DataSelf AI Options and Data Protection

DataSelf gives organizations control over whether and how artificial intelligence interacts with their data environment. AI is optional, and clients can choose an approach that aligns with their security, privacy, governance, and business requirements.

Depending on the selected option, AI tools may receive:

• No client information

• Data warehouse metadata and business definitions, but no transactional data values

• Authorized query results from the curated data warehouse

• Permission to perform specifically authorized ETL and data-modeling activities

The selected level can be changed at any time through written authorization from the client.

How DataSelf Protects Client Data

DataSelf separates operational source systems from reporting, analytics, and AI through a controlled data warehouse environment. Data is extracted from authorized sources, organized into business-ready models, and made available according to the authentication, permissions, and security policies configured for the client.

This architecture helps protect operational systems and provides AI and analytics tools with more clearly defined information. Instead of requiring an AI assistant to navigate the client’s ERP, CRM, or other transactional system directly, DataSelf can provide access to curated data models containing established relationships, calculations, mappings, groupings, and business definitions.

Depending on the deployment and selected options, protection can include:

• Separation between operational source systems and the analytical environment

• Authentication and user-specific permissions

• Row-level security that limits which records an authorized user can access

• Controlled, read-only access to data warehouse models through DataSelf MCP+

• Metadata-only AI workflows that do not provide transactional data values to AI tools

• Restricted tools and permissions for agentic ETL activities

• Human authorization for changing the approved level of AI participation

AI services are also subject to the security, privacy, retention, and contractual terms of the selected AI provider. Clients should approve the AI providers and configurations used with their DataSelf environment.

Available AI Options

No AI

Under the No AI option, DataSelf operates without AI-enabled features or touchpoints. No client data or metadata is submitted to AI tools through DataSelf.

Data extraction, transformation, data modeling, reporting, dashboards, and analytics continue to operate using traditional DataSelf capabilities. This option allows an organization to benefit from the DataSelf data warehouse and business intelligence platform without connecting its DataSelf environment to an AI service.

This option is appropriate when organizational policies prohibit AI use or when the client prefers to introduce AI at a later stage.

AI for Data Analysis

Under this option, authorized users may use approved AI assistants to analyze information from the client’s curated data warehouse.

DataSelf MCP+ provides controlled, read-only access to authorized data models. This helps AI work with known, organized, and business-ready information rather than attempting to interpret raw operational tables or directly navigate the client’s source systems.

When an authorized user asks a question, the AI service may process the metadata and data warehouse query results needed to prepare its response. Access can be constrained by the security configured for the user and environment. Where implemented, row-level security can further limit the records available to each user.

Examples include:

• Identifying customers with declining revenue or margins

• Explaining changes in sales, inventory, purchasing, receivables, or cash flow

• Investigating unusual transactions or business trends

• Summarizing authorized data warehouse results

• Helping users explore business information through conversational questions

Read-only access prevents the AI assistant from modifying the data warehouse through DataSelf MCP+. It does not guarantee that every AI-generated interpretation will be correct. Users should validate important conclusions and decisions against the underlying business information.

AI for Data Modeling Analysis

This option allows AI to assist with understanding and analyzing how the client’s data warehouse is organized without providing business transactional data values to AI tools.

AI tools may receive metadata and definitions such as:

• Table and field names

• Data types and model structures

• Relationships and joins

• KPI calculations and business formulas

• Mappings, classifications, and groupings

• Transformation logic

• Documentation and model descriptions

This information can help AI explain existing business logic, identify documentation gaps, review model consistency, and assist with the design of reporting-ready models. For example, AI might explain how gross margin is calculated or identify which mappings place customers into specific reporting groups.

Under this option, AI tools receive model metadata and definitions, but no transactional data values. Metadata can still reveal information about an organization’s systems and business structure, so access should remain limited to approved AI providers, users, and use cases.

Agentic ETL and Data-Warehouse Modeling

Agentic ETL extends AI assistance from analysis and recommendations to specifically authorized actions. When enabled, AI agents may use approved DataSelf tools to perform ETL and data-modeling activities.

Authorized activities may include:

• Adding source tables to the extraction process

• Creating or modifying transformations

• Building or adjusting data models

• Creating calculated fields and business logic

• Running or stopping data-refresh processes

• Loading selected tables or groups of tables

• Assisting with documentation and model maintenance

Agentic activities are limited to the tools, environments, permissions, and approval requirements configured within the client’s DataSelf security environment. The information and actions available to an AI agent depend on the client-approved configuration.

Because agentic activities can change the data warehouse or its refresh processes, organizations should define which users can request actions, which actions require human review, and which environments the agent may modify. Important modeling and production changes should be tested and validated according to the client’s normal change-management practices.

Choosing an Option

The appropriate option depends on the organization’s objectives and policies:

• Choose No AI when AI must not interact with the DataSelf environment.

• Choose AI for Data Analysis when users should be able to analyze authorized data warehouse information conversationally.

• Choose AI for Data Modeling Analysis when AI should help explain or review business rules and models without receiving transactional data values.

• Choose Agentic ETL and Data-Warehouse Modeling when authorized AI agents should help build, operate, or maintain the data warehouse.

AI options can be introduced gradually. For example, an organization might begin with metadata-only model analysis, later enable read-only data analysis for selected users, and then authorize specific agentic ETL activities after establishing appropriate approval and change-management processes.

Changing the Selected Option

DataSelf does not require the use of AI. The client may change its approved level of AI participation through written authorization. The authorization should identify the selected option, approved AI provider or providers, authorized users, applicable environments, and any limitations or approval requirements.

Contact DataSelf to review the available options and select the configuration that best aligns with your organization’s data, security, governance, and analytical requirements.