Compare Two Customer Support Prompts
A 'be nice and helpful' support prompt against a policy-bounded one — compared on risk, because support prompts fail on risk first.
Overview
Support prompts have a failure mode the other categories don't: a confident reply that promises something policy can't deliver. That makes risk the first dimension to compare, before tone or speed. This resource loads a friendly-but-unbounded prompt against one with an explicit policy boundary and escalation rule, so the comparison shows how 'be nice and fix it' scores when the question is what could go wrong.
Workflow
-
Compare with Risk & Ambiguity focus
The loaded pair makes risk the deciding dimension — exactly how support prompts should be judged.
-
Read A's risk gaps
No policy boundary, no escalation rule, and 'whatever it takes' — each one is a promise an agent can't keep.
-
Note what B controls
Verification before promises, an escalation trigger, a length cap, and a next-step ending — the anatomy of a safe reply prompt.
-
Test your own reply prompt
Paste your current support prompt as A against B's structure. If risk is the losing dimension, fix that before anything else.
Why This Works
- Risk-first comparison matches how support prompts actually fail in production — one over-promise costs more than ten awkward tones
- An explicit policy boundary converts 'be helpful' from a liability into a deliverable
- Escalation rules in the prompt mean the hard cases route to humans by design, not by luck
Best for
- Support teams putting AI-drafted replies in front of real customers
- Prompts for refund, billing, and account disputes where wrong promises cost money
- Standardising reply prompts across agents with different instincts
Not for
- Building a support system prompt from scratch — that's the System Prompt Generator's job
- Tone-only polish on a single reply — no comparison needed for that
Use cases
- Choosing which support reply prompt becomes the team standard
- Auditing an existing support prompt for over-promising risk before agents rely on it
- Showing a support lead why 'do whatever it takes' is a liability in a prompt