APEX · Workload Governance Agent

One next action for every workload.

Where should this workload run, and who changes it? APEX learns the job’s context, recommends Run, Optimize, Redirect, or Review, and sends the change to its owner.

  1. Trace the job
  2. Assign the owner
  3. Act on one recommendation

Your business data stays in your account Production changes wait for your approval

What a workload is

The job behind the cost, not the account it landed on.

Native tools often stop at the warehouse, the reservation, or the project. APEX names the work.

dbt

A model, test, or run. The name most teams already use.

Fivetran

A connector sync on a schedule.

Airbyte

An open-source sync beside Fivetran.

Airflow

The DAG that starts the warehouse or slot work.

Snowpark

DataFrame and Python jobs inside Snowflake.

Dynamic Table

Snowflake’s declarative pipeline. A refresh, not a warehouse.

Snowpipe

A continuous or batch load into Snowflake.

Task

A Snowflake Task. Serverless or warehouse-backed.

Cortex

Snowflake AI functions and token spend.

Cortex Analyst

A question against a semantic view.

Cortex Search

A search service and the queries against it.

Cortex Agent

A Snowflake agent run.

Streamlit

A Streamlit in Snowflake session.

Dataform

BigQuery’s pipeline tool. The job Google teams schedule.

Scheduled query

A BigQuery scheduled query.

Continuous query

A BigQuery continuous query.

BigQuery ML

A model training or prediction job.

Notebook

A Snowflake notebook or a BigQuery Studio notebook.

dbt Agent

An agent acting on a dbt project.

Looker

A Looker query or dashboard refresh.

Looker Studio

A Looker Studio report on BigQuery.

Tableau

An extract or a live query.

Power BI

A dataset refresh.

Sigma

A Sigma workbook query.

Connected Sheets

A Sheets refresh against BigQuery.

The recommendation

Four answers. Your team still makes the change.

Redirect names a better path. It does not move the workload. Optimize keeps the requirement and uses less compute.

Run

Keep it moving. Performance is met and the spend is in line.

Optimize

Use less compute for the same requirement. Your team approves it.

Redirect

A better path. Your team makes the move.

Review

Needs a person before it becomes scheduled spend.

What APEX does not do

Read the run. Recommend. Stop there.

It does not block the query

APEX is not in the path of the workload. It reads how the run went.

It does not apply the change

A recommendation waits for the review your team already uses.

Table contents stay put

Snowflake stays in Snowflake. BigQuery stays in BigQuery. The input is how the workload ran and what it cost.

Snowflake and BigQuery

Same decision. The words of each platform.

Cortex is a Snowflake workload. BigQuery uses jobs, slots, and reservations. The recommendation is still one of the four.

Snowflake

Credits tied to the job

dbt, Snowpark, Dynamic Tables, and Cortex, not only the warehouse.

dbt, Fivetran, Snowpark the work
Warehouses the cost
Agents and Cortex the new bill

Snowflake

BigQuery

Jobs and slots, with a next step

dbt, Dataform, scheduled queries, BigQuery ML, and Looker.

Jobs the work
Slots the cost
Reservations where it runs

BigQuery

Availability

On Microsoft Marketplace, with a free trial.

APEX is listed as a SaaS offer for Snowflake workload governance. Teams who buy through Microsoft can start the trial from the listing and connect an account in about five minutes. Read the announcement on the wire, or see every Anavsan release on the press page.

See one workload get one recommendation.

The example is on the home page. Book a Demo when you want the same decision walked through on your account.