AI query optimization tool for data warehouses

Yes. In fact, the major cloud data warehouses now have workload-aware optimization that is very close to what you describe. The best choice depends heavily on your warehouse.

If you're on BigQuery

This is probably the closest match to your exact request.

BigQuery's partitioning and clustering recommender analyzes up to 30 days of workload execution data and uses ML to predict which partitioning/clustering configurations could reduce workload cost. It gives you the table, recommended column, and estimated monthly savings.docs.cloud.google.com

Its materialized-view recommender similarly examines the previous 30 days for repetitive query patterns and recommends incremental MVs when the expected query savings outweigh maintenance costs.docs.cloud.google.com

There's also history-based optimization, where BigQuery learns from previous executions of the same or similar queries and automatically applies optimizations when it has high confidence they'll help.docs.cloud.google.com

And Gemini Cloud Assist can analyze individual SQL queries and suggest query-level improvements.docs.cloud.google.com

So a BigQuery setup can effectively give you:

query history → workload analysis → partition/cluster recommendation → MV recommendation → query-plan optimization

If you're on Redshift

I'd put Amazon Redshift's built-in automation at the top of the list if you want the warehouse itself to make changes.

Redshift Automatic Table Optimization continuously observes how queries interact with tables and uses AI methods to select sort and distribution keys. It can actually modify tables rather than merely telling you what to do.docs.aws.amazon.com

Even more interestingly, Redshift's autonomics include automated materialized views: it can create and refresh MVs based on observed query patterns and ML algorithms.docs.aws.amazon.com

If you're on Snowflake

Snowflake has become particularly interesting in 2026. Optima Planning learns from past executions to improve plans for recurring workloads, while newer Optima capabilities continuously analyze workload patterns for metadata/pruning and indexing-related optimization.docs.snowflake.com

Snowflake also has AI-driven query optimization through Cortex Code, plus automatic materialized-view rewriting.docs.snowflake.com

If you're on Databricks

Databricks' automatic liquid clustering is almost exactly the workload-driven behavior you're asking about. It analyzes a table's historical query workload, identifies candidate clustering columns, and changes them as query patterns change—while considering whether the expected data-skipping savings justify the clustering cost.docs.databricks.com

That's particularly compelling if you're moving away from manually choosing partitions/Z-ordering on Delta tables.

One important caveat: "indexing" means different things

Traditional index recommendation is much more common in OLTP databases than modern warehouses. For example, DBtune analyzes real PostgreSQL workload statistics and produces ranked, concrete index recommendations—including composite, covering, expression, and partial indexes—with estimated speedups and write/storage costs.www.dbtune.com

For BigQuery/Snowflake/Databricks, the analogous levers are generally partitioning, clustering/data layout, search optimization, materialized views, and query-plan optimization, rather than conventional B-tree indexes.

My recommendation: don't start with a generic "AI SQL optimizer." If you tell me which warehouse you're using (Snowflake, BigQuery, Redshift, Databricks, Synapse, ClickHouse, Postgres, etc.), I can identify the best native AI/autotuning capability and, if necessary, the best third-party tool, including what query-history data it needs, what changes it can recommend automatically, and how I'd set up a safe evaluation against your workload.

Yes, modern cloud data warehouses and third-party tools increasingly use machine learning and heuristic analyzers to parse query history, detect sequential bottlenecks, and recommend or automate indexes, partitioning, and materialized views.energent.ai

The best options depend on whether you want a native feature tied to your specific cloud data warehouse or a third-party AI/observability tool that plugs into your stack.

1. Native Warehouse Optimizers (Best if you're locked into a specific ecosystem)

• Snowflake (Snowflake Copilot & Advisor): Snowflake tracks query history automatically. Its built-in Advisor and performance features analyze recurring scan patterns, clustering keys, and search optimization service needs. It surfaces recommendations for automatic clustering or search optimization when queries struggle with large micro-partitions.
• Google BigQuery (BigQuery Recommender): BigQuery uses machine learning on your historical query logs (typically over a 30-day window) to automatically generate optimization recommendations. It suggests partitioning or clustering improvements, unutilized materialized views, and cost-saving query rewrites directly in the GCP console or via the Recommender API.querio.ai
• Databricks (Lakehouse Optimizer / Predictive Optimization): If you are running on a lakehouse architecture with Delta Lake, Databricks provides predictive optimization which automatically handles compaction (bin-packing small files) and Z-ordering/clustering based on how your tables are queried.medium.com
• AWS Redshift (Redshift Advisor): Redshift analyzes workload query history to recommend optimal sort and distribution keys, materialized views, and compression encodings.

