Free QA scorecard templates.
Three scorecards you can copy and use today: one for AI agents, one for compliance-heavy calls, and a general auto-QA card. Read every line on the page. Email yourself the editable copy if you want it.
Read every line before you take it.
Three scorecards, previewed in full on the page. Pick the one closest to your work, adapt the weights to your priorities, and use it. The editable copy is one email away if you want it.
| Dimension | What it checks | Weight |
|---|---|---|
| Resolution | Did the agent solve the customer's actual problem, not just end the conversation. | 20 |
| Containment quality | Was it genuinely resolved, not deflected, looped, or quietly abandoned. | 15 |
| Groundedness | Were the agent's statements true, on policy, and free of invented facts. | 20 |
| Compliance | Required disclosures present, identity verified, data handled correctly. | 15 |
| Empathy and tone | Acknowledged the customer and read sentiment, rather than answering a form. | 10 |
| Brand voice | On brand, plain, no filler or robotic padding. | 10 |
| Action accuracy | Correct actions taken: the right refund, update, or handoff. | 10 |
| Total | A passing conversation clears your threshold across all seven. | 100 |
| Check | What it checks | Weight |
|---|---|---|
| Identity verification | Right-party contact confirmed before any account detail is shared. | 20 |
| Required disclosures | Recording notice, and the debt-collection disclosure where it applies. | 20 |
| PCI handling | No full card number read back, repeated, or captured in the clear. | 20 |
| Consent and contact rules | Honors do-not-call status and permitted contact times. | 15 |
| Prohibited statements | No threats, misrepresentation, or false urgency. | 15 |
| Complaint handling | Complaints recognized, logged, and routed correctly. | 10 |
| Total | Treat the first three as auto-fail gates, not point deductions. | 100 |
| Category | What it checks | Weight |
|---|---|---|
| Greeting and identification | Branded open; the agent and the purpose are identified. | 15 |
| Needs discovery | Asked the right questions to understand the real issue. | 20 |
| Resolution | Issue resolved, or correctly escalated with a reason. | 25 |
| Empathy and tone | Acknowledged the customer and matched the moment. | 15 |
| Compliance basics | Required notices given and information stated accurately. | 15 |
| Closing and next steps | Clear summary, next steps, and a proper close. | 10 |
| Total | Resolution carries the most weight, because it is the point. | 100 |
We will email all three scorecards as editable files you can drop into your QA tool or a spreadsheet. The preview above stays free either way.
Error: Contact form not found.
Ungated to read here. Email only if you want the editable files.
A scorecard is only as good as what you do with it.
Building one from scratch? The Field Guide on AI Agent QA scorecards walks through the design choices.
About the templates.
Are these really free?
Yes. The full scorecards are on this page, ungated. The only thing the email gets you is an editable file so you do not have to retype them.
Can I use them with another QA tool?
Yes. They are plain scorecards, not a QEval format. Use them in a spreadsheet, your current QA tool, or QEval. The design is what matters.
Is the compliance scorecard legal advice?
No. It is a practical starting point. Compliance requirements vary by industry and jurisdiction, so have your own compliance or legal team confirm what applies to you before you rely on it.
How do these relate to QEval scoring?
QEval scores against your scorecard, whatever it is. These templates mirror the dimensions QEval is built to evaluate, so they are a good fit, but you are free to bring your own.
We score against yours, whatever it is.
Take a template, adapt it, and bring it. Start a pilot and we will score a sample of your conversations against your scorecard and show you each score pinned to its evidence.