October 2026	System stable operation	⏳ Pending
M7: Closure	October 2026	Project closure and handover	⏳ Pending

13.2 Phase 1 Feature-to-Milestone Mapping

July (M2)        August (M3)       September (M4-M5)  October (M6-M7)
─────────────────────────────────────────────────────────────────────
F-01 Platform    ████████
F-02 Data Import ████████
F-10 Status Trk  ████████
F-03 Screening             ████████
F-04 Enrichment            ████████
F-05 Comm. Plan            ████████
F-06 Review Gate           ████████
F-07 Handoff Pkg           ████████
F-08 Reporting                       ████████
F-09 Pilot Enable                   ████████
F-11 Notifications                   ████████
                                     UAT Testing  Pilot Live  Stabilize&Close

14. Assumptions, Dependencies & Risks

14.1 Assumptions

ID	Assumption	Impact
A-01	Dodge and ConstructConnect APIs remain stable and available during the project	Data import dependency
A-02	Pilot users are willing to participate in testing and provide feedback	UAT and pilot success
A-03	AI model accuracy meets minimum business requirements (≥ 80%)	System usability
A-04	Project team is staffed, no key role vacancies	On-time delivery

14.2 Dependencies

ID	Dependency	Dependent Party	Impact	Mitigation
D-01	Dodge API access & documentation	Data provider	Blocks data import development	Request access early, confirm API docs
D-02	ConstructConnect API access	Data provider	Same as above	Same as above
D-03	SSO system integration	IT Infrastructure team	Blocks auth development	Coordinate integration timing early
D-04	AI model training data	Business team	Affects AI accuracy	Collect historical labeled data
D-05	Pilot user list confirmation	Business owner	Affects UAT scheduling	Confirm by end of August

14.3 Risk Register

ID	Risk	Probability	Impact	Risk Level	Mitigation Strategy
R-01	Dodge/CC API changes or unavailability	Medium	High	🔴 High	Design data source abstraction layer; manual import fallback
R-02	AI model accuracy below target	Medium	High	🔴 High	Reserve model tuning time; rules engine fallback
R-03	Low pilot user adoption	Medium	Medium	🟡 Medium	Early training; simplify workflows; rapid feedback iteration
R-04	Project staff changes	Low	High	🟡 Medium	Designate backups for key roles; document-driven management
R-05	Scope creep	Medium	Medium	🟡 Medium	Strict change management; Phase 2 backlog
R-06	Data migration quality issues	Low	Medium	🟢 Low	Pre-migration data cleansing; batch migration verification

15. Acceptance Criteria

15.1 Functional Acceptance Criteria

Acceptance criteria for each functional requirement are defined in Section 6. Overall functional acceptance must meet:

All P0 requirements: 100% implemented and tested
All P1 requirements: ≥ 90% implemented and tested
P2 requirements: delivered as feasible, do not block go-live

15.2 UAT Acceptance Criteria

Acceptance Item	Standard
Test case coverage	≥ 90% of functional requirements have test cases
Test pass rate	≥ 95%
Critical defects	0 P0 defects
Major defects	≤ 5 P1 defects (with fix plan)
Pilot user acceptance	Pilot users sign off

15.3 Go-Live Checklist

[ ] All P0 features pass UAT
[ ] Performance metrics met (Section 7.1)
[ ] Security review passed
[ ] Data backup & recovery process verified
[ ] Ops monitoring & alerting configured
[ ] User manual and training materials ready
[ ] Pilot user training completed
[ ] Emergency rollback plan ready

16. Glossary

Term	Abbreviation	Definition
Opportunity	Opp	Potential construction project business opportunity
Researcher	—	Role responsible for screening, enriching, and assessing opportunities
Scheduler	—	Role responsible for receiving handed-off opportunities and arranging follow-ups
Enrichment	—	Process of supplementing and completing opportunity information
Handoff Package	—	Structured information package from Researcher to Scheduler
Human Review Gate	—	Key manual review node in AI workflow
Agent-assisted	—	AI-assisted mode: AI provides recommendations, human confirms
Agent-operated	—	AI autonomous mode: AI executes, then notifies human
Human-gated	—	Human-gated mode: AI prepares, human must approve to proceed
Dodge	—	Construction project information data service provider
ConstructConnect	CC	Construction project information data service provider
RBAC	Role-Based Access Control	Role-based access control
RTO	Recovery Time Objective	Target time for system recovery after failure
RPO	Recovery Point Objective	Target data recovery point after failure
UAT	User Acceptance Testing	User acceptance testing
WBS	Work Breakdown Structure	Work breakdown structure

17. Appendix

Appendix A: Charter Key Points Summary

Core points from the Project Charter (Approved 2026-06-23), for PRD alignment reference.

Project Name: GIGA CRM
Assessment/Approval Date: 2026/06/23
Phase 1 Objective: Build AI Opportunity Platform
Core Users: Researchers, Schedulers
Data Sources: Dodge, ConstructConnect
Architecture: AI-first CRM
Deliverables: 11 items (D-01 ~ D-11)

Appendix B: Open Items

ID	Open Item	Owner	Due Date	Status
O-01	Dodge API endpoint & authentication method	—	—	Pending
O-02	ConstructConnect API endpoint & authentication method	—	—	Pending
O-03	SSO system type & protocol	—	—	Pending
O-04	AI model selection (cloud vs. on-premise)	—	—	Pending
O-05	Deployment environment (cloud / on-premise)	—	—	Pending
O-06	Pilot user list & scale	—	—	Pending
O-07	Frontend tech stack selection	—	—	Pending

Appendix C: Requirements Traceability Matrix Template

Req ID	Requirement Description	Source (Charter Clause)	Priority	Status	Design Doc	Dev Task	Test Case
F-01-01	User registration & login	Charter-Deliverable D-01	P0	Draft	TBD	TBD	TBD
F-02-01	Dodge data import	Charter-Deliverable D-02	P0	Draft	TBD	TBD	TBD
...	...	...	...	...	...	...	...

End of Document

This PRD is Draft v0.1 framework version. It requires the following process before finalization: 1. Business review — Confirm scope and priorities 2. Architect review — Confirm technical feasibility 3. Development team review — Confirm implementation approach 4. Project manager approval — Confirm schedule and resources 5. Final release — v1.0 Baseline

Generated from GIGA_CRM_PRD_Framework_v0.1_EN.md — Deployed on Cloudflare Pages