Dremio Blog

17 minute read · February 9, 2026

From Bottlenecks to Breakthroughs: A Hands-on Intro to Agentic Analytics for the Data Analyst

Alex Merced Alex Merced Head of DevRel, Dremio
Start For Free
From Bottlenecks to Breakthroughs: A Hands-on Intro to Agentic Analytics for the Data Analyst
Copied to clipboard

Table of Contents

From Bottlenecks to Breakthroughs: A Hands-on Intro to Agentic Analytics for the Data Analyst

Key Takeaways
What is the Agentic Lakehouse? (Beyond the Hype)
Takeaway 1: Your Foundation is Ready (The Free Trial Walkthrough)
Takeaway 2: Organizing the Chaos (Namespaces and Tables)
Takeaway 3: The Agent as Your Analytical Co-pilot
SQL Generation for Medallion Architecture
Self-Documenting Metadata
Takeaway 4: Instant Visualization from Gold Datasets
Takeaway 5: SQL-Powered AI (Analyzing Silver Data)
Sentiment Analysis in SQL
Unlocking "Dark Data"
Conclusion: The Future of the Self-Managing Lakehouse

Key Takeaways

  • Data analysts face challenges like architectural friction, which slows down analysis and decision-making.
  • Dremio's Agentic Lakehouse uses advanced technologies for fast, AI-driven insights from fragmented data sources.
  • Key features include Query Federation, Autonomous Reflections, and an AI Semantic Layer, enhancing operational efficiency.
  • The platform offers a free trial with instant access, allowing users to build projects without lengthy approvals.
  • Dremio empowers users to analyze and visualize data quickly and intuitively, turning complex queries into actionable insights.

For many data analysts, daily reality is defined by "architectural friction." You have a critical business question, but the answer is buried under fragmented data silos, brittle ETL pipelines, and queries that take forever to run. This friction turns promising data lakes into inefficient "data swamps," where the cycle time between a question and an answer is measured in days. In this environment, learning doesn't compound; it stalls.

Enter Agentic Analytics. This isn't just another chatbot; it’s a paradigm shift where the data platform itself acts as an autonomous partner. Imagine asking a question in plain English, "Which suppliers are driving our lowest on-time-in-full (OTIF) rates?", and receiving an accurate, visualized answer in seconds. Dremio’s Agentic Lakehouse makes this possible by providing the three missing ingredients for AI-driven analysis: deep business context, universal access, and interactive speed.

Try Dremio’s Interactive Demo

Explore this interactive demo and see how Dremio's Intelligent Lakehouse enables Agentic AI

What is the Agentic Lakehouse? (Beyond the Hype)

The Agentic Lakehouse is more than a repository; it is the "brain" of your data operation. To a Lakehouse Architect, this means moving beyond passive metadata to a platform that actively manages itself. Dremio achieves this by sitting atop an open foundation of Apache Iceberg and Apache Polaris.

To deliver "interactive speed," Dremio relies on a high-performance Massively Parallel Processing (MPP) architecture powered by Apache Arrow, the open-source columnar format that eliminates costly serialization overhead. When you query data in the lake, Dremio’s C3 (Columnar Cloud Cache) ensures that frequently accessed data stays close to compute, delivering the sub-second response times usually reserved for expensive proprietary warehouses.

Dremio eliminates the "analytics bottleneck" through three core pillars:

  • Query Federation: Access data directly where it lives, S3, Snowflake, or PostgreSQL, without the risk and cost of data movement.
  • Autonomous Reflections: Unlike traditional materialized views that require manual tuning, Dremio’s engine learns from query patterns and automatically creates optimized physical layouts to accelerate performance behind the scenes.
  • AI Semantic Layer: This is where you "teach" the AI your business logic. By using a Layered View Strategy, transitioning from a Preparation Layer (raw data) to a Business Layer (joins and logic) and finally an Application Layer (tailored for specific users), you provide the structured context an AI Agent needs to be accurate.

"Discovery, exploration, and analysis that could previously take hours can now be done in minutes with Dremio's AI Agent."

Takeaway 1: Your Foundation is Ready (The Free Trial Walkthrough)

The greatest hurdle for analysts is often waiting for IT to approve infrastructure. Dremio’s "Next Gen" cloud trial provides an "instant-on" experience that bypasses this administrative drag.

