The CFM Copilot, an AI-driven conversational assistant integrated within Sprinklr CFM, enables business users to effortlessly query survey data using natural language. This eliminates the need for advanced analytical skills or manual exploration of dashboards. Seamlessly embedded in both Analytics and Custom Dashboards, the Copilot provides highly confident and well-founded answers, which are consistently supported by actual survey responses.

The Copilot helps to reduce the complexity involved in extracting meaningful insights from survey data, particularly when multiple custom fields, filters, hierarchies, or surveys are in play. Instead of relying on specialized analysts or extensive dashboard exploration, users benefit from a self‑serve conversational interface that can interpret natural language questions and deliver accurate, verifiable insights. The goal is to enable intuitive data querying, ensure every response is grounded in actual survey inputs or visualized data, and support advanced capabilities such as trend identification, comparisons, root‑cause exploration, thematic analysis, emerging issue detection, and multi‑turn follow‑up interactions.

Business Use Cases

1. Eliminates manual analysis for common questions (trends, comparisons, distributions).

2. Bridges skills gap with a natural language, self-serve interface.

3. Grounded answers reduce risk of misinterpretation and build trust.

4. Faster decisions via concise, data-backed summaries + charts.

5. Context-aware multi-turn exploration (filters, time, segments, hierarchies).

This solution enables natural language–to–insight interaction with no specialized skills required, delivers grounded and auditable answers with sources displayed in a dedicated third pane, supports multi‑turn depth by preserving context across follow‑up questions for progressive analysis, and accelerates speed‑to‑insight by instantly rendering trend lines, comparisons, and distributions.

Examples

• Trend & Comparison for CX Scores
  Example query: “How has NPS changed across quarters for the Shell sites in the region of Arizona?”
  Delivers a trend summary with an NPS chart, along with suggested refinements such as “Compare with Texas” or “Show Tier 1 only.”

• Root Cause & Themes for Low Scores
  Example queries: “What are the top complaints from low CSAT responses?” or “Why did NPS drop in July for Tier 1 cities?”
  Applies relevant filters, surfaces themes, sentiment, and drivers, and offers follow‑ups like “Show top 3 themes by city.”

• Emerging Issues & Sentiment Monitoring
  Example queries: “Which new issues have appeared recently in open comments?” or “Has positive sentiment improved in the last 60 days?”
  Highlights emerging topics versus baseline, shows sentiment trends, and identifies impacted regions, sites, or channels.

  ​

  

Key Capabilities

• Convert natural language into valuable insights without needing any specialized skills.

• Verifiable and reliable answers are provided, with their sources displayed in a dedicated third pane.

• The depth of multi-turn interactions is achieved as follow-up exchanges maintain context, allowing for progressive analysis.

• Rapid insights are achieved through the immediate display of trend lines, comparisons, and distributions.

Prerequisites

You would need CFM Copilot permissions at the App Level.

​

Accessing CFM Copilot

The navigation steps are listed below:

1. Navigate to Spinklr Insights and then access Customer Feedback Management Persona App.

2. You can access Co-pilot in these two ways:

   1. Analytics Tab: First, access a Survey. From there, navigate to the Analytics tab to open Sprinklr Copilot.

      

   2. Custom Dashboards: In the left pane, proceed to Custom Dashboards. Then, choose a custom dashboard of your preference, and finally, open Sprinklr Copilot.

      

Key Functionalities

Copilot delivers contextual, conversational insights by understanding intent, grounding answers in survey data, and supporting interactive exploration with full transparency and explainability. The functionalities are listed below:

• Conversational input parsing and intent detection: Copilot understands natural language questions, identifies intent and key entities, and adapts to informal or incomplete inputs.

• Response generation grounded in available survey data: Generates answers strictly based on surveys, responses, metrics, and widgets available in your environment.

• Auto‑detection of relevant surveys and custom fields: Automatically identifies the most relevant surveys, questions, and custom fields to reduce manual setup.

• Multi‑turn clarifications for ambiguous queries: Asks targeted follow‑up questions when queries are vague to ensure accurate results.

• Support for drilldowns, filters, and follow‑ups: Allows iterative exploration through filters, drilldowns, and follow‑up questions that stay within the same data context.

• Visual and textual answer support: Presents insights using charts, statistics, and concise summaries for faster comprehension.

• Feedback, copy, and share options: You can provide feedback using thumbs‑up or thumbs‑down, copy responses, or share insights with stakeholders to support collaboration and continuous improvement.

  

• Source transparency through a dedicated pane: Copilot shows the exact surveys, responses, or widgets used to generate an answer in the third pane, giving you full visibility into the data backing each insight.

  

  

• Edge Case Handling: Copilot intelligently detects when your question spans multiple surveys with similar data, such as shared fields like Region or Site. Instead of guessing, it guides you to choose the right survey before proceeding, ensuring insights are always accurate, relevant, and trustworthy. This keeps your analysis focused and eliminates the risk of drawing conclusions from the wrong data set.

• Permissions & Security: Copilot respects your data access at all times and only surfaces insights from surveys you’re authorized to view. If a question requires comparing multiple surveys and you don’t have access to all of them, Copilot will clearly let you know and guide you to continue within your permitted scope, ensuring secure, transparent, and reliable analysis without interruptions.

• Multi-Survey Handling: When you ask comparative or aggregated questions across multiple surveys, Copilot automatically checks your access. If you have permission for all required surveys, it fetches and processes the data seamlessly. If not, Copilot clearly explains the limitation and guides you to continue the analysis within the surveys you’re allowed to access.

