Context Settings Capabilities in Kyvos

How it helps

1. Business purpose and usage: Describing the business purpose and usage of the Semantic model or AI Space helps AI to create more precise and relevant responses.

2. Domain-aware: AI understands the specific business domain or industry (such as retail, finance, healthcare, or supply chain) and tailors its analysis accordingly.

3. Focused on key performance indicators (KPIs): Business Context identifies the Key Performance Indicators (KPIs) most critical to evaluating business performance. It tells the AI which metrics to prioritize when analyzing data.

4. Aligned with business goals and decision-making needs: This ensures that responses are not just data-driven, but business-driven.

Sample screens

Query instructions settings

The Query Instructions feature of AI settings is designed to guide how natural language queries in Kyvos Dialogs are interpreted and translated into executable queries.

It essentially acts as a rule layer on top of the semantic model, helping the AI understand user intent more accurately and respond in a way that aligns with business expectations.

These rules include:

• Business defaults

• Cross‑entity relationships

• Time logic & comparison logic

How it helps

1. Defines how to interpret business terms: Query Instructions let you specify how certain words or phrases should be understood.
   For example:

   • Revenue: Net Sales measure

   • Active customers: customers with transactions in the last 30 days

This removes ambiguity in user queries.

2. Controls default assumptions: When users ask incomplete questions like “Show sales”, Query Instructions can define:

   • Default time period (e.g., current month or year)

   • Default aggregation (sum, average, etc.)

So, the system always returns meaningful results.

3. Standardizes ambiguous keywords
   Terms like top, best, growth, and performance are subjective.
   You can define:

   • “Top” = highest by revenue

   • “Growth” = % change vs previous period

This ensures consistent output across users.

4. Improves time intelligence handling
   It helps interpret phrases such as:

   • “Last year”

   • “Month-to-date (MTD)”

   • “Year-to-date (YTD)”

by mapping them to correct date filters and logic.

5. Guides ranking and filtering logic
   For queries like “Top 10 products”, Query Instructions define:

   • Which metric to rank by

   • Sorting order

   • Any default filters

6. Enhances conversational accuracy
   In multi-step queries, it helps maintain context and ensures follow-up questions are interpreted correctly.

7. Reduces incorrect or inconsistent results
   Instead of AI guessing, Query Instructions enforce predefined logic, improving trust and reliability.

Examples

Example -1

User asks: “Show top performing regions last year”

• Semantic model identifies Region and Revenue

• Query Instructions defines:

• Business Context Rules: If a user asks for Top N, Top performing, Highest N items and does not specify the value of N, then we automatically use N = 10.

• Query Formulation Rules: Include important KPIs like Revenue and Profit in the generated query.

Result: Accurate, business-aligned output without manual intervention.

Example 2

User asks: “Show year over year performance of Northen region”

• Semantic model identifies Region and Revenue

• Query Instructions defines:

• relative year
  Logic: Always build the “relative year” as 2024.

• **year over year**

Triggers: YoY, year over year, year-over-year

Break down context: Compare two sequential years. Subtract the measure values for two sequential years within an extra column called “Difference”. If not mentioned, then always use the relative year compared to (relative year - 1 year).

Logic: Always use the same measures. Filter the query using members specified by the command or question.

Result: Correct result and aligned with business logic.

Sample screens

Summary instructions settings

The Summary Instructions feature in AI Settings lets you define custom instructions or context that the AI should follow when generating natural language summaries. This feature helps by making AI-generated summaries more useful, consistent, and aligned with business needs.

How it helps

1. Makes summaries business-relevant
   Without instructions, AI may give a generic description of data.
   With Summary Instructions, you can tell it:

   • what metrics matter (revenue, margin, growth)

   • what to highlight (trends, anomalies, comparisons)

Summaries focus on what your business actually cares about.

2. Ensures consistency across users
   Different users might ask similar questions but expect different styles.
   Summary Instructions enforce:

   • a fixed tone (executive, analytical, simple)

   • a consistent format (bullets, short summary, insights first)

Everyone gets uniform, professional summaries.