Step-by-Step Setup:

  1. Sign Up: Navigate to the Dremio sign-up page and authenticate via Google, Microsoft, GitHub, or email.
  2. Automatic Provisioning: Dremio immediately creates an Organization and your first Project.
  3. Managed Storage: By default, the trial includes managed storage. You can upload CSVs or Parquet files and start querying immediately without connecting a credit card or an S3 bucket (though Dremio supports S3 for custom catalog storage when you're ready to scale).

Architect's Insight: This low-friction entry allows you to move from "curious" to "querying" in minutes. It’s a sandbox where you can build proof-of-concepts without the typical "capacity planning guessing game."

Takeaway 2: Organizing the Chaos (Namespaces and Tables)

Open up the SQL editor where you can run the following SQL to seed some namespaces and tables in your Dremio Catalog.

-------------------------------------------------------------------------------

-- 1. SETUP: Create Raw Folder

-------------------------------------------------------------------------------

CREATE FOLDER IF NOT EXISTS Zendesk_Clone;

CREATE FOLDER IF NOT EXISTS Zendesk_Clone.Support;

-------------------------------------------------------------------------------

-- 2. DDL: Create Raw Tables (3 Tables)

-------------------------------------------------------------------------------

CREATE TABLE IF NOT EXISTS Zendesk_Clone.Support.AGENTS (

    AGENT_ID VARCHAR,

    NAME VARCHAR,

    LEVEL VARCHAR -- 'L1', 'L2'

);

CREATE TABLE IF NOT EXISTS Zendesk_Clone.Support.CUSTOMERS (

    CUST_ID VARCHAR,

    EMAIL VARCHAR,

    TIER VARCHAR -- 'Free', 'VIP'

);

CREATE TABLE IF NOT EXISTS Zendesk_Clone.Support.TICKETS (

    TICKET_ID VARCHAR,

    CUST_ID VARCHAR,

    AGENT_ID VARCHAR,

    CREATED_TS TIMESTAMP,

    RESOLVED_TS TIMESTAMP,

    DESCRIPTION VARCHAR

);

-------------------------------------------------------------------------------

-- 3. SEED DATA (50+ Records)

-------------------------------------------------------------------------------

-- AGENTS

INSERT INTO Zendesk_Clone.Support.AGENTS VALUES

('A-1', 'Alice', 'L1'), ('A-2', 'Bob', 'L2');

-- CUSTOMERS

INSERT INTO Zendesk_Clone.Support.CUSTOMERS VALUES

('C-1', '[email protected]', 'Free'), ('C-2', '[email protected]', 'VIP');

-- TICKETS (50 Rows)

INSERT INTO Zendesk_Clone.Support.TICKETS VALUES

('T-1', 'C-1', 'A-1', '2023-01-01 10:00:00', '2023-01-01 11:00:00', 'Login issue'),

('T-2', 'C-1', 'A-1', '2023-01-01 12:00:00', '2023-01-01 13:00:00', 'Reset password'),

('T-3', 'C-1', 'A-1', '2023-01-01 14:00:00', '2023-01-01 15:00:00', 'Billing'),

('T-4', 'C-1', 'A-1', '2023-01-01 16:00:00', '2023-01-01 17:00:00', 'URGENT outage'),

('T-5', 'C-2', 'A-2', '2023-01-01 10:00:00', '2023-01-01 10:05:00', 'VIP Request'),

('T-6', 'C-2', 'A-2', '2023-01-01 11:00:00', '2023-01-01 11:05:00', 'Help'),

('T-7', 'C-1', 'A-1', '2023-01-02 10:00:00', '2023-01-02 11:00:00', 'Bug'),

('T-8', 'C-1', 'A-1', '2023-01-02 12:00:00', '2023-01-02 13:00:00', 'Bug'),

('T-9', 'C-1', 'A-1', '2023-01-02 14:00:00', '2023-01-02 15:00:00', 'Bug'),

('T-10', 'C-1', 'A-1', '2023-01-02 16:00:00', '2023-01-02 17:00:00', 'Bug'),

('T-11', 'C-1', NULL, '2023-01-03 10:00:00', NULL, 'Unassigned'), -- Open

('T-12', 'C-1', 'A-1', '2023-01-03 10:00:00', '2023-01-02 10:00:00', 'Time Travel'), -- Error

