Appendix 10-F: Deferred Items Register¶

Overview¶

This appendix documents items that have been intentionally deferred from the initial Phase 2B AI implementation documentation. Each item includes the rationale for deferral and the trigger/timeline for completion.

Deferral Principle: Items are deferred when they either: 1. Depend on outputs from implementation (cannot be created until system is live) 2. Require data collection over time before meaningful content can be produced 3. Are better suited for a specific implementation phase

Deferred Documentation Items¶

Appendix 10-A: Dialogflow CX Agent Export (JSON)¶

Attribute	Details
Item	Complete Dialogflow CX agent configuration export
Format	JSON blob (agent.json, intents/.json, flows/.json, etc.)
Reason for Deferral	Export can only be generated from a live GCP agent
Completion Trigger	After Dialogflow CX agent is built in GCP
Owner	AI Operations Team
Estimated Effort	30 minutes (export + documentation)

What Will Be Included: - agent.json - Agent settings and configuration - intents/*.json - All 10 intent definitions with training phrases - entityTypes/*.json - Custom entity definitions - flows/*.json - All 7 flow definitions - pages/*.json - All 27 page configurations - webhooks/*.json - Webhook configurations

Storage Location: Git repository (encrypted for sensitive data)

Appendix 10-B: Webhook Code (Python)¶

Attribute	Details
Item	Complete Python Flask webhook implementation
Format	Python file (.py) with full endpoint code
Reason for Deferral	Full implementation requires API specifications from Abhavtech backend team
Completion Trigger	During implementation phase when backend APIs are finalized
Owner	Development Team
Estimated Effort	4-8 hours

What Will Be Included:

abhavtech-fulfillment-webhook/
+-- main.py                 # Flask application entry point
+-- handlers/
|   +-- order_handler.py    # /orders endpoint
|   +-- account_handler.py  # /accounts endpoint
|   +-- billing_handler.py  # /billing endpoint
|   +-- product_handler.py  # /products endpoint
|   +-- support_handler.py  # /support endpoint
+-- utils/
|   +-- api_client.py       # Abhavtech API integration
|   +-- response_builder.py # Dialogflow response formatting
|   +-- auth.py             # Authentication utilities
+-- requirements.txt        # Python dependencies
+-- README.md               # Deployment instructions

Current Status: Code snippets provided in Chapter 10 Section 10.8

Appendix 10-C: Training Phrase Library¶

Attribute	Details
Item	Complete training phrase list for all intents
Format	CSV/Excel with phrases in English and Hindi
Reason for Deferral	Complete list requires SME review and Hindi translation verification
Completion Trigger	During implementation phase, before Dialogflow CX agent build
Owner	AI Operations + Business SMEs
Estimated Effort	8-12 hours

Summary from Chapter 10:

Intent	English Phrases	Hindi Phrases	Total
order.status	20	10	30
order.track	15	8	23
product.inquiry	20	10	30
product.pricing	15	8	23
account.balance	18	10	28
account.info	15	8	23
support.general	20	10	30
support.troubleshoot	25	10	35
billing.inquiry	20	10	30
agent.handoff	25	10	35
TOTAL	193	94	287

Deliverable Format:

training_phrases.csv:
intent_name,phrase,language,category
order.status,"Where is my order",en,direct_query
order.status,"मेरा ऑर्डर कहाँ है",hi,direct_query
...

Appendix 10-D: Test Case Matrix¶

Attribute	Details
Item	Complete test case matrix for AI features
Format	Excel workbook with multiple sheets
Reason for Deferral	Detailed test cases require finalized flow designs and UAT planning
Completion Trigger	During UAT planning phase
Owner	QA Team + AI Operations
Estimated Effort	16-24 hours

Summary from Chapter 10:

Category	Test Cases	Description
Intent Recognition	50	Verify correct intent detection for all phrases
Entity Extraction	30	Verify data capture (order numbers, products, etc.)
Conversation Flow	25	Multi-turn navigation and context retention
Escalation Paths	15	Handoff to agents with context preservation
Error Handling	20	Fallback, timeout, and recovery scenarios
Hindi Language	20	Hindi-specific interactions
TOTAL	160

Deliverable Structure:

AI_Test_Case_Matrix.xlsx:
+-- Sheet 1: Intent Recognition Tests
+-- Sheet 2: Entity Extraction Tests
+-- Sheet 3: Conversation Flow Tests
+-- Sheet 4: Escalation Tests
+-- Sheet 5: Error Handling Tests
+-- Sheet 6: Hindi Language Tests
+-- Sheet 7: Test Execution Tracker
+-- Sheet 8: Defect Log

Deferred Implementation Items¶

Predictive Routing¶

Attribute	Details
Feature	WxCC Predictive Routing with ML-based agent matching
Reason for Deferral	Requires minimum 6 months of historical interaction data
Completion Trigger	Phase 3 (Month 7+) after sufficient data collection
Owner	Contact Center Operations + IT
Prerequisites	Phase 2A + 2B operational, data collection enabled

Data Collection (Starting Phase 2A): - Call outcomes (wrap-up codes) - Handle time per agent - FCR by agent/queue - Customer intent (from VA) - Sentiment scores - CSAT survey responses

Expected Benefits (When Enabled): - FCR improvement: 82% -> 88% - CSAT improvement: 4.3 -> 4.5 - AHT reduction: 5.5 min -> 5.0 min

Custom Speech Model Training¶

Attribute	Details
Feature	Custom-trained speech recognition model for Abhavtech terminology
Reason for Deferral	Evaluate need based on baseline accuracy metrics
Completion Trigger	If intent recognition accuracy < 90% after 3 months
Owner	AI Operations + Google Cloud Support
Prerequisites	3 months of interaction data, accuracy metrics below threshold

Evaluation Criteria:

Metric	Threshold	Action
Intent Recognition	≥ 90%	No custom training needed
Intent Recognition	85-90%	Add training phrases, re-evaluate
Intent Recognition	< 85%	Initiate custom model training
Entity Extraction	≥ 95%	No changes needed
Entity Extraction	< 95%	Review entity patterns, add synonyms

If Custom Training Required: 1. Export conversation logs from Dialogflow CX 2. Prepare training dataset (minimum 10,000 utterances) 3. Engage Google Cloud professional services 4. Train and validate custom model 5. Deploy to production environment 6. Monitor accuracy improvement

Completion Tracking¶

Item	Target Date	Actual Date	Status	Notes
Appendix 10-A	Phase 2B Week 2	-	⏳ Pending	After agent build
Appendix 10-B	Phase 2B Week 1	-	⏳ Pending	During implementation
Appendix 10-C	Phase 2B Week 1	-	⏳ Pending	Before agent build
Appendix 10-D	Phase 2B Week 3	-	⏳ Pending	During UAT planning
Predictive Routing	Phase 3 (Month 7)	-	⏳ Deferred	Data collection ongoing
Custom Speech Model	TBD	-	⏳ Conditional	If accuracy < 90%

Document Update History¶

Date	Author	Changes
Jan 2026	Collaboration Architecture Team	Initial creation

Notes¶

Living Document: This appendix should be updated as deferred items are completed
Ownership: Each item has a designated owner responsible for completion
Dependencies: Some items depend on others (e.g., 10-C must be done before 10-A)
Quality Gate: Deferred items should go through the same review process as original documentation