Every audit comes back with what the agent did well, what to fix, and what to actually do about it, attached to the call it came from.
“Work on your empathy” is not coaching. It gives an agent nothing to change on the next call, and it is the natural output of a QA process that produces category scores and stops.
Every audit returns three lists: strengths, areas to improve, and suggested actions. They are written against the specific call, so a supervisor opens a coaching conversation already knowing which moment to replay.
The strengths matter as much as the improvements. A feedback process that only ever surfaces failures gets avoided by everyone in it.
coaching.strengths / .improvements / .suggestedActions
A failed call has a proximate cause and an underlying one. The agent skipped the disclosure; the script puts the disclosure after verification. Coaching the first without fixing the second produces a frustrated agent and the same failure next week.
Where an audit finds problems, it returns an analysis of what drove them rather than only recording that they happened. When the same pattern appears across a batch, that is usually a process finding rather than a performance one.
rootCauseAnalysis · per audit
The disclosure was skipped on all four failing calls in this batch, and in each the agent moved straight from greeting to account lookup. The pattern is consistent with the opener script, not with individual performance: the script places the disclosure after verification.
Not every miss is equal, and a compliance report that lists them flat makes you re-triage it by hand every time.
Violations come back graded low, medium, high or critical, alongside the keyword checks that ran and a compliance score for the call, so the four calls that need a manager today are separable from the forty that need a note.
We grade against the rules you configure. We do not certify you as compliant with any particular regulation; that assessment is yours to make against your own obligations.
compliance.violations[].severity · low → critical
Nine calls need a manager today. The other 253 need a note.
| Severity | Count |
|---|---|
| low | 148 |
| medium | 74 |
| high | 31 |
| critical | 9 |
Several tools in this category advertise coaching plans that arrive on a schedule. Ours does not, and pretending otherwise would be easy to disprove in week two.
| Criterion | QXAI |
|---|---|
| Strengths, improvements and actions per call | YesYesOn every audit, tied to that call. |
| Root-cause analysis | YesYesReturned where an audit finds problems. |
| Severity-graded compliance violations | YesYeslow / medium / high / critical. |
| Per-agent trends over time | By nameBy nameThe dashboard groups audits by the agent name recorded on them. Agents are not yet linked to user accounts, so the grouping is only as consistent as the name entered. |
| Scheduled or weekly coaching plans | NoNoThere are no scheduled coaching plans. Coaching is produced per audit, when the audit runs. |
| Delivering feedback to the agent | NoNoNo acknowledgement or sign-off flow. Turning an audit into a conversation is still a human job. |
| Agent training simulations | NoNoNo role-play or practice mode. We audit real calls. |
| Email digests to supervisors | Alerts onlyAlerts onlyAlert rules can fire on a fatal failure, a low score or a compliance violation. There is no periodic summary. |