('T-13', 'C-1', 'A-1', '2023-01-03 12:00:00', '2023-01-03 13:00:00', 'Issue'),

('T-14', 'C-1', 'A-1', '2023-01-03 13:00:00', '2023-01-03 14:00:00', 'Issue'),

('T-15', 'C-1', 'A-1', '2023-01-04 10:00:00', '2023-01-04 11:00:00', 'Issue'),

('T-16', 'C-1', 'A-1', '2023-01-04 11:00:00', '2023-01-04 12:00:00', 'Issue'),

('T-17', 'C-1', 'A-1', '2023-01-04 12:00:00', '2023-01-04 13:00:00', 'Issue'),

('T-18', 'C-1', 'A-1', '2023-01-04 13:00:00', '2023-01-04 14:00:00', 'Issue'),

('T-19', 'C-1', 'A-1', '2023-01-04 14:00:00', '2023-01-04 15:00:00', 'Issue'),

('T-20', 'C-1', 'A-1', '2023-01-04 15:00:00', '2023-01-04 16:00:00', 'Issue'),

('T-21', 'C-1', 'A-1', '2023-01-05 10:00:00', '2023-01-05 11:00:00', 'Issue'),

('T-22', 'C-1', 'A-1', '2023-01-05 11:00:00', '2023-01-05 12:00:00', 'Issue'),

('T-23', 'C-1', 'A-1', '2023-01-05 12:00:00', '2023-01-05 13:00:00', 'Issue'),

('T-24', 'C-1', 'A-1', '2023-01-05 13:00:00', '2023-01-05 14:00:00', 'Issue'),

('T-25', 'C-1', 'A-1', '2023-01-05 14:00:00', '2023-01-05 15:00:00', 'Issue'),

('T-26', 'C-1', 'A-1', '2023-01-05 15:00:00', '2023-01-05 16:00:00', 'Issue'),

('T-27', 'C-1', 'A-1', '2023-01-06 10:00:00', '2023-01-06 11:00:00', 'Issue'),

('T-28', 'C-1', 'A-1', '2023-01-06 11:00:00', '2023-01-06 12:00:00', 'Issue'),

('T-29', 'C-1', 'A-1', '2023-01-06 12:00:00', '2023-01-06 13:00:00', 'Issue'),

('T-30', 'C-1', 'A-1', '2023-01-06 13:00:00', '2023-01-06 14:00:00', 'Issue'),

('T-31', 'C-1', 'A-1', '2023-01-06 14:00:00', '2023-01-06 15:00:00', 'Issue'),

('T-32', 'C-1', 'A-1', '2023-01-06 15:00:00', '2023-01-06 16:00:00', 'Issue'),

('T-33', 'C-1', 'A-1', '2023-01-07 10:00:00', '2023-01-07 11:00:00', 'Issue'),

('T-34', 'C-1', 'A-1', '2023-01-07 11:00:00', '2023-01-07 12:00:00', 'Issue'),

('T-35', 'C-1', 'A-1', '2023-01-07 12:00:00', '2023-01-07 13:00:00', 'Issue'),

('T-36', 'C-1', 'A-1', '2023-01-07 13:00:00', '2023-01-07 14:00:00', 'Issue'),

('T-37', 'C-1', 'A-1', '2023-01-07 14:00:00', '2023-01-07 15:00:00', 'Issue'),

('T-38', 'C-1', 'A-1', '2023-01-07 15:00:00', '2023-01-07 16:00:00', 'Issue'),

('T-39', 'C-1', 'A-1', '2023-01-08 10:00:00', '2023-01-08 11:00:00', 'Issue'),

('T-40', 'C-1', 'A-1', '2023-01-08 11:00:00', '2023-01-08 12:00:00', 'Issue'),

('T-41', 'C-1', 'A-1', '2023-01-08 12:00:00', '2023-01-08 13:00:00', 'Issue'),

('T-42', 'C-1', 'A-1', '2023-01-08 13:00:00', '2023-01-08 14:00:00', 'Issue'),

('T-43', 'C-1', 'A-1', '2023-01-08 14:00:00', '2023-01-08 15:00:00', 'Issue'),

('T-44', 'C-1', 'A-1', '2023-01-08 15:00:00', '2023-01-08 16:00:00', 'Issue'),

('T-45', 'C-1', 'A-1', '2023-01-09 10:00:00', '2023-01-09 11:00:00', 'Issue'),

('T-46', 'C-1', 'A-1', '2023-01-09 11:00:00', '2023-01-09 12:00:00', 'Issue'),

('T-47', 'C-1', 'A-1', '2023-01-09 12:00:00', '2023-01-09 13:00:00', 'Issue'),

