Working Guidelines

Ask Me Working Guidelines

1. Lumenore Ask Me Can’t Explain Data Schema or Structure

What does this mean?
The AI can’t tell you how your dataset is organized, column names, data types, or relationships between tables.

Don’t ask:

• “How many columns are in the Sales table?”
• “What type of data is in the Customer ID column?”

Ask:

Check the data schema manually in the platform’s data settings. There, you’ll find:

• Column names
• Data types (number, string, date)
• Field-level details

Tip: If you’re unsure of column names, open the dataset before querying the AI.

2. Column Type Detection Is Not Automatic

What does this mean?
Lumenore Ask Me does not automatically detect column types, especially when the data is ambiguous. By default, it may interpret fields like Rank, Age, or Year as numeric measures, even when they should function as categorical dimensions.

Example:
If the Year column contains values like 2020, 2021, 2022, the AI might treat it as a numeric field rather than a time-based dimension.
As a result, queries like “Sales by year” may not work as expected.

Tip: To ensure accurate insights, follow data readiness best practices:

• Define all fields correctly (dimensions vs. measures).
• Use the Schema Manager to set column types explicitly.
• Label time-based columns like Year as a dimension, not just a number.

3. Basic Chart Customization  

What does this mean?
Lumenore Ask Me supports basic visualizations (like bar, line, and pie charts), but it does not support advanced customizations, such as:

• Customized data labels
• Customized axis labels
• Font styles
• Multiple visualization changes in one query
• Advanced conditional formatting
• Advanced visualization support

Don’t ask:

• “Make this chart blue and label the X-axis as ‘Months’.”
• “Give values in millions.”
• “Change precision of data label to 2 decimals.”
• “Show sales by month and change it to a bar chart.”

Ask:

• “Show sales by month as a bar chart.”
• “Change this into a bar chart.”
• “Change the color of bars to red.
• “Hide X axis labels”
• “Hide legend”

4. AI Doesn’t Respond to Casual or Non-Data Queries

What does this mean?
Lumenore Ask Me is focused only on business and data queries. It won’t respond to unrelated or casual requests.

Don’t ask:

• “Tell me a joke.”
• “Who won the match yesterday?”

Ask:

• “What were the top 5 products sold last month?”
• “Show customer churn by region.”

Tip: Use the AI only for data analysis tasks.

5. Lumenore Ask Me Doesn’t Support Complex, Multi-Step Logic in a Single Query

What does this mean?
Lumenore Ask Me is designed to process one logical step at a time. It cannot handle queries that combine multiple conditions or require nested logic in a single request.

Example 1:

Don’t Ask:
“Compare sales in regions where sales are greater than 500K and break it down by  category.”

Instead, Break It Down Step-by-Step:

1. “Filter regions where sales > 500K.”
   1. “Add category-level data to the regions already filtered.”

Example 2:

Don’t Ask:
“Compare sales and expenses by region and show it in line chart.”

These queries contain too many actions in one sentence—filtering, aggregating, comparing, and visualizing—which can lead to incorrect or no results.

Instead, Break It Down Step-by-Step:

1. “Show total sales.”
2. “Add total expenses.”
3. “Add region.”
4. “Change it to line chart”

Tip: Build complex queries gradually. Use simple, focused instructions in sequence to guide the AI through each part of the analysis.

6. AI Remembers Only 3 Previous Questions (Short-Context Length)

What does this mean?
Lumenore Ask Me has a short context window; it remembers only your last 3 queries. Anything beyond that gets forgotten.

Good usage:

1. “Show 2023 sales.”
2. “Compare with 2022.”
3. “Break down by region.”

If you ask a 4th question like:

• “Now show by channel.”
  → It may forget that you were comparing 2023 and 2022.

Fix: “Show 2023 and 2022 sales by channel.”

Tip: Repeat important context in new questions.

7. Always specify column names when applying filters

What does this mean?
Since data is not shared with AI, it’s important to mention the column name when applying filters to clarify which field the filter should apply to. Additionally, ensure the filter value matches exactly with the data.

Don’t ask:

• “Give me sales for John”

Ask:

• “Give me sales where Employee = ‘John’”

8. Filters Must Match Dataset Values Exactly

What does this mean?
Lumenore Ask Me requires exact matches between your filter values and the data in your dataset. If the value you enter doesn’t exactly match what’s stored, the AI won’t return any results.

Don’t ask:

• Ask to filter by “United States” when the dataset has “USA”.

Ask:

• Use the exact spelling and format: “USA”.

Note: If the filter value doesn’t match exactly, the AI may return a “No data found” message, even if relevant data exists.

Tip: Always check your dataset for the correct spelling, format, or abbreviation before applying filters.

9. AI Gets Confused by Similar or Vague Column Names

What does this mean?
Lumenore Ask Me may get confused when your dataset contains columns with ambiguous or similar names, such as “Cat1,” “Cat2,” “Total A,” or “Total B.” It may not know which column to reference, leading to errors or incorrect results.

Don’t ask:

• “Show sales for Cat1”
  The AI might choose the wrong column or fail to return a valid result.

Ask:

• Rename columns to clear, descriptive names like “Category_Electronics” and “Category_Furniture.”

