Leveraging CFM Copilot for Analyzing CFM Dashboards
Updated
CFM Copilot is an AI-powered conversational analytics assistant embedded in Sprinklr Customer Feedback Management (CFM). It allows you to query survey data in natural language without manually building widgets, changing filters, or creating complex reporting formulas.
You can use CFM Copilot to identify trends, compare results, summarize feedback, investigate changes in customer experience metrics, and translate survey insights into actionable next steps.
Democratize survey analytics by enabling business users to find insights without depending on reporting teams.
Accelerate time to insight by replacing manual cross-tabulation and formula drafting with direct answers.
Drive action from feedback by connecting scores, comments, segments, and trends to clear next steps.
Tip: Ask a specific question and include the metric, segment, and time period whenever possible. For example, “How has NPS changed quarterly across Arizona locations?”
Where Can You Use CFM Copilot?
CFM Copilot is available from individual survey analytics and multi-widget custom dashboards.
Survey Analytics: Open CFM Copilot from the top-right header of an individual survey’s Analytics tab.
Scope: Automatically limited to the active survey.
Custom metrics: Supports survey-specific metrics such as Completion Rate, Custom CSAT, calculated CSAT, Composite Experience Score, and Custom NPS Index.
Controls: Expand or collapse the panel, copy response transcripts, provide feedback, and review grounded sources.
Custom Dashboards: Open CFM Copilot from the global action bar of a multi-widget CFM Custom Dashboard.
Scope: Respects active dashboard filters, date ranges, tabs, configured dimensions, and RowScope.
Cross-widget analysis: Compares and correlates metrics across widgets, segments, tiers, and categories included in the dashboard.
Note: Dashboard context and RowScope define the analysis boundary. Copilot does not silently substitute another survey or expand beyond the data you are permitted to access.
Permissions and Governance
User entitlement: You would need CFM Copilot permissions at the App Level.
.png "CFM Copilot Permissions (1).png")
Custom metric governance: Follows existing permissions in Reporting-Setting-Custom Metrics, including metric visibility, dimension restrictions, survey access, RowScope, and masking rules.
AI killswitch: Administrators can enable or disable CFM Copilot tenant-wide through AI+ Studio Feature Management. Disabling it immediately stops all runtime LLM calls for the partner environment.
Operating Modes
Capability | Quick Answer | Deep Research |
Primary use | Quick facts, single-metric lookups, and point-in-time checks. | Strategic reviews, diagnostic investigations, and root-cause analysis. |
Output | A concise, focused answer. | A structured brief with a summary, tables, drivers, and recommendations. |
Statistics | Descriptive statistics such as mean, median, percentage, and count. | Advanced analysis such as correlation, chi-square tests, ANOVA, regression, and driver analysis. |
Comments | Cost-aware sampling for large comment volumes. | Theme clustering, sentiment shifts, and representative quote extraction. |
Investigation style | User-driven follow-up questions. | Multi-step analysis that surfaces anomalies and segment outliers. |
Standard Mode / Quick Answer
Use Quick Answer for targeted operational questions that require a direct response without additional narrative.
Example: “What is our Email CSAT for this quarter?”
Expected output: The calculated score, response count, and date boundary.
Deep Research
Use Deep Research for broad questions that require a complete, decision-ready analysis.
Example: “How has NPS performed over the past two quarters, and what is driving the changes?”
Expected output: An executive summary, period-over-period trends, segment breakdowns, driver analysis, customer comment themes, grounded sources, and prioritized actions tied to evidence.
Trust, Scope, and Data Integrity
Strict survey and dashboard scoping: Copilot works within the selected survey or active dashboard, including partner, filter, date range, tab, widget, and RowScope context.
Data-backed metrics: Numbers, percentages, and sample counts are calculated from retrieved survey records. Low sample sizes are identified rather than overstated.
Defensible statistical claims: Copilot distinguishes correlation from causation and avoids unsupported causal claims.
Security and governance: Partner isolation, field masking, redaction, survey-level access controls, and RowScope are enforced.
Traceable sources: Responses include grounded sources that help you inspect the questions, widgets, records, and counts behind an answer.
Business-readable language: Answers use customer-facing labels and calendar dates instead of backend keys or raw JSON.
Best Practices
Ask clear, specific questions.
Include a time period when asking about trends or metric changes.
Use the exact configured name when referring to a custom metric.
Use follow-up questions to refine filters, segments, or time periods.
Review grounded sources to verify the data used for each response.
Treat diagnostic findings as evidence-based patterns, not proof of causation.