Emotional Sensing

Description

Detects emotions, intent signals, and business indicators in text — going beyond basic sentiment to identify signals like enthusiasm, hesitation, buying intent, frustration, and churn risk. Paste customer messages, support tickets, reviews, or sales conversations and get each signal scored on a 0-1 scale with severity levels and action recommendations. Supports single statements or batches of up to 10.

Core Capabilities

• Multi-dimensional detection: emotions (joy, anger, confusion), intent (purchase intent, selling opportunity), and business signals (churn risk, frustration, satisfaction, loyalty risk, urgency)
• 4-level severity scoring: Low (0-0.3), Moderate (0.3-0.6), High (0.6-0.8), Critical (0.8-1.0)
• Batch processing up to 10 statements with cross-statement pattern analysis and high-risk alerts
• Action recommendations prioritized by business impact

Good Query Examples

• "Analyze this renewal email from our biggest client: 'We appreciate the partnership but need to reassess priorities for next quarter. Let's schedule a call to discuss options.'" (subtle churn signals in polite language)
• "Score purchase intent in this inbound lead message: 'We're evaluating solutions for Q2 rollout across 3 departments and your platform came up in our shortlist'"
• "Batch analyze these 8 support tickets and flag the highest-urgency ones: [paste tickets, one per line]"
• "Check the tone of this draft before I send it to the client: 'Per our last conversation, I wanted to follow up on the outstanding items that require your immediate attention'"

Bad Query Examples

• "ok" or "fine" — text too short for reliable signal detection. Fix: provide meaningful text with enough context
• "Analyze the sentiment on Nike's Instagram page" — the agent analyzes text you paste in, it cannot fetch external content. Fix: paste the actual text, or use Instagram Analysis for social account analysis
• "How do millennials feel about sustainability?" — this asks about population-level attitudes, not specific text. Fix: use GWI Spark for audience sentiment

Use-Case Ideas

• Triage support tickets by urgency and frustration level
• Identify at-risk customers from their communications before churn
• Detect buying signals in sales conversations for pipeline prioritization
• Tune message tone before sending — check if your draft sounds enthusiastic, hesitant, or pushy

Limitations

• Text-only: no voice, images, or external content fetching — you must paste the text
• Maximum 10 statements per batch
• Point-in-time analysis only (no sentiment tracking over time)
• Scores are AI-generated probabilities, not certainties
• Subtle or mixed emotions may be simplified into primary categories
• Intent detection depends on contextual cues and may need domain-specific phrasing

Input/Output Spec

• Input: Text to analyze (single or up to 10 statements, one per line), optional additional signal categories
• Output: Markdown report with scored signals (0-1), severity levels, business impact alerts, and prioritized action recommendations. Batch mode adds cross-statement statistics and high-risk alerts.

Disambiguation Signals

• If the query asks "what do people think about X" or "how do people feel about X" → this is an opinion-gathering request, not text analysis. Route to Ask the Crowd: Opinion. However, if the query includes existing text (e.g., customer comments, survey responses, support tickets) and asks to detect the emotional signals in it → that's Emotional Sensing.
• If the query provides a statement and asks to understand potential reactions to it → route to Ask the Crowd: Opinion (generates written feedback from 5 personas)
• If the query asks to rate agreement with a statement on a scale → route to Ask the Crowd: Survey (1–5 Likert scale from 100 respondents)
• If the query asks to pick the best option among 3 alternatives → route to Let the Crowd Decide (40 virtual voters)
• If the query asks about general audience attitudes or demographics → route to GWI Spark
• If the query says "analyze the tone of…" or "what emotions are in this text" and includes or pastes actual text → this IS Emotional Sensing