Grade every call.
Judge the whole cohort.
Custom scorecards score each conversation against your playbook. Cohort AI QA packs surface hallucinations, resolution, sentiment, and trends across hundreds of transcripts.
Call scorecard
Maple Dental · Reception
Sarah M.
+49 170 ··· 42
Scored against this assistant’s custom criteria — not a generic industry template.
Person score
0/100
Passed · ≥70
- Used clinic greeting0
Said “Maple Dental, how can I help?”
- Asked recording consent0
- Offered two appointment slots0
- No invented pricing0
Score every call — then judge the cohort
Scorecards grade each conversation against your rubric. Cohort AI QA packs show how quality moves across a date range.
Per-call AI scorecards
Define your rubric once — greeting, consent, booking steps, prohibited claims — and every call is scored automatically.
Custom evaluation criteria
Boolean and scored checks that match your playbook, not a generic industry template.
Language & hallucination packs
Cohort runs flag invented facts, weak knowledge-base recall, and language quality across hundreds of calls.
Resolution & sentiment
See whether calls actually resolved, how callers felt, and which questions keep coming up.
Performance trends
Compare score and resolution before and after a prompt or model change — with a clear trend line.
Catch regressions early
Run Full QA after a deploy so a bad prompt does not scale across your entire outbound list.
Your rubric on every call — pass, fail, and why
Turn playbooks into checks the AI grades after hangup. Required criteria fail the scorecard; optional ones guide coaching.
- Custom criteria per assistant
- Pass / fail with per-check scores
- Surfaces in call history and webhooks
Assistant scorecard
Rubric runs on every completed call
- Used approved greetingRequired
- Asked recording consentRequired
- Offered at least two slotsRequired
- Did not invent pricingRequired
- Confirmed next step
Run a pack across hundreds of calls in one job
Pick Full QA, Language & Hallucinations, Resolution & Sentiment, or Performance Trends. Choose a date range, spend credits on the analysis, and open the Call QA Overview.
- Transcript-based analysis (no audio WER in v1)
- Trends, top questions, and resolution rates
- REST API and MCP for runs and packs
Cohort AI QA
PacksFull QA
Score, resolution, hallucinations, sentiment, trends, top questions.
86
Avg score
81%
Resolved
312
Calls
Grade the conversation. Then judge the fleet.
Scorecards catch the single bad call. Cohort QA shows whether yesterday’s prompt change helped — or hurt — across hundreds of transcripts.
Built on the same transcripts as Analytics — with packs, credits, and dashboards dedicated to quality assurance.
Common questions about Quality & QA
Scorecards grade each call against criteria you define on the assistant. Cohort AI QA runs a pack (Full QA, Language & Hallucinations, Resolution & Sentiment, or Performance Trends) across many calls in a date range and opens a dashboard.
Stop guessing whether quality slipped
Turn on scorecards for every call, run a cohort pack after your next prompt change, and see what actually moved.