3. Improves decision-making
   By guiding AI to:

   • highlight anomalies

   • call out risks or opportunities

   • focus on KPIs

Summaries become actionable, not just descriptive.

4. Reduces ambiguity in AI output
   AI can interpret data in multiple ways. Instructions remove that ambiguity by telling it:

   • what perspective to take (finance, sales, operations)

   • what level of detail to use

More accurate and relevant insights.

Example

Summary Instructions:

“You are analyzing a retail business. Focus on revenue, profit, and growth trends. Highlight any significant increase or decline. Provide a concise executive summary with key insights.”

Output will:

• Talk about revenue & profit

• Highlight trends (up/down)

• Be short and business-focused

Sample screen

Verified queries settings

A Verified Queries feature of AI settings is designed to provide pre-approved, trusted query patterns and deterministic queries that the AI can use while responding to high‑value business questions in Kyvos Dialogs.

When a user asks a question, the system semantically searches for the most relevant Verified Queries, retrieves the closest matches, and uses them to guide SQL generation.

Verified Queries make Kyvos Dialogs more reliable, consistent, and business-approved by grounding AI responses in pre-validated query logic rather than relying entirely on dynamic generation.

How it helps

1. Ensures accurate and trusted responses: Verified Queries are created and validated by BI or data experts. When similar user questions are asked, the AI can reuse or adapt these queries—ensuring results are always correct and aligned with business logic. This ensures standardized answers across recurring KPI scenarios

2. Acts as a reference for the AI: Instead of generating queries purely from scratch, the AI learns from these predefined queries. This improves interpretation and reduces the chances of incorrect query generation.

3. Handles complex business scenarios: Some calculations or logic can be complicated (e.g., cohort analysis, custom KPIs). Verified Queries capture these complexities, so users can simply ask questions without needing to understand the logic.

4. Improves consistency and reliability across users: Different users may ask similar questions in different ways. Verified Queries ensure that all of them get consistent results based on the same approved logic.

5. Boosts performance and efficiency: Since these queries are already tested and optimized, the system can execute them more efficiently compared to generating entirely new queries each time.

6. Supports natural language variations: A single Verified Query can support multiple phrasings, such as:

   • “Sales by region last year”

   • “Regional revenue for previous year”

The AI maps to the same verified logic.

7. Builds user trust in AI responses: As outputs are based on validated queries, business users gain confidence that the data is reliable.

Example

A Verified Query is defined for:
“Top 10 customers by revenue in the last quarter”

If a user asks:
“Who are my best customers recently?”

The AI maps this to the verified query and returns accurate results using approved logic.

Sample Screens

Columns Metadata settings

Columns Metadata in the AI settings helps the system better understand individual columns (fields), enabling Kyvos Dialogs to interpret user queries more accurately. It provides detailed context for each column, enabling the AI to map natural language to the correct data elements and generate more accurate SQL.

How it helps

1. Adds descriptive context to columns
   You can define clear descriptions for columns, such as:

   • rev_amt → “Total revenue after discounts”

   • cust_flag → “Indicates active customers”

This helps the AI understand what each field actually represents.

2. Improves mapping of user queries
   When a user asks: “Show total revenue”, the AI uses column metadata to identify the correct field, even if the actual column name is technical.

3. Supports synonyms
   You can associate multiple business terms with a single column in Tags parameter:

   • Revenue = Sales = Net Sales

   • Customer = Client = Buyer

This helps Kyvos Dialogs understand different user phrasing without confusion.

4. Reduces ambiguity and errors
   If multiple columns seem similar (e.g., gross sales vs net sales), metadata helps the AI distinguish between them and choose the correct one.

5. Enhances query generation quality
   With better column understanding, the system generates more accurate and optimized queries.

6. Improves user experience in Dialogs
   Users don’t need to know schema details. They can ask questions naturally, and the AI uses column metadata to interpret correctly.

Example

User asks: “Show active customers by region”

·        Columns Metadata defines:

o   “Active customers” → customer column with specific flag/logic

o   “Region” → geography dimension  

Result: Accurate mapping and correct output.

Sample Screens