How to Create and Use AI Prompt Columns in Microsoft Dataverse
You can create and use ai prompt columns in Microsoft Dataverse by adding columns that generate data using generative ai models. These columns let you design prompts that analyze your data and provide instant insights. When you add ai prompt columns, you unlock advanced natural language understanding and real-time ai-driven interactions. Many organizations report faster development, better security, and seamless integration with Power Apps and Power BI. You gain flexibility to customize prompts and connect them to related dataverse tables for smarter automation.
Key Takeaways
AI prompt columns in Dataverse let you use AI to analyze data and get instant insights, improving automation and decision-making.
You need proper licenses, permissions, and environment setup before creating AI prompt columns to ensure smooth operation.
Create AI prompt columns by adding new columns with the 'Prompt' data type and customize prompts with clear instructions and relevant data.
Test and review AI outputs carefully to ensure accuracy and trustworthiness before using them in your business processes.
Use AI prompt columns to automate tasks like sentiment analysis, feedback classification, and generating replies, saving time and boosting productivity.
Prerequisites for Dataverse AI Columns
Before you start creating AI prompt columns, you need to make sure your setup meets a few important requirements. These steps help you unlock the full power of generative AI in your dataverse database and ensure a smooth experience with plug-in features.
Licenses and Permissions
You need the right licenses and permissions to use AI prompt columns in Microsoft Dataverse. Most users require a Power Apps or Power Automate license that supports AI Builder features. If you want to create or manage AI models, you must have the Environment Maker role. To publish AI models that change the dataverse schema, such as adding new columns or tables, you need System Customizer or System Administrator privileges. These roles also allow you to run category classification models and manage plug-in operations. Always check that your security roles match your responsibilities in the dataverse accelerator app or power app environment.
Environment Setup
Setting up your environment is a key step. You must have access to a power app environment where you can create and manage tables. Assign the Environment Maker role to users who will build AI models and prompts. System Customizers or System Administrators should handle publishing models that add new columns or tables. For prediction models that create new data columns, only System Customizers can publish them at first. Category classification models also require System Customizer or System Administrator roles to run. Make sure you have the right access to dataverse tables and plug-in features before you begin.
Tip: Assign security roles carefully to avoid issues when working with AI prompt columns and plug-in components.
Enabling AI Features
To enable AI features in your Microsoft Dataverse environment, follow these steps:
1. Sign in to the Power Platform admin center. 2. Go to Environments and select your environment. 3. Open Settings, then Product, and choose Features. 4. Under AI Builder, turn on the AI prompts feature.
If you manage several environments with environment groups, you can enable AI prompts for all of them at once:
In the Power Platform admin center, select Environment groups.
Choose the group you want.
Go to the Rules tab and enable the AI prompts toggle.
Save your changes.
Publish the rules to apply them to every environment in the group.
When you publish environment group rules, individual admins cannot change the AI prompts setting. This ensures consistency across your plug-in and dataverse accelerator app deployments.
Creating AI Prompt Columns
Accessing Dataverse Tables
You start by opening your Dataverse environment in Power Apps or Power Platform. Find the dataverse table where you want to add AI prompt columns. You can use the table designer to view and manage your data structure. This is where you will add new columns and set up plug-in features.
Tip: Make sure you have the right permissions before you begin. Only users with the correct roles can add or edit columns in a dataverse table.
Adding AI Prompt Columns
To add an AI prompt column, follow these steps:
Open your dataverse table in the table designer.
Select Add column.
Enter a name for your new column, such as "Customer Sentiment" or "Feedback Category."
For the data type, choose Prompt. This option lets you use AI to generate text based on your data.
Click Add a new prompt. This opens the AI prompt setup window.
When you add a prompt column, you can connect it to your data. In the Knowledge section, select columns from your table or related tables. These columns become knowledge objects for the AI. You can filter knowledge by choosing a column, picking a filter attribute (like text, number, or date), and entering filter values. This helps the AI focus on the most relevant data.
Note: You can reference columns from related tables to give the AI more context. For example, you might use product details from another table to help the AI understand customer feedback better.
Customizing Prompts
Customizing prompts is key to getting accurate and useful results from your AI prompt columns. Here are some best practices:
Write clear and specific instructions for your prompt. Tell the AI exactly what you want it to do.
Break down complex tasks into smaller steps. This makes your prompts easier for the AI to follow.
Use the "Test Your Copilot" panel to try out your prompt and see how the AI responds. Adjust your instructions until you get the results you want.
Add context by referencing columns from your table or related tables. This helps the AI understand the data and provide better answers.
Use retrieval-augmented generation (RAG) if you need the AI to use information from external sources. This can improve the accuracy of your results.
You can also use prebuilt prompt templates from the AI Builder Prompt Library. These templates help you get started quickly and ensure your prompts follow best practices.
Alert: Always test your prompts before using them in production. AI-generated answers should be checked for accuracy, especially when they affect business decisions.
