customer support response

Status: ready — not yet trained.

Traits

Empathy & Acknowledgment

Acknowledges the specific situation with genuine understanding ↔ Generic sympathy boilerplate or no acknowledgment at all

Whether the response demonstrates genuine understanding of the customer's situation and frustration — acknowledging the specific problem they described — versus generic "I understand your frustration" boilerplate or jumping straight to a solution without validating the experience.

Resolution Specificity

Concrete steps or definitive answer to the specific problem ↔ Vague suggestions or deflection to generic resources

Whether the response provides a concrete, actionable resolution — specific steps to fix the problem, a clear workaround, or a definitive answer — versus vague suggestions like "try again later" or "please refer to our help center" with no targeted guidance.

Expectation Setting

Clear next steps, timeline, and ownership ↔ No indication of what happens next or when

Whether the response sets clear expectations about what happens next — timelines, who will follow up, what the customer needs to do versus what the team will handle — versus leaving the customer uncertain about whether their issue is being tracked or when they'll hear back.

Personalization

References the customer's specific context and details ↔ Templated response disconnected from what was actually asked

Whether the response addresses the customer's specific context — their account state, the feature they mentioned, the exact error they hit — versus a templated reply that could be sent to anyone regardless of what they actually wrote.

Tone Calibration

Tone appropriate to the situation and severity ↔ Robotic, over-warm, or mismatched to the situation

Whether the tone matches the situation — professional warmth without being saccharine, appropriate urgency for serious issues, conversational without being flippant — versus robotic formality, excessive exclamation marks, or a mismatch between tone and severity.

About

Scores the quality of a customer support response.

What it measures

Whether a support reply actually helps the customer move forward. The model rewards responses that acknowledge the specific situation with genuine empathy, provide concrete resolution steps (not "please refer to our help center"), set clear expectations on timeline and ownership, personalize to the customer's actual context, and calibrate tone to match the severity of the issue.

Feed a single support response as input text. Works for email replies, chat messages, and ticket responses.

Limitations

• Optimized for customer-facing support responses in English. Internal escalation notes or ticket metadata will score low.
• Does not verify whether the suggested resolution is technically correct — only whether the response structure and communication quality are present.
• Proactive outreach messages (not responding to a customer issue) may not fit this model well; consider u22a8.retention-message instead.

Pairs well with

• u22a8.retention-message — when the goal shifts from resolving an issue to saving a relationship
• u22a8.crisis-comms — public-facing incident communication at a broader scale

Docs

• Tiers — categorical labels (Strong, Solid, Developing, Weak) assigned per trait
• Breaks — the per-trait trained boundaries between tiers