• Response Capabilities Summary: Copilot delivers clear, actionable insights through concise textual answers, supported by visualizations where relevant, with detailed breakdowns by key entities such as region, site, or custom fields. Each response is interactive, allowing you to drill down into details or refine the analysis with follow‑up questions and filters.

Query Types Supported by CFM Copilot

CFM Copilot is designed to answer reporting and analytical questions using survey response data. It can help you explore trends, compare metrics, summarize dashboards, and derive insights from survey responses. However, it does not perform actions or configuration changes within the platform.

Unsupported Query Types: Non‑Reporting (Action‑Based)

CFM Copilot cannot execute actions, modify configurations, or create reporting assets.

Examples of unsupported requests:

• Change distribution programs or modify business rules.

• Take case management actions.

• Create or edit dashboards, tabs, sections, or widgets.

• Configure reports or visualizations.

Sample queries

• "Create a new dashboard for Support CSAT".

• "Add a widget that shows CSAT trends by week".

• "Plot a bar chart for NPS by region on Tab 2".

Supported Query Types: Reporting & Analytics

1. Dashboard Summarization: Copilot can generate high-level summaries of dashboard components by analyzing the underlying survey data displayed in dashboards, tabs, sections, or widgets.

   Example

   "Summarize this dashboard."

   "Summarize this widget."

2. Survey Response Q&A: Copilot can answer questions based on survey responses, including supported survey question types and custom fields.

   Queries Referencing Dashboard Components

   You can ask questions about specific widgets, tabs, or sections.

   Examples

   "In the 'CSAT by Channel' widget, what is the average CSAT?"

   "Which region has the lowest NPS in the 'NPS by Region' widget?"

   Queries Without Referencing Dashboard Components

   You can also ask questions directly against the survey data.

   Examples

   "What is the overall CSAT average for the last quarter?"

   "How many bot responses did we receive last month?"

   "What is the average NPS by distribution channel?"

   Supported Analysis Types

   CFM Copilot supports a wide range of analytical queries, including:

   Analysis Type	Support
   Descriptive ("What is...?")	✅ Supported
   Comparative Analysis	✅ Supported
   Trend Analysis	✅ Supported
   Qualitative Text Summarization	✅ Supported
   Diagnostic / Root Cause Analysis	✅ Supported with guardrails

   Note: Diagnostic insights are exploratory in nature. Copilot highlights patterns and likely contributing factors but does not establish statistical causation.

3. Diagnostic and Root Cause Analysis: In Normal Mode, Copilot can help investigate changes in key metrics by identifying areas where performance shifts appear concentrated and surfacing related patterns in the data.

   Example queries

   "Why did NPS drop last month?"

   "Why did satisfaction decline in Tier 1 stores?"

   "Why did CSAT increase in Europe?"

   How Copilot Performs Diagnostic Analysis

   When a diagnostic query is submitted, Copilot follows a structured analysis approach:

   1. Identifies the metric change by comparing the requested period against a comparable previous period.

   2. Examines key segments such as region, store, hierarchy, distribution channel, custom fields, and customer profile attributes to identify where the largest changes occurred.

   3. Evaluates text analytics themes to highlight feedback topics that increased or decreased significantly.

   4. Analyzes driver questions, where available, to identify questions with significant score movement.

   5. Generates a diagnostic summary that consolidates these findings into a concise explanation of where the change appears most concentrated.

   Diagnostic Language Guardrails

   To ensure responsible and accurate analysis, Copilot avoids causal statements such as:

   "This caused the drop."

   "These drivers impacted NPS."

   "These factors led to the decline."

   Instead, it uses evidence-based language such as:

   "The largest decline occurred in..."

   "The change appears concentrated in..."

   "Customer feedback indicates an increase in..."

   Edge Cases

   Copilot handles special scenarios appropriately:

   Low sample size: Indicates when there is insufficient data to identify meaningful patterns.

   No clear concentration: States when changes are not concentrated within a specific segment or theme.

   Missing time period: Prompts users to specify the time period for analysis.

4. Custom Metrics — Supported: CFM Copilot can resolve and analyze explicitly referenced Custom Metrics in both Analytics and Custom Dashboards. Custom Metrics are treated similarly to standard metrics for supported reporting scenarios while respecting permissions, filters, RowScope, masking, and dashboard context.

   Supported Custom Metric Types

   Numeric

   Percentage

   Currency

   Date Time

   Time Difference

   Supported Query Scenarios

   Analysis Type	Example
   Descriptive	"What is the average Composite Experience Score?"
   Comparative	"Compare Completed Rate across regions."
   Trend	"How has our Custom NPS Index changed over the last quarter?"
   Segmentation	"Show Completed Rate by store tier."
   Drill-down	"Show Custom CSAT for underperforming stores in Region A."
   Relationship Analysis	"Is there a correlation between Completed Rate and NPS?"

   Custom Metric Resolution Rules

   1. Custom Metrics must be referenced by their exact names.

   2. Concept-based matching is not supported (for example, "completion percentage" will not automatically resolve to a metric named "Completed Rate").

   3. If multiple metrics match the query, Copilot asks for clarification.

   4. If a Custom Metric conflicts with a standard metric name, Copilot prioritizes the standard metric and requests clarification.

   5. If a referenced metric cannot be found, Copilot informs the user rather than generating inferred results.

   Scoping Behavior

   Analytics

   Custom Metrics are evaluated within the survey where they are defined.

   Existing Analytics filters remain in effect.

   Custom Dashboards

   Metric computations respect RowScope, dashboard filters, widget context, permissions, masking, and selected time ranges.

   Natural-language prompts cannot override a Custom Metric's configured survey scope or dashboard-level restrictions.