Referencing Columns as Knowledge
To make your AI prompt columns smarter, you can reference columns from the same or related tables as knowledge sources. Here’s how you do it:
Add your Dataverse tables as knowledge sources in Copilot Studio.
Define synonyms and glossary terms for your columns. This helps the AI understand special terms or codes in your data.
Enable Dataverse search so the AI can look up information in your columns.
Mark the columns you want as searchable in the Quick Find View in Power Apps.
Publish your changes to make the columns available for AI prompts.
If you want to speed up indexing, you can re-add the knowledge source after making changes.
Practical Examples
AI prompt columns have many real-world uses. Here are some examples:
For example, you can use a prebuilt prompt to analyze customer feedback and detect if it is positive, negative, or neutral. Another prebuilt prompt can classify feedback into categories like bug, feature request, or compliment. You can even set up a prompt to suggest actions based on the feedback and product details, such as offering a refund if the product price is below a certain amount and the feedback is negative.
Plug-in Tip: AI prompt columns work with plug-in features, so you can automate actions based on AI results. For example, you can trigger a plug-in to send an email reply when the AI detects negative sentiment.
By following these steps and best practices, you can create powerful AI prompt columns that help you analyze data, automate responses, and improve customer service. You can use AI Builder prompts and prebuilt prompt templates to save time and ensure high-quality results. Always provide enough context and test your prompts to get the best performance from your AI plug-in features.
Using Dataverse AI Columns
Inputting Prompts
You can start using dataverse ai columns by inputting natural language prompts that guide the AI to generate valuable insights. This process helps you create clear and specific instructions for the AI, ensuring that the suggestions you receive are relevant to your business needs. Follow these steps to input prompts effectively:
Sign in to Power Apps, Power Automate, or Copilot Studio.
Go to the AI hub and select Prompts. If you use Copilot Studio, navigate to Agents, then Tools, and add a new Prompt tool.
Enter a custom name for your prompt, such as "Customer Sentiment" or "Suggested Actions."
Write your prompt or choose a template that matches your scenario.
Add input variables by selecting Add content. Choose Text or Image/Document inputs to allow dynamic content at runtime.
Link a knowledge object to your Dataverse data if you want the AI to use business context for better suggestions.
Provide sample values for each input to help the AI understand the expected data.
Test the prompt to review the AI response and make sure it meets your requirements.
Save the prompt when you are satisfied with the results.
Tip: Always test your prompts with real data to see how the AI suggestions perform in your environment. This helps you refine your instructions and improve the quality of ai generated content.
Reviewing AI Output
After you input prompts and receive ai suggestions, you need to review and validate the AI output to ensure accuracy. This step is important because it helps you trust the responses to inquiries and make informed decisions based on ai columns. Here is a recommended process for reviewing AI-generated results:
Copy the AI-generated FetchXML query from the Copilot response by selecting the 'Show code' link.
Create a new cloud flow in Power Automate and add a 'List rows' Dataverse action.
Paste the copied FetchXML query into the 'FetchXML query' field under advanced options.
Save and test the flow manually to review the returned JSON results.
Examine the grouped fields and aggregated counts in the JSON output to verify if the data matches your query intent.
Edit and rerun the FetchXML queries directly in the code area to fine-tune results based on new criteria.
Follow prompt best practices: be specific, experiment with prompt phrasing, and provide feedback to improve AI output accuracy.
Address any Copilot error messages or known limitations as part of your validation process.
Note: Reviewing AI output helps you catch errors early and ensures that the ai suggestions align with your business goals.
Managing Data with AI
Using ai suggestions in Dataverse transforms how you manage and interact with your data. You can sort, filter, and visualize AI-generated results just like any other column in your tables. This makes it easy to find trends, spot issues, and take action based on real-time insights.
Sorting and Filtering: You can sort feedback by sentiment, filter for negative reviews, or group data by category. This helps you focus on the most important information.
Visualizing Data: Use built-in visualization tools to create charts and dashboards. For example, display the number of positive, neutral, and negative sentiments in a bar chart. Select a segment to filter the underlying data instantly.
Automation and Agentic Experiences: AI prompt columns enable automation and in-app agentic experiences. For example:
Add a sentiment analysis column to product reviews. The AI generates sentiment values based on review content, making it easy to track customer satisfaction.
Use a Document Processor Agent to automate document workflows, such as extracting data from invoices and routing them for validation.
Deploy a Customer Brief Agent to generate executive summaries from business databases, supporting meetings and decision-making.
Implement a Lead Manager Agent to triage and manage inbound leads at scale, using ai suggestions to prioritize follow-up actions.
Agents built with Copilot Studio can access your Dataverse data as a conversational knowledge base. They can answer questions, update records, and generate custom outputs like summaries or sentiment assessments. This unified approach connects structured and unstructured data, allowing agents to reason over your business information and deliver actionable ai suggestions.
