Scorecards drive how the AI grades every call your team takes. This guide covers how to write scorecard items that score accurately and are easy to coach from.
Prompts should be very short
3 sentences max. The scoring is categorical (Yes / No / N/A, did it happen or not), so long prompts add nothing. Use simple plain English; cut filler and boilerplate. Short plain prompts are easier to maintain, debug, and coach from.
One behavior per step
Split different behaviors into different steps. A single Yes/No on a two-part behavior is ambiguous and you cannot coach from it.
All items are scored together in one AI call
Every item can see the others, so items can reference each other if you need it.
Use quotes only if you want verbatim
A quote tells the AI to require those exact words. Sometimes that is exactly what you want, for example a required compliance disclosure. Otherwise describe the behavior in plain English and skip examples; the AI pattern matches aggressively and will over-index on any example you give it.
N/A is always an option on every item
There is no checkbox. If you want N/A used predictably, say when to use it in the item prompt. N/A is neutral, not a penalty: an N/A item drops out of the score entirely.
The AI only sees the conversation
It gets the transcript plus basic context (company name, agent name, call direction). It cannot score anything that happens outside the call, like “logged the lead in the CRM” or “dispositioned the call properly.”
Weights are applied after scoring, not by the AI
Give more weight to items that matter more. But do not write “this step is very important” in the prompt; the AI never sees weights and that sentence just adds noise.
Routing
Many customers do well with an appointment setting scorecard, a confirmations scorecard, and a generic scorecard for everything else. Set up call tags for each and then a scorecard routing rule. Know how routing behaves:
- A rule matches only if the call has all of the rule’s required tags (AND, not OR).
- A campaign match outranks tag matches, since the campaign comes from the phone system while tags are AI-assigned.
- Tags are assigned by AI, so routing quality depends on tagging quality. If calls land on the wrong scorecard, check the tags first.
- If no rule matches, the call falls back to the default scorecard for its direction (inbound or outbound).
Minor things
Write like a human. Use as few words as necessary to express what you want, no more, no less. Proofread each prompt to make sure you did not accidentally add conflicting instructions.