DataSelf uses AI in two complementary ways: AI-assisted features help users perform data and analytics work, while agentic features allow an AI agent to perform authorized tasks through DataSelf tools.
The simplest distinction is:
-
AI-assisted: AI helps you decide what to do and create what is needed.
-
Agentic: AI can use authorized tools to help carry out the work.
Agentic functionality does not necessarily mean fully autonomous operation. The agent works within the tools, permissions, and approval requirements made available through the DataSelf environment.
Key Differences
|
Capability |
AI-Assisted |
Agentic |
|---|---|---|
|
Primary purpose |
Help the user perform work |
Perform authorized work on the user’s behalf |
|
AI’s role |
Explain, generate, recommend, and analyze |
Plan steps, select tools, execute tasks, and report results |
|
User’s role |
Review the output and perform or apply the operation |
Define the objective, authorize the scope, and review the outcome |
|
Access to DataSelf tools |
Not necessarily required |
Required through supported tools such as DataSelf MCP+ |
|
Ability to change the environment |
Typically provides proposed content or instructions |
Can perform supported changes when authorized |
|
Multi-step execution |
Usually guided by the user |
Can coordinate multiple supported steps |
|
Permissions |
Limited to the application and information provided |
Controlled by authentication, authorization, tool access, and environment permissions |
|
Example |
Generate SQL for a new transformation |
Create or update the transformation and run the appropriate job |
AI-Assisted Features
AI-assisted features help users develop, understand, manage, and troubleshoot data and analytics solutions more efficiently.
The user remains the primary operator. AI may generate content, explain an environment, recommend an approach, or identify a likely problem, but the user reviews the output and initiates or applies the resulting operation.
AI-assisted capabilities can help users:
-
Generate SQL queries and transformation logic.
-
Create formulas, calculations, and KPIs.
-
Recommend data-cleansing and optimization steps.
-
Identify possible relationships among tables and fields.
-
Assist with the design of dimensions, facts, and star or galaxy schemas.
-
Explain ETL statements, jobs, scripts, and data models.
-
Diagnose errors and recommend corrective actions.
-
Create reports, dashboards, charts, and analytical summaries.
-
Document data sources, transformations, models, and business definitions.
-
Ask questions about trusted Data Warehouse data using natural language.
For example, a user might ask AI to:
“Write the SQL needed to add the salesperson category to this transformation.”
The AI can generate the SQL and explain its logic. The user can then review, test, and apply it.
Agentic Features
Agentic features extend AI assistance by allowing an AI agent to interact with authorized DataSelf tools and perform supported operations on the user’s behalf.
The user provides the objective, and the agent can determine which available tools and steps are needed to complete the task. Depending on its permissions, the agent may retrieve information, execute queries, modify supported objects, run processes, or coordinate a multi-step operation.
DataSelf MCP+ provides supported tools and contextual information to an MCP-compatible AI agent. The agent can then interact with DataSelf ETL+, the DataSelf Data Warehouse, and other supported components within the authenticated user’s permitted environment.
Agentic capabilities can help users:
-
Explore entities, data sources, tables, statements, jobs, and scripts.
-
Retrieve job configurations, steps, status, and execution information.
-
Refresh source-schema metadata.
-
Identify and respond to source-system schema changes.
-
Mirror source tables into the Data Warehouse.
-
Create or modify Data Warehouse tables and views.
-
Generate and execute transformation statements.
-
Run controlled, read-only Data Warehouse queries.
-
Start or stop ETL jobs.
-
Create, update, or execute supported external scripts.
-
Investigate failed data loads using metadata and execution details.
-
Perform authorized multi-step data-engineering procedures.
For example, a user might ask:
“Refresh the source metadata, identify any new sales fields, add the appropriate field to the staging table, and run the sales load.”
An authorized agent can retrieve the relevant metadata, determine which objects and tools are required, prepare or perform the supported changes, and report the outcome.
Comparing Common Requests
“Create a staging table for this source query.”
AI-assisted:
AI generates the recommended table definition and explains how to create it. The user reviews and executes the SQL.
Agentic:
The agent inspects the source metadata, generates the table definition, creates the table through an authorized tool, and reports the result.
“Why did last night’s ETL job fail?”
AI-assisted:
The user provides the error or job information, and AI explains the likely cause and recommends a solution.
Agentic:
The agent retrieves the job configuration, processing steps, status, and available execution details; investigates the failure; and may perform an authorized corrective action.
“Refresh the sales data.”
AI-assisted:
AI explains which process to run and what should be reviewed before starting it.
Agentic:
The agent identifies the appropriate job, starts it through an authorized tool, monitors the available status, and communicates the outcome.
“Add a new source field to the Data Warehouse.”
AI-assisted:
AI helps identify the required extraction, transformation, table, and model changes.
Agentic:
The agent can inspect the changed source schema and perform the supported extraction, transformation, and Data Warehouse modifications after receiving the required authorization.
How They Work Together
AI-assisted and agentic features are not competing approaches. They represent different levels of AI participation within the same data and analytics lifecycle.
A typical workflow may include both:
-
AI explores the available metadata and helps the user understand the requirement.
-
AI recommends a transformation, model, or processing change.
-
The user reviews and authorizes the proposed approach.
-
An agent uses approved DataSelf tools to perform the supported operation.
-
The DataSelf platform applies the validated logic consistently during future processing.
AI assistance makes expertise and development faster and more accessible. Agentic capabilities extend that value by reducing the manual effort required to carry out supported tasks.
Access, Permissions, and Human Oversight
Agentic operations are limited to the tools and information exposed by DataSelf MCP+ and the permissions available within the connected environment.
Depending on the configuration, an agent may have:
-
Metadata-only access.
-
Read-only data access.
-
Permission to execute selected processes.
-
Permission to create or modify supported objects.
-
Access to only specific entities, jobs, scripts, or environments.
Authentication, authorization, data permissions, tool restrictions, approval requirements, and auditing determine what an AI application or agent may access and do.
Users should review AI-generated logic, test significant changes in a non-production environment, and require human approval for sensitive or high-impact operations. Agentic functionality does not eliminate the need for security, backups, testing, change management, and appropriate human oversight.
Summary
DataSelf AI-assisted features help users understand, design, analyze, and troubleshoot.
DataSelf agentic features allow authorized AI agents to use supported tools to help execute and manage the work.
Together, they help organizations accelerate data engineering and analytics while retaining control over their data, business rules, permissions, and production environments.