 Tip:

• Always use meaningful column names in your dataset.
• Using the Schema Manager, add column descriptions under the “Description” field to help the AI better understand the context.

10. Limited Support for Advanced Chart Types and Features

What does this mean?
Lumenore Ask Me currently supports only basic visualizations and does not support advanced chart types or formatting features such as:

• Dual-axis charts
• Map charts
• Bubble charts
• Group tables
• Pivot tables
• Filters or slicers
• Waterfall charts
• And other advanced visuals

Don’t Ask:

• “Show sales by country in a map chart.”

Since map charts aren’t supported, the AI will default to a basic bar chart or return an incorrect response, as shown in the screenshot below:

Ask:

• “Show sales by country.”

Lumenore Ask Me will automatically select the most appropriate chart type for the data.

Tip: Let the AI recommend the best chart format based on the question.

11. LQL variables are not supported in Ask Me

LQL variables are not supported in the Ask Me module, so any variables defined using Lumenore Query Language will not function within this environment. To ensure compatibility and accurate query execution, users should create and use SQL variables instead. These are fully supported in Ask Me and allow seamless interaction when users input queries using natural language.

Best Practices for Getting the Most Out of Lumenore Ask Me

1. Ask Simple and Clear Questions

Avoid ambiguity, use full words, and be specific.

Examples:

• “Show revenue for 2023 by region.”
• “Display Q3 sales for Product A as a line chart.”

Don’t ask:

• “Give me the numbers for stuff.”

Tip: Always use actual column names from your dataset.

2. Refine Your Question with Specific Details for More Accurate Responses

What does this mean?
If the AI’s answer is too broad, vague, or incorrect, try to rephrase your query with more context or precision.

Don’t Ask:

• “What about profit?”
  This is too general and may lead to incomplete or irrelevant results.

Ask:

• “Show profit margin for Electronics in March 2023.”

If the AI does not calculate it correctly, you can guide it further by specifying the formula. AI can generate simple formulas, but for complex formulas, it is recommended to create an SQL variable in the data dictionary.

“Profit margin is calculated as (Profit ÷ Sales) × 100.”

Tip:
The more clearly you define your question, the better the AI can interpret it. Providing formulas or additional context helps the AI deliver more accurate and meaningful insights.

3. Customize Step-by-Step

Break each part of your request into separate steps.

Example:

1. “Show sales by product category.”
2. “Now show it as a bar chart.”
3. “Filter by Q2 2023.”

This ensures accuracy and reduces AI errors.

4. Avoid Using Pronouns Like “I,” “My,” or “We”

The AI doesn’t know personal context.

Don’t Ask:
“Show my team’s numbers.”

Ask:
“Show sales for the Western Team from the Team column.”

5. Overriding the Master Agent for Specific Queries

What does this mean?

If a user asks a question like “What is the sales trend?”, the Master Agent may route the query to the DS Agent by default. However, if the dataset already contains a column named “Sales Trend”, the ideal agent to handle the query should be the NLQ Agent.

Don’t Ask:

• “What is the sales trend?”
  This could cause the query to be routed to the wrong agent (the DS Agent) and may not give the correct result.

Ask:

• “What is the sales trend using the NLQ agent?”
  To ensure the correct agent processes the query, manually select the NLQ Agent from the dropdown before submitting the question. This will route the query appropriately and return the correct result using the existing column.

Summary

Troubleshooting

1. When Something Goes Wrong
If the AI fails to process your request or returns an error, try the following steps:

• • Clear the context by typing clear context to start fresh.
  • Ensure your schema is published and properly configured.
  • Check your AI model – upgrade to GPT-4o for complex queries.
  • Verify column definitions (e.g., distinguish between measures and dimensions).
  • Review your AI credit usage to ensure you haven’t exceeded limits.

2. No Data Found
If the AI fails to process your request or returns a “No Data Found” message, try the following steps:

• • Ensure the filter is applied to the correct column.
  • Verify that the value you’re filtering (e.g., region = “abv”) exists in the dataset.
  • Check for typos or case sensitivity in the filter value.
  • Remove or simplify filters to rule out conflicts from multiple conditions.
  • Preview your dataset to confirm it’s populated and correctly refreshed.
  • Review the dataset schema to validate field names and data types.
  • Confirm the dataset is connected and published in the current workspace.

3. Visualization Limitations for Data Science Queries

When a user asks a Natural Language Query (NLQ) query, such as “Show sales by region,” the platform will first display the analysis, and then the user can interact with the visualization agent. For example, after seeing the sales by region chart, the user can use the visualization agent to change the bar colors to red, which will modify the chart’s appearance.

However, if the user asks a data science-related query, such as “Trend analysis for sales order date month,” the system will display the corresponding analysis. But, if the user attempts to run a visualization query, such as “Change bar colors to red” on a data science chart, an error message will appear, stating that the platform currently does not support visualization changes on data science charts.

Follow the steps below to avoid errors:

• Clear Context: Type “clear context” to reset and start fresh.
• New Analysis: Begin a new analysis to apply visualizations.
• Avoid Data Science Queries: To use the visualization agent, avoid data science queries, as they do not support visualization changes.