('T-48', 'C-1', 'A-1', '2023-01-09 13:00:00', '2023-01-09 14:00:00', 'Issue'),

('T-49', 'C-1', 'A-1', '2023-01-09 14:00:00', '2023-01-09 15:00:00', 'Issue'),

('T-50', 'C-1', 'A-1', '2023-01-09 15:00:00', '2023-01-09 16:00:00', 'Issue');

Takeaway 3: The Agent as Your Analytical Co-pilot

Dremio’s UI-based AI Agent isn't just a SQL generator; it is a proactive co-pilot that uses Semantic Search to discover the right data products based on the Wikis and Labels you provide.

SQL Generation for Medallion Architecture

You can ask the agent to help you build a "Medallion Architecture." In Dremio, these are Logical Views, not physical copies. This means you get the organization of Bronze, Silver, and Gold layers without the cost of duplicating petabytes of data.

Improve support stats in Zendesk_Clone.Support.

   1. Bronze: Raw View of TICKETS, AGENTS, CUSTOMERS.

   2. Silver: 

      - Join: TICKETS -> AGENTS.

      - SLA Breach: If (Resolved - Created) > 4 Hours (for 'VIP' > 1 Hour).

   3. Gold: 

      - Agent Scorecard: Avg Resolution Time per Agent.

   Show the SQL.

Will likely get response like this:

-------------------------------------------------------------------------------

-- BRONZE LAYER: Raw Views

-------------------------------------------------------------------------------

-- Bronze: Raw TICKETS

CREATE OR REPLACE VIEW Zendesk_Clone.Support.bronze_tickets AS

SELECT * FROM Zendesk_Clone.Support.TICKETS;

-- Bronze: Raw AGENTS

CREATE OR REPLACE VIEW Zendesk_Clone.Support.bronze_agents AS

SELECT * FROM Zendesk_Clone.Support.AGENTS;

-- Bronze: Raw CUSTOMERS

CREATE OR REPLACE VIEW Zendesk_Clone.Support.bronze_customers AS

SELECT * FROM Zendesk_Clone.Support.CUSTOMERS;

-------------------------------------------------------------------------------

-- SILVER LAYER: Enriched with Joins & SLA Breach Logic

-------------------------------------------------------------------------------

CREATE OR REPLACE VIEW Zendesk_Clone.Support.silver_tickets_enriched AS

SELECT 

    t."TICKET_ID",

    t."CUST_ID",

    t."AGENT_ID",

    t."CREATED_TS",

    t."RESOLVED_TS",

    t."DESCRIPTION",

    a."NAME" AS agent_name,

    a."LEVEL" AS agent_level,

    c."EMAIL" AS customer_email,

    c."TIER" AS customer_tier,

    TIMESTAMPDIFF(HOUR, t."CREATED_TS", t."RESOLVED_TS") AS resolution_hours,

    CASE 

        WHEN t."RESOLVED_TS" IS NULL THEN NULL

        WHEN c."TIER" = 'VIP' AND TIMESTAMPDIFF(HOUR, t."CREATED_TS", t."RESOLVED_TS") > 1 THEN true

        WHEN c."TIER" != 'VIP' AND TIMESTAMPDIFF(HOUR, t."CREATED_TS", t."RESOLVED_TS") > 4 THEN true

        ELSE false

    END AS sla_breach

FROM Zendesk_Clone.Support.TICKETS AS t

LEFT JOIN Zendesk_Clone.Support.AGENTS AS a 

    ON t."AGENT_ID" = a."AGENT_ID"

LEFT JOIN Zendesk_Clone.Support.CUSTOMERS AS c 

    ON t."CUST_ID" = c."CUST_ID";