2. Third-Party AI & Observability Platforms (Best for multi-cloud or deep tuning)

• Dedicated Database AI Tools: Platforms like Energent.ai specialize in schema-aware intelligence, parsing query logs, learning evolving application patterns, and proactively suggesting specific indexes or structural changes.energent.ai
• Data Observability & FinOps Tools (Monte Carlo, Datadog, Chaos Genius): While primarily used for data reliability and cost monitoring, their AI layers increasingly flag high-compute, recurring slow queries and map out structural data model flaws (like heavy data skew or missing partition prunes).

A Word of Caution on AI-Driven Schema Changes

While AI is great at spotting patterns (e.g., "You filter by date and region on 80% of queries, so you should partition by those" ), always cross-reference recommendations with your ETL pipeline constraints. For instance, over-indexing or poorly chosen streaming partition keys can quietly tank your data ingestion speeds or bloat your storage costs.icertglobal.com

To narrow down the best tool or manual strategy:

• Which data warehouse are you currently using (e.g., Snowflake, BigQuery, Redshift, Databricks)?
• Are your slow queries mostly heavy aggregations (ideal for materialized views) or massive full-table scans (ideal for partitioning/clustering)?

Yes. There are both native platform features built directly into modern cloud data warehouses and third-party AI optimization tools designed specifically to scan query history, analyze access patterns, and autonomously recommend or implement performance boosters like materialized views and indexing strategies.

1. Native Cloud Data Warehouse Features

Major cloud data platforms use internal machine learning and cost-based analysis engines to evaluate your active query logs natively:

• Google BigQuery (Materialized View Recommender): BigQuery automatically analyzes historical query execution characteristics over a rolling window (e.g., the past 30 days). It features a built-in Materialized View Recommender that scans your query logs to identify repetitive patterns and suggest optimal materialized views, complete with smart tuning to transparently rewrite incoming queries. You can view these recommendations in the Google Cloud Console or via BigQuery's recommendation API.docs.cloud.google.com
• Snowflake (Optima Indexing & Optima Planning): Snowflake features automated optimization capabilities like Snowflake Optima , which continuously studies workloads from your query history. When it detects recurring point-lookup or predictable patterns that would benefit from indexing, it proactively creates and maintains search indexes automatically. Its companion, Optima Planning , learns from recurring executions (like BI dashboards or scheduled pipelines) to dynamically refine query plans for massive speedups.snowflake.com+1
• Amazon Redshift (Redshift Advisor): Redshift Advisor continuously inspects your cluster’s system tables and query logs. It outputs automated recommendations for sorting and distribution keys, unoptimized queries, and tables that could benefit from materialized views.

2. Specialized Third-Party & AI Tools

If you are running multi-cloud environments, open-source data lakes, or traditional database backends alongside your warehouse, external AI-driven query optimizers can ingest query logs:

• EverSQL / Polyform: Tools like EverSQL use AI to analyze slow-query logs and suggest structural changes, while newer platforms like Polyform leverage AI models paired with metadata/execution validation to propose efficient query rewrites, structural modifications, and storage strategies by comparing historical workloads in a test sandbox.medium.com+1
• Databases with Hypo-Indexing (e.g., PostgreSQL/DuckDB ecosystems): While more transactional/hybrid, tools utilizing hypothetical indexes (like HypoPG ) allow optimization scripts to test indexing strategies against captured query logs without actually locking or consuming real storage for the index upfront.medium.com

To narrow down which approach fits best, could you tell me:

• Which cloud data warehouse are you currently using (e.g., Snowflake, BigQuery, Redshift, Databricks)?
• Are you looking for a fully automated native tool or an external analyzer to audit logs across multiple systems?

Yes. There are several AI-assisted and automated tools that analyze data warehouse query workloads and recommend performance improvements, including materialized views, partitioning, clustering, indexing-like structures, and query rewrites.

