Introduction
Astrato AI Features are a core part of the self-service capabilities. The AI features help users in all areas of the products, from giving fields and table business-friendly names to analyzing dashboards and surfacing hidden insights.
This article covers the management and setup of the AI features.
AI Settings
The Administration->System Settings -> AI modal manages the AI settings.
Providers
Custom AI providers can be configured for your instance. Use the "Configure" button to set up and manage your AI provider integrations.
Currently, the supported AI providers are:
Snowflake Cortex LLM (20+ models available)
Astrato Default Provider (Azure OpenAI 40-mini)
subject to fair usageAzure OpenAI
OpenAI
Snowflake Cortex LLM ❄️
When Snowflake Cortex LLM is used in Astrato, data does not need to be shared - we simply share the query and Snowflake does the rest of the work.
This removes the need for any additional data transfer, business contracts or security processes.
ℹ️ Some Astrato features share metadata, such as semantic models, which include measure definitions, field names and join criteria.
ℹ️ If you have not used Snowflake Cortex LLM before, please ensure you set it up and meet these Prerequisites:
Snowflake AccountAdmin role access
Astrato workspace with AI configuration enabled
Cortex enabled on your account
🔐 Step 1: Enable Cross-Region Cortex (if needed)
ALTER ACCOUNT SET CORTEX_ENABLED_CROSS_REGION = 'ANY_REGION';
👤 Step 2: Set Up Permissions in Snowflake
USE ROLE ACCOUNTADMIN; CREATE ROLE cortex_user_role; GRANT DATABASE ROLE SNOWFLAKE.CORTEX_USER TO ROLE cortex_user_role; GRANT ROLE cortex_user_role TO USER some_user;
Replace some_user with the username used by Astrato.
🔧 Step 3: Configure in Astrato
Go to Settings → AI Provider
Select Provider Type:
Snowflake Cortex LLMChoose a Name (e.g., Snowflake Cortex LLM (llama3.1-8b))
Under Data Connection, select the source (e.g.,
Demodata source)Under Model, pick from available models like:
llama3.1-8bclaude-3-5-sonnetdeepseek-r1
Click Apply
🧠 Astrato will now use Snowflake Cortex without data ever leaving your warehouse.
💡 Snowflake Cortex LLM Tips from Astrato
1️⃣ Start with llama3.1-8b if unsure, it’s optimized for cost/performance.
2️⃣ You can change models at any time without reconnecting the data.
3️⃣ Custom & fine-tuned models can be entered in the text box, if supported by Snowflake.
4️⃣ If you are able to process data in multiple regions, enable
Astrato Default Provider (Azure OpenAI 40-mini)
No settings are needed.
* This provider is best suited for testing the AI features
ℹ️ For customers who signed up on or before 20th March 2025, you will need to Add a new Provider here.
Configure a new provider, and choose Azure Open AI 4o-mini [provided by Astrato] as the provider to test out the AI features. We recommend using your own AI Provider for increased flexibility and model choices.
Open AI
setting an OpenAI provider
Provider Type- OpenAI
Name- Business-friendly name
API Key- enter the OpenAI Key
API URL- enter the OpenAI API URL, for example, https://api.openai.com/v1/
Support External Embeds- Enables the usage of the AI feature when Astrato is embedded in other software.
Model- Set the OpenaI Model that will be used.
Azure OpenAI
Setting an Azure OpenAI provider.
Provider Type- Azure OpenAI
Name- Business-friendly name
API Key- enter the OpenAI Key
API URL- enter the Azure OpenAI API URL, for example, https://azure-openai-184867.openai.azure.com/
Support External Embeds- Enables the usage of the AI feature when Astrato is embedded in other software.
Deployment Name- add API deployment name
API Version- set the API version
Model- Set the OpenaI Model that will be used.