Callout: Using ai suggestions in Dataverse helps you automate repetitive tasks, improve decision-making, and deliver faster responses to inquiries. You can rely on ai columns to provide real-time insights and streamline your business processes.
AI Suggestions for Formula Columns
Integrating Power Fx
You can use ai suggestions for formula columns to simplify how you build business logic in Microsoft Dataverse. This preview feature lets you enable "AI suggestions for formula columns" in your environment settings. After you turn it on, you can create a new formula column in the Maker Portal. Select "Get formula suggestions" and describe what you want in plain language. The AI will generate a suggested formula using Power Fx, which you can review and adjust to fit your needs.
For example, if you want to calculate "Days since Opportunity was Won or Lost," you can enter that as your prompt. The AI will provide a formula draft that you can copy, paste, and refine. You should always check the logic and syntax before using it in your production tables. This feature supports formulas that reference a single table and is currently available in US regions. You can troubleshoot and improve the ai suggestions by using tools like ChatGPT if you encounter errors.
Tip: AI Copilot in Power Platform helps you create and refine formulas using natural language. You do not need deep coding skills to automate your business logic.
Enhancing Data Workflows
AI suggestions for formula columns help you automate calculations and improve data workflows. When you use these suggestions, you reduce manual work and ensure data consistency. Formula columns can handle tasks like calculating total costs or expected delivery dates. These calculations can trigger workflows that notify your team or update records automatically.
Many organizations have seen measurable improvements. For example, automating formula columns reduced onboarding emails by 50% at AECOM. Coca-Cola Bottling Company United cleared a backlog of 50,000 orders with automated workflows, allowing one CRM agent to manage orders more efficiently. Integrating ai suggestions with tools like Copilot Studio speeds up time to market and boosts customer satisfaction.
You save time by letting AI generate formulas for you.
You reduce errors because the AI follows your instructions closely.
You improve workflow responsiveness by automating repetitive tasks.
Using ai suggestions for formula columns gives you a strong starting point for building business logic. Always review the suggestions to make sure they match your requirements and data structure.
Troubleshooting and Best Practices
Common Issues
When you work with AI prompt columns in Microsoft Dataverse, you may face some common issues. Sometimes, the AI might not return the expected results. You might see empty values or responses that do not match your instructions. This can happen if the prompt is unclear or if the knowledge source is missing. You may also notice delays in AI responses when your table has a large amount of data. If you see errors, check your permissions and make sure the AI features are enabled in your environment.
Tip: Always review your prompt instructions and knowledge sources if you notice unexpected results.
Validating AI Results
You should always validate the output from your AI prompt columns. Start by testing your prompts with real data. Compare the AI-generated results with what you expect. If you see mistakes, adjust your prompt or update your knowledge sources. Use sample data to check if the AI can handle different scenarios. You can also ask team members to review the results for accuracy. This process helps you maintain high data quality in your Dataverse tables.
Tips for Effective Prompts
To get the best results from AI prompt columns, follow these best practices:
Write clear and simple instructions for the AI.
Use examples in your prompts to guide the AI.
Reference the right columns and related tables for context.
Test your prompts often and update them as your data changes.
Monitor AI activity to spot trends or errors early.
Good prompt design leads to more reliable and useful AI suggestions.
You can unlock powerful insights in Dataverse by following these key steps:
Create custom prompts in AI Builder for your business needs.
Structure AI outputs using JSON for easy mapping.
Integrate prompt columns with Power Automate to update and process data.
Use expert tutorials to guide your setup.
Customizing prompts lets you tailor AI to your data, making workflows smarter and more secure. Try different prompt scenarios to see what works best for your business. For advanced learning, explore:
Document Processor and Lead Manager Agents
Power Platform connector SDK for external data integration
Experiment with AI prompt columns to transform your business processes and stay ahead with the latest AI tools.
FAQ
What is an AI prompt column in Dataverse?
An AI prompt column lets you use generative AI to create or analyze data in your Dataverse tables. You design prompts that guide the AI to generate text, classify feedback, or suggest actions based on your data.
Can I reference related tables in my AI prompts?
Yes, you can reference columns from related tables as knowledge sources. This gives the AI more context and helps it provide better answers. For example, you can use product details from another table to improve feedback analysis.
How do I test if my AI prompt works correctly?
You can use the "Test Your Copilot" panel to try your prompt with real data. Adjust your instructions until you get the results you want. Always review the AI output before using it in your business processes.
Are AI prompt columns secure for sensitive data?
AI prompt columns follow Dataverse security roles and permissions. Only authorized users can view or edit the data. Always review your security settings to protect sensitive information.
What are some common uses for AI prompt columns?
You can use AI prompt columns for sentiment analysis, feedback categorization, drafting email replies, or suggesting next steps. These columns help you automate tasks and gain insights from your data quickly.