The important distinction is that some tools analyze query history and recommend changes, while others use workload analysis and machine learning to automatically implement or maintain optimizations.

Best tools to evaluate

• 

1. Google BigQuery

(Strong for recommendations)

BigQuery's built-in recommenders analyze historical workload behavior to identify opportunities for materialized views, partitioning, and clustering. Its materialized-view recommender looks for repeated query patterns and weighs potential execution savings against maintenance costs.docs.cloud.google.com+2

Best for: BigQuery-native warehouses, especially teams looking to reduce query cost and latency without building their own tuning system.

Materialized-view recommendations

---

• 

2. Databricks SQL and Genie Code

Databricks provides query-history performance insights, actionable tuning suggestions, and an AI-assisted optimization workflow. Genie Code can rewrite queries for approval, while predictive optimization automatically handles certain data-layout and statistics-maintenance tasks. Predictive I/O also uses learned access patterns to accelerate supported scans.docs.databricks.com+2

Best for: Lakehouse environments where the bottleneck may be file layout, data skipping, joins, or query structure rather than traditional indexes.

Query performance insights

---

• 

3. Amazon Redshift

Redshift's Automated Materialized Views (AutoMV) use workload monitoring and machine learning to select and maintain useful materialized views. Eligible queries can then be transparently rewritten to use them, without requiring application SQL changes.docs.aws.amazon.com+1

Best for: AWS warehouses where you want automatic optimization with minimal manual intervention.

Automated materialized views

---

• 

4. Snowflake

Snowflake has workload-oriented query analysis and several built-in optimization mechanisms: materialized views, automatic clustering, search optimization, and query acceleration. These address different access patterns, although its storage optimizations are not all an AI-generated recommendation system.docs.snowflake.com+2

Best for: Snowflake teams seeking native performance diagnostics and managed optimization features.

Snowflake performance options

---

• 

5. Azure SQL Database

Query Store captures historical SQL queries, execution plans, and runtime statistics. Azure SQL also offers an Index Advisor for index recommendations. This is particularly relevant if your warehouse is relational SQL and you specifically want traditional index suggestions.learn.microsoft.com+1

Best for: SQL Server-style workloads where index selection and execution-plan regressions are major concerns.

Query Store and performance tuning

What these tools can actually recommend

A key caveat: data warehouses don't all use traditional indexes. For example, BigQuery relies heavily on partitioning and clustering, while Snowflake uses micro-partition pruning and specialized search optimization. The right tool depends on the engine.docs.snowflake.com+2

What if you want an AI tool that works across your warehouse?

If your goal is to scan query logs, rank opportunities, and generate concrete proposals rather than rely on one cloud provider's built-in advisor, I'd evaluate a custom workload-analysis pipeline alongside the native tools.

A useful system would:

1. Ingest query history, execution plans, runtime statistics, bytes scanned, and compute costs.
2. Group similar SQL statements into workload patterns.
3. Identify expensive scans, repeated aggregations, join bottlenecks, and redundant computation.
4. Generate candidate materialized views, indexes or clustering keys, and SQL rewrites.
5. Estimate performance gains and ongoing storage, refresh, and maintenance costs.
6. Validate candidates against representative workloads before recommending production changes.

An LLM can help explain findings and generate candidate SQL, but it should work alongside query-plan analysis and cost modeling, not replace them. Recommendations should be tested before deployment.

My recommendation

Start with the native optimizer for your warehouse if you use BigQuery, Redshift, Snowflake, or Databricks. Those tools already have access to engine-specific workload information that a generic AI assistant may not have.

If you need one analysis layer across multiple engines, a custom system that combines query-log mining, plan analysis, workload simulation, and an LLM is worth investigating. I haven't established that a single off-the-shelf product is best across all these platforms.

One question that would help me narrow this to the most relevant tools: Which data warehouse are you running?

Your environment

Snowflake

Google BigQuery

Amazon Redshift

Databricks SQL / Delta Lake

PostgreSQL or another relational warehouse

Multiple warehouses

---

What should the tool do?

Recommend changes for engineers to review

Automatically implement and validate safe optimizations

Rank opportunities by ROI and cost savings

---

What data can it access?

Query logs and execution plans

Query logs, plans, and schema metadata

Read-only access to the warehouse

[Find the best-fit tools]