Capability | Description | Notice & Risks for Third-Party AI Providers |
Insights | Enable AI-powered insight generation to discover patterns and meaningful information in your data automatically. This feature includes:
* This feature uses the built-in AI endpoint; the defined providers aren't supported yet. | ℹ️ Data from query results is neccesary to be shared for this capability
- Risk of bias in AI-generated insights. |
Creator Copilot | The Copilot feature assists users with Creator role permissions in automatically generating:
* This feature uses the built-in AI endpoint; the defined providers aren't supported yet. |
ℹ️ Data is not shared for this capability. Data structures & query requests are shared (fields, measures, aggregations, filter values). |
Field Name Suggestions | This AI-powered feature helps Creator role users generate business-friendly field names in the Dataview Editor. It suggests intuitive and meaningful names for fields based on the meta data in the table and context provided. * This feature uses the built-in AI endpoint; the defined providers aren't supported yet. |
ℹ️ Data is not shared for this capability. Data structures & query requests are shared (fields, measures, aggregations, filter values). |
Self-service Custom Report | Enable natural language custom report generation to allow users to create reports using conversational language. This feature makes report creation more accessible to users who may not be familiar with traditional query languages. | - Natural language queries may be misinterpreted, leading to incorrect reports.
ℹ️ Data is not shared for this capability. Data structures & query requests are shared (fields, measures, aggregations, filter values). |
Measure Generation | This feature allows users with Creator permissions to create measures in natural language.
When enabled, users can describe the measure they want to create in plain language, and the system will generate the appropriate measure definition. | ℹ️ Data is not shared for this capability. Data structures & query requests are shared (fields, measures, aggregations, filter values). |
Actions | Astrato actions include a block for custom prompting. Request custom insights, ask for validation, generate reports, recommendations and actions. | ℹ️ Data sharing is determined by the creator of the action workflow. |
Disclaimer
At Astrato Analytics, we are committed to ensuring that our AI-powered features align with the latest regulatory frameworks, including the EU AI Act. While we serve a diverse range of customers, including those in healthcare, the majority of our AI applications fall into the minimal or limited risk categories, ensuring compliance without imposing significant regulatory burdens.
1. Minimal Risk AI Systems
Most of our AI-powered features, such as insight generation, field name suggestions, and measure generation, are designed to assist users in data analysis and reporting. These systems do not make autonomous decisions or significantly impact individuals, placing them in the minimal risk category under the EU AI Act.
For these features, no additional compliance requirements apply, as they pose no significant risk to users or society.
2. Limited Risk AI Systems
For our systems, we ensure transparency by clearly informing users when they are interacting with AI and providing explanations of how the AI works.
This aligns with the EU AI Act's requirements for limited risk systems, ensuring users are aware of AI involvement without imposing heavy compliance obligations.
3. High Risk Considerations (Healthcare Customers)
For customers in healthcare or other high-risk sectors, we take additional precautions to ensure compliance with the EU AI Act. While Astrato itself does not directly provide high-risk AI systems, we support our customers in meeting their regulatory obligations by:
Ensuring data governance and quality standards are maintained.
Providing tools for transparency and explainability in AI-driven insights.
Enabling human oversight to validate AI-generated outputs.
Our platform is designed to empower users in high-risk sectors to comply with the EU AI Act while leveraging AI responsibly.
4. No Unacceptable Risk AI Systems
Astrato Analytics does not develop or deploy AI systems that fall into the unacceptable risk category (e.g., social scoring or manipulative AI). Our focus is on providing tools that enhance productivity and decision-making without compromising ethical or regulatory standards.
Key Takeaways for Stakeholders:
Low Regulatory Burden: The majority of Astrato's AI features are minimal or limited risk, requiring no additional compliance measures beyond transparency and user awareness.
Support for High-Risk Sectors: For healthcare customers, we provide the tools and framework to ensure compliance with the EU AI Act's high-risk requirements.
Commitment to Ethical AI: We prioritize transparency, explainability, and user control in all our AI-powered features, ensuring they are both effective and compliant.
By focusing on minimal and limited risk AI systems, Astrato ensures that our customers can leverage the power of AI without facing significant regulatory challenges. For high-risk use cases, we provide the necessary support to ensure compliance while maintaining the highest standards of data privacy and security.