-------------------------------------------------------------------------------

-- GOLD LAYER: Agent Scorecard with Avg Resolution Time

-------------------------------------------------------------------------------

CREATE OR REPLACE VIEW Zendesk_Clone.Support.gold_agent_scorecard AS

SELECT 

    "AGENT_ID",

    agent_name,

    agent_level,

    COUNT(*) AS total_tickets,

    COUNT(CASE WHEN "RESOLVED_TS" IS NOT NULL THEN 1 END) AS resolved_tickets,

    AVG(resolution_hours) AS avg_resolution_hours,

    SUM(CASE WHEN sla_breach = true THEN 1 ELSE 0 END) AS sla_breaches,

    ROUND(

        CAST(SUM(CASE WHEN sla_breach = true THEN 1 ELSE 0 END) AS DOUBLE) / 

        CAST(COUNT(CASE WHEN "RESOLVED_TS" IS NOT NULL THEN 1 END) AS DOUBLE) * 100, 

        2

    ) AS sla_breach_pct

FROM Zendesk_Clone.Support.silver_tickets_enriched

WHERE "AGENT_ID" IS NOT NULL

GROUP BY "AGENT_ID", agent_name, agent_level

ORDER BY avg_resolution_hours;

Self-Documenting Metadata

Documentation is the foundation of trust. In Project Preferences, enable "Generative AI features." You will see a button to auto-generate Wikis (rich text descriptions). By inspecting the schema and sampling the data, the AI documents your datasets for you, making the semantic layer smarter over time.

Takeaway 4: Instant Visualization from Gold Datasets

Once your "Gold" datasets are ready, the AI Agent can move from "intent to execution" by generating visualizations.

  • Show me a bar chart comparing total tickets handled by each agent from gold_agent_scorecard
  • Create a bar chart of average resolution hours per agent from gold_agent_scorecard

Takeaway 5: SQL-Powered AI (Analyzing Silver Data)

One of Dremio's most powerful "magic moments" is the ability to call LLMs directly within SQL using AI Functions. This follows the "Ingest Anywhere, Consume Here" philosophy: use Spark for heavy-duty ingestion, but use Dremio as the "brain" for consumption.

Sentiment Analysis in SQL

Enrich your Silver-layer customer tickets by categorizing sentiment with a simple SELECT statement:

SELECT 
    "TICKET_ID",
    "DESCRIPTION",
    customer_tier,
    agent_name,
    resolution_hours,
    AI_CLASSIFY(
        'Classify the priority of this support ticket: ' || "DESCRIPTION",
        ARRAY['High', 'Medium', 'Low']
    ) AS ticket_priority
FROM Zendesk_Clone.Support.silver_tickets_enriched
WHERE "DESCRIPTION" IS NOT NULL
ORDER BY "CREATED_TS" DESC
LIMIT 20;

Unlocking "Dark Data"

You can even query unstructured data like PDFs. By combining LIST_FILES with AI_GENERATE, you can scan an S3 bucket of PDF invoices and extract structured fields directly into an Iceberg table (this would have to be an object storage source you have connected Dremio):

SELECT AI_GENERATE(('Extract invoice total', file) 

WITH SCHEMA ROW(total_amount DECIMAL)) AS invoice_data 

FROM TABLE(LIST_FILES('@Invoices/2025_Q1')) 

WHERE file['path'] LIKE '%.pdf';

Conclusion: The Future of the Self-Managing Lakehouse

The shift to Agentic Analytics is a transition from "passive metadata" (knowing where data is) to "agentic context" (the platform understanding what data means). By unifying federation, a robust semantic layer, and autonomous performance tuning, Dremio transforms the lakehouse from a static repository into an active partner.

The core challenge of modern analytics isn't the AI model, it's the data foundation. As you move from a "data swamp" to an Agentic Lakehouse, the cycle of learning compounds, allowing you to focus on decisions rather than infrastructure.

When AI is seamlessly integrated into every layer of the data stack, what manual tasks will you be happy to never do again?

Stop managing bottlenecks and start delivering breakthroughs. Start your Dremio free trial today.

Try Dremio Cloud free for 30 days

Deploy agentic analytics directly on Apache Iceberg data with no pipelines and no added overhead.

Start For Free

Article Topics

Product Insights from the Dremio Blog