Is Your AI Program Working?
Read the Trend, Not the Number.
AI Outcomes shows whether call quality, compliance and customer satisfaction are moving month by month, scores how mature your AI operation is — and what would move it up — and, if you put an AI on the phone, what each of its calls amounted to and what a resolved call cost.
Trend First, Number Second
Every metric leads with whether it is improving, flat or declining across up to twelve months, then shows the sparkline, then the latest value. A flat 41% and a climb from 22% to 41% are different stories — a single tile tells them the same way.
Quality and Compliance by Month
Mean call score and compliance rate month by month, graded against your own rubric and compliance rules — the same questions whether your calls are answered by people, by an AI, or both.
Two CSAT Instruments, Never Blended
Survey CSAT is what customers told you. Inferred CSAT is the AI’s per-call read of how the conversation landed, across every scored call. Both are charted side by side, labelled, and never averaged together.
AI Operating Maturity Score
A 100-point score across five dimensions — outcomes, operating model, training, trust and business discipline — computed from what your system actually did, not from a questionnaire. Every point you have not earned comes with a sentence saying what would earn it.
What the AI Voice Agent Resolved
For teams running the AI voice agent: the share of calls it resolved end to end, the share that reached a person, whether the AI disclosure played, and estimated cost per resolved call — with every finished call stacked by outcome, month by month.
Handoff Quality
When the AI hands a caller over: median and 90th-percentile time to a person, how many people were rung, and whether the account context and call summary were ready before your agent picked up.
Numbers You Can Take Into a QBR
An outcome report is only useful if nobody has to footnote it later. These rules are built into how every figure on the page is computed.
Small months show a dash, not a percentage. Below ten counted calls in a month, rates are withheld. One escalation out of three calls is three calls, not a 33% escalation rate — and a trend built on that would be an artefact of volume.
Resolved means resolved, not deflected. Dropped calls and calls cut off at the duration limit count against the resolution rate, never toward it. The fastest way to improve the headline number is not to get worse at answering the phone.
Cost is per resolved call, and stated as a floor. Cost per call falls whenever an AI gives up faster, so we divide by calls it actually resolved. The figure is model spend only and is labelled as a lower bound — carrier minutes are not included.
AI and human scores are not presented as a ranking. Both groups of calls are graded by the same rubric, but the AI hands its hardest calls to your team. The page says so on screen, so nobody walks into a QBR claiming the AI outscores the people.
A partial month says it is partial. If a month was computed from an incomplete read, a notice names the month above every chart, instead of quietly showing a lower number.
Every team
Quality, satisfaction and maturity
Call quality and compliance trends, survey and inferred CSAT, and the AI operating maturity score. None of it needs an AI on the phone — a fully human team gets the same questions answered, and the maturity score is most useful to the teams furthest from the top of it.
Teams running the AI voice agent
Plus containment, cost and handoffs
Once the voice agent takes calls, three more panels appear on their own: the headline resolution, escalation, disclosure and cost trends; where every call ended up; and handoff quality. AI-answered calls are scored by the same rubric as your team’s, so both appear on the quality chart.
Who Benefits from AI Outcomes
For the Customer
Because handoffs are measured, a team can see — and fix — the calls where the person picking up after the AI did not yet have your account or a summary. Those are the calls where you get asked to explain everything twice.
For the Agent
Handoff quality is measured, not assumed. If escalated callers keep reaching you before their context does, it shows up on the page as a number someone is accountable for rather than as a complaint you have to make.
For the Business
A month-by-month answer to “is the AI program paying off” — resolution, cost per resolved call, quality and satisfaction — with a maturity score that names the next concrete step instead of handing you a grade.
Related
Know What Your AI Is Worth
Book a demo and we'll walk through the trends, the maturity score and the gaps it names, using your own call data.
Request a DemoAvailable on every plan to admins and managers. The voice-agent panels appear for Enterprise teams running the Live AI Voice Agent.