11. Making the Most of Infinity's AI-Powered Features
Here's a detailed section on making the most of Infinity's AI features in a GitBook-compatible format:
Infinity’s AI is designed to assist users in exploring, understanding, and analyzing data effectively. However, AI thrives on context—the more context you provide, the better its performance. Below are key ways you can enhance AI’s capabilities within Infinity.
1. Adding Table and Data Source Descriptions
AI performs best when it understands the structure and meaning of your data. While Infinity automatically identifies common data sources like Salesforce, Google Analytics, or YouTube, other data sources require additional context to improve AI accuracy.
Provide a Description for the Data Source
Explain what the data source is used for.
Describe the type of data it contains.
Highlight any nuances in the data (e.g., whether it includes historical data, real-time updates, or aggregated metrics).
Enhance Table-Level Descriptions
Clearly describe the purpose of each table within a data source.
Define relationships between tables, primary and foreign keys, and dependencies.
Specify column details, including data types, enumerations, and possible values.
Example: If you ask the AI to generate a chart showing "Sales in North America," the AI must understand whether the region column contains "North America," "NA," or some other format.
To simplify this process, use the Data Source Wizard, which walks you through these steps and helps capture all relevant details.
2. Optimizing AI Assistance for Charts
Each chart in Infinity has an AI-powered bot that assists users in understanding data. To maximize its effectiveness, provide additional context at three key levels:
1. Chatbot Prompt
Define key terms and acronyms that users might use in queries.
Example: If a chart displays sales data, users may ask about CAGR (Compound Annual Growth Rate). Adding this explanation ensures the bot understands and responds correctly.
Include details about data nuances, such as specific calculation methodologies, custom metrics, or segmentation.
2. Summary Prompt
When a user requests a summary, AI will reference these instructions.
Specify focus areas, such as highlighting trends, anomalies, or specific business insights.
Example: "If summarizing this chart, focus on year-over-year sales growth, key drivers of change, and regional performance variations."
Define the level of detail required—whether it should be concise, technical, or high-level.
Customize formatting if needed, such as bullet points, tables, or structured text.
3. Delta-Feed Prompt
Infinity’s platform can monitor material changes in data.
Define what constitutes a material change for your business.
Example: "For Customer A, a 2% drop in sales is significant, whereas for Customer B, a 10% change matters."
Specify any thresholds or key areas AI should watch.
Use the Delta-Feed Wizard to structure this information effectively.
3. Improving AI Query Responses
The AI-powered bot helps users interpret charts by answering questions, but it relies on clear and detailed queries. To improve its effectiveness:
Phrase questions clearly, providing as much context as possible.
Include specific examples of the types of responses you expect.
Remember that the bot’s knowledge is limited to the data and context provided—it does not have external knowledge beyond the dataset.
4. Providing Feedback for Continuous Improvement
If you notice inaccuracies or limitations in AI responses, provide feedback. This helps refine the model over time, improving its ability to assist users. Key feedback mechanisms include:
Reporting incorrect responses or misinterpretations.
Suggesting refinements for prompts and summaries.
Highlighting missing context that could enhance AI accuracy.
By actively providing context, structuring prompts effectively, and refining AI outputs, you can ensure that Infinity’s AI delivers highly relevant and insightful analysis tailored to your business needs.
Maximize the AI’s effectiveness within Infinity by adding context, structuring prompts, and providing feedback. 🚀
Last updated