AI Modeling & Build

After BRD completion, ekai generates production-ready artifacts. The AI MODELING & BUILD tab contains six sub-tabs for each artifact type.

Prerequisites

Complete the Capture Requirements step to generate your Business Requirements Document.

---

Artifact Sub-tabs

Tab	Content
DATA LINEAGE	Visual diagram of data flow
DATA CATALOG	Table and column documentation
BUSINESS GLOSSARY	Term definitions
METRICS/KPIS	Calculated measures with SQL
DATA VALIDATION	dbt tests and quality rules
DBT PROJECT	Complete dbt project code

---

Data Lineage

Visual representation of how data flows from source to output:

Lineage Diagram Elements

Source Tables

Yellow boxes - raw data from warehouse

Staging Models

Green boxes - cleaned/renamed data

Intermediate

Pink boxes - business logic

Final Models

White boxes - marts and facts

Controls

• Zoom — + / - buttons
• Pan — Click and drag
• Fullscreen — Expand view
• Download — Export as image

---

Data Catalog

Auto-generated documentation for all tables and columns:

Catalog Entry Example

fct_card_activation

Fact table capturing card activation events and metrics for onboarding analysis. Grain: One row per card issued.

Column Name	Data Type	Description
card_id	NUMBER	Primary key - Unique identifier from Worldline
customer_id	NUMBER	Foreign key to dim_customers
card_type	TEXT	Type of card (Debit, Credit, Prepaid)
card_issuance_date	DATE	When the card was issued
card_activation_date	DATE	When activated (nullable)
activation_lag_days	NUMBER	Days between issuance and activation
is_activated	BOOLEAN	Whether card has been activated
is_valid_activation	BOOLEAN	Data quality flag
customer_segment	TEXT	Customer segment for analysis

---

Business Glossary

Standardized term definitions:

Glossary Format

Term	Definition & Logic
Customer 360 View	A complete, unified view of each customer combining all banking relationships—deposits, loans, cards, and interactions—into a single record.
Active Relationship	A banking product or account that is currently open and operational. Closed or inactive accounts are not counted.
Deposit Balance	Total amount deposited across all active savings and checking accounts, shown in Euros.
Outstanding Loan Balance	Total amount owed across all active loans, shown in original loan currency.
Card Utilization Rate	Percentage of credit limit currently utilized across all active cards.

---

Metrics/KPIs

Calculated measures with SQL formulas:

Metrics Format

Name	Description	Calculation
Total Deposit Balance	Sum of all deposit balances (EUR)	SUM(total_deposit_balance_eur) WHERE deposit_account_count > 0
Average Deposit per Customer	Average deposit balance	AVG(total_deposit_balance_eur) WHERE deposit_account_count > 0
Card Utilization Rate	Credit usage percentage	(SUM(outstanding) / SUM(credit_limit)) * 100
Activation Rate	Cards activated percentage	COUNT(activated) / COUNT(total) * 100

---

Data Validation

Quality rules implemented as dbt tests:

Validation Rule Example

1. Grain Validation

Attribute	Value
Rule	One Row Per Customer
Category	GRAIN
Severity	CRITICAL
Test Type	Uniqueness
Expected Result	Zero rows returned
Failure Action	BLOCK deployment

Implementation:

SELECT customer_id, COUNT(*) as row_count
FROM mart_customer_360
GROUP BY customer_id
HAVING COUNT(*) > 1

---

DBT Project

Complete dbt project with all models:

Project Structure

├── README.md
├── dbt_project.yml
├── logs/
├── macros/
├── mart_relations.json
├── models/
│   ├── intermediate/
│   ├── marts/
│   │   └── customer/
│   │       ├── _customer__models.yml
│   │       ├── mart_customer_360.sql
│   │       └── sources.yml
│   └── staging/
├── models_targets.json
├── package-lock.yml
└── packages.yml

Code View

The right panel shows model SQL with syntax highlighting:

models:
  - name: mart_customer_360
    description: "Single denormalized table containing one row per customer"
    tests:
      - dbt_utils.equal_rowcount:
          compare_model: ref('stg_customers')
    columns:
      - name: CUSTOMER_ID
        description: "Unique identifier for the customer"
        tests:
          - not_null
          - unique

---

Share with Collaborators

Use the Share button to add team members:

---

Next Steps

• Execute DBT — Build and run the project
• Publish — Deploy to Snowflake Intelligence