Home / How ekai works

Your warehouse and your experts go in. Verified definitions come out.

First ekai discovers your data and draws the architecture plan. Then semantics and ontology: your experts define what it means, ekai proves it against your warehouse and publishes it to your agents.

Book a demo

SemanticsSales and revenue

Business question

What was net revenue last quarter?

ekai asks your expert

Does net revenue include tax and discounts?

Answered by Finance

No. Exclude both.

Generated togetherWaitingGeneratingDone

Business GlossaryAI Modeling

MetricsAI Modeling

LineageAI Modeling

dbt modelfct_orders.sql

Validation rules and reviewWaitingReviewingPass

Data Validation · Expected result: 0 violations

Semantic Model Review · ekai reviewing agent

Publish approval checkpointApproved at the checkpointSnowflake semantic views and Cortex Agents, Databricks Genie and a dbt project in your own Git$1.32Mnet revenue, last quarter

models/marts/fct_orders.sqldbt model

select
    order_id,
    customer_id,
    order_date,
    amount - tax - discount as net_revenue
from {{ ref('stg_orders') }}
where status in ('completed', 'settled_gw')  -- legacy gateway counts, per Finance

Business glossaryGlossary

Net revenue

Order amount net of tax and discounts.

Legacy gateway (settled_gw) orders count as real sales.

SUM(AMOUNT - TAX - DISCOUNT)From the Metrics artifact in AI Modeling

Owner: Finance

Illustration

ekai takes actual figures and creates verified, accurate answers.

AI should use your numbers, never calculate them.

Deterministic + probabilistic

Deterministic

Your warehouse calculates

net revenueamount − tax − discount$1.32MThe same on every run

Semantic layer
Each number defined by your expert and checked.
Ontology
How those definitions connect across your business.
AI agent delivery lifecycle
Agents built, tested and monitored on those definitions.

Probabilistic

AI uses the number

What was net revenue last quarter?$1.32MFrom Finance's definition

Discovery01 Connect 02 Model the structure 03 Architecture plan

Semantics and ontology04 Generate 05 Verify 06 Publish

Discovery · 01

Connect

ekai connects to your warehouse (Snowflake, Databricks, BigQuery, Redshift and more), links the tables you choose and profiles them where they sit.

What goes in: Read access to your warehouse, the tables you link and your context files
What comes out: A data profile of the linked tables
Where people step in: Your admin grants read access for Discovery

Discovery›Sales warehouseConnection Details

Snowflake5 tables linked · 2 context files · Modified 20 Oct 2026

View Connection Parameters

TablesLink the tables you choose

SALES.ORDERS
SALES.ORDER_ITEMS
CRM.CUSTOMERS
REF.REGIONS
REF.FISCAL_CALENDAR

Data Profile5 tables and 27 columns profiledView Profile

ekai product · demo data

Discovery · 02

Model the structure

The Onboarding Agent asks about your domain, like which schema is the source of truth, and writes the Source Onboarding Summary.

What goes in: The data profile and your context files
What comes out: The Onboarding Document, a Source Onboarding Summary
Where people step in: You answer the Onboarding Agent's questions

Discovery›Sales warehouseOnboarding

Onboarding DocumentSource Onboarding Summary

1Domain & Systems
2Business Processes
3Conceptual Model
4Conventions & Practice
5Transformation Plan
6Facts for Schema
7Unresolved Gaps

Onboarding Agent

Which schema is the source of truth for orders?

1SALES, loaded nightly from the order system2LEGACY, kept for history onlyAnswered 1 question

ekai product · demo data

Discovery · 03

Architecture plan

The Schema Agent maps keys and relationships across your tables and draws the entity relationship diagram that semantics builds on.

What goes in: The Onboarding Document
What comes out: Keys, relationships, documentation and an entity relationship diagram (DBML)
Where people step in: You review the relationships in the agent chat

Discovery›Sales warehouseSchema & Relationships

SALES.ORDERSORDER_ID7 columns

CRM.CUSTOMERSCUSTOMER_ID5 columns

SALES.ORDER_ITEMSORDER_ITEM_ID7 columns

REF.REGIONSREGION_CODE4 columns

REF.FISCAL_CALENDARFISCAL_DATE4 columns

REGION_CODECUSTOMER_IDREGION_CODEORDER_DATEORDER_ID

CUSTOMERSREGIONSREGION_CODE
ORDERSCUSTOMERSCUSTOMER_ID
ORDERSREGIONSREGION_CODE
ORDERSFISCAL_CALENDARORDER_DATE
ORDER_ITEMSORDERSORDER_ID

Schema AgentKeys · Relationships · Documentation · Entity relationship (DBML)

ekai product · demo data

Semantics and ontology · 04

Generate

The Business Requirements agent asks Finance whether net revenue includes tax and discounts, then the AI Modeling Agent builds the lineage and metric.

What goes in: The architecture plan and your expert's answers
What comes out: Business requirements, the semantic model and a dbt transformation
Where people step in: Your expert leads, and your engineer reviews the code

Semantics›Sales and revenueBusiness Requirements · AI Modeling

Business Requirements agentDoes net revenue include tax and discounts?Exclude both · answered by Finance

Business Glossary
Lineage
Sales and Revenue Semantic Model Review
Data Catalog
Metrics
Data Validation

SourceStagingMarts

ORDERSstg_ordersfct_net_revenue

CUSTOMERSstg_customersdim_customer

REGIONSstg_regionsdim_region

net_revenueMetrics · on fct_net_revenue

SUM(amount − tax − discount)

Illustration · demo data

Semantics and ontology · 05

Verify

Each definition becomes a validation rule with an expected result, and the read-only ekai reviewing agent gives a binding verdict.

What goes in: The semantic model and the dbt project
What comes out: Data validation rules, dbt test results and the verdict of an independent ekai reviewing agent
Where people step in: You approve at each checkpoint before anything publishes

Semantics›Sales and revenueAI Modeling · Data Validation

TypeRuleExpected result

GrainOne row per ORDER_ID0 violations

ReferentialEvery CUSTOMER_ID exists in dim_customer0 violations

ReferentialEvery REGION_CODE exists in dim_region0 violations

Measurenet_revenue = amount − tax − discount0 violations

Measurenet_revenue >= 0 on every order0 violations

Sales and Revenue Semantic Model ReviewThe ekai reviewing agent is read-only, and its verdict is binding. dbt tests run when the Data Transformation agent builds.

Pass

Illustration · demo data

Semantics and ontology · 06

Publish

The Data Transformation agent asks which target to publish to and waits for your approval, and Code Sync pushes the dbt project to your Git.

What goes in: The verified model
What comes out: Snowflake semantic views and Cortex Agents, Databricks Genie spaces and metric views, BigQuery data agents, or dbt MetricFlow for Postgres and Redshift, plus a dbt project in your Git
Where people step in: You approve the publish checkpoint, under your review process and branch protection

Semantics›Sales and revenueData Transformation · Code Sync

Data Transformation agent · tasks

Full-build approval checkpoint + dbt build
Validation audit (ekai reviewing agent)
Publish approval checkpoint + publish

Where should I publish Sales and revenue?

1Snowflake semantic view2Cortex agent3Databricks Genie4BigQuery Data Agent5dbt MetricFlow6Don't publish

Code Sync · Push to GitGitHub App · GitLab · BitBucket · GitHub PAT

Illustration · demo data · MCP, per model: coming soon

All of this comes together with governance and security at the heart of it.

Every definition becomes a validation rule, with a clear record of who declared it, and the dbt project is versioned in your Git.

DefinitionsOwnerExpected result

Net revenueamount − tax − discount

Finance0 violations

Active customercompleted order, last 90 days

Your expert0 violations

Net profit marginbetween 0 and 1, every row

Finance0 violations

Regionevery REGION_CODE in REF.REGIONS

Your engineer0 violations

Raw rows stay in your warehouseOwned, reviewed and versioned in your Git

Illustration · demo data

Read-only discovery

Discovery reads your tables in place. Semantics builds dbt models into its own ekai schema, separate from your source tables, and never edits them. Raw rows never leave your warehouse.

Deploy where it fits

Self-hosted, ekai SaaS or a Snowflake Native App, listed on the Snowflake Marketplace.

An owner on every definition

Every definition carries the name of the expert who owns it.

Versioned in your Git

Pushed to GitHub, GitLab or Bitbucket, under your review process and branch protection.

Activity and History

Every change is recorded: Activity on each semantic model, History on each item, and your own Git review process.

SOC 2 Type II

Audited with zero exceptions.

See architecture and security

You own every file ekai builds, in standard formats.

A standard dbt project in your own Git, plus the AI Modeling artifacts: business glossary, lineage, data catalog, metrics and data validation. Use ekai to build them and keep them for good.

AI Modeling › Metrics

net_revenue
  Description: Order amount net of tax and discounts.
  Formula: SUM(AMOUNT - TAX - DISCOUNT)
  Grain: one row per ORDER_ID
  Source: SALES.ORDERS

active_customers
  Description: Customers with a completed order in the last 90 days.
  Formula: COUNT(DISTINCT CASE WHEN is_active THEN CUSTOMER_ID END)
  Source: CRM.CUSTOMERS, SALES.ORDERS

Demo data, in the formats ekai generates.

Common questions

Last updated October 28, 2026

How can I build data products faster?

ekai turns a new business need into a trusted data product in hours, not months. It maps your data, learns the rules from your experts, and builds a dbt project you own.

What access does ekai need?

Read access for Discovery, which profiles the tables you link in place. To build semantic models, write access for its own ekai schema, separate from your source tables, where the Data Transformation agent creates its models. ekai never edits your source tables.

What leaves my environment?

Your raw rows never leave your warehouse. ekai works from metadata and statistical summaries, and the privacy policy sets out what can be shared with the AI providers you use.

Who approves what ekai builds?

Your people. Inferred relationships, definitions and metrics are reviewed, and the Data Transformation agent stops at approval checkpoints before it builds and before it publishes.

Where do the outputs live?

In your own Git, as a standard dbt project pushed from Code Sync to GitHub, GitLab or Bitbucket.

Your warehouse and your experts go in. Verified definitions come out.

Net revenue

ekai takes actual figures and creates verified, accurate answers.

Your warehouse calculates

Semantic layer

Ontology

AI agent delivery lifecycle

AI uses the number

Connect

Model the structure

Architecture plan

Generate

Verify

Publish

All of this comes together with governance and security at the heart of it.

Read-only discovery

Deploy where it fits

An owner on every definition

Versioned in your Git

Activity and History

SOC 2 Type II

You own every file ekai builds, in standard formats.

Common questions