Features / Support
Insights
What is actually happening in the queue
Volume, resolution, response times, satisfaction and AI spend in one place, and a risk score that flags a customer drifting toward the exit before they go.
Who resolved it
What it does
- Volume and resolution over any period you choose
- Response and resolution times against your targets
- Satisfaction, by agent and by product
- At-risk customers, scored from real signals rather than a guess
Without it
Everyone has a theory about what customers complain about
The chart says the number went down
Resolution rate dropped and the dashboard has no way to ask why. Somebody exports a CSV, somebody else joins it to another CSV, and by the time there is an answer it is next week and the number has moved again.
You find out they left in the renewal call
Four reopened tickets in March. 'This is getting hard to justify internally' in April. Silence in May. Every one of those was already in your inbox, and none of them was ever on a screen anyone looked at.
Nobody can say what the AI cost you
There is one number on the invoice and no way to break it into what it was spent on, which model spent it, or whether the expensive model is quietly doing the cheap work.
How it works
Count what happened, over a window you choose
Volume, resolution rate, first response median and 90th percentile, how many conversations are still waiting for a first reply, and the tags the queue actually clustered on. A preset period or a custom date range.
What that includes
- Volume
- resolution rate
- first response median
- 90th percentile
Split every resolution by who did it
AI on its own, AI-assisted, a person, or the customer solving it themselves. Attribution is written at the moment of resolution and counted from those records, not inferred from the transcript later.
What that includes
- AI on its own
- AI-assisted
- a person
Score the accounts that are drifting
Ten signals feed a score from 0 to 100 and a band from low to critical: reopened tickets, escalation rate, sentiment drift, weakening ratings, churn-intent language, a plan downgrade, an account that has gone quiet, an inactive champion, and more. Each scored contact shows its top three factors with a suggested next move.
Make one bad week not read as a crisis
The band rises when several signals fire together, not when one spikes, and a sensitivity setting decides how many it takes. New accounts and trials are scored in their own cohort for their first 60 days so onboarding noise does not fill the at-risk list.
Put the AI bill next to the work it did
Spend broken out by the kind of work and by the model that did it, over the same window as everything else, exportable as CSV.
What it means for you
The warning arrives while you can still act
An account drifting toward the exit shows up with the three things that moved it and something to do about each, rather than as a cancellation email two weeks after the decision was made.
Your numbers survive contact with an engineer
Each figure is a count of records over a stated window. Where there is no data it says no data instead of drawing a zero, and anything your role cannot see says that too.
You can answer 'is this worth it' with both halves
What the AI handled and what the AI cost are on the same screen, over the same dates. The conversation about whether to keep going stops being a matter of taste.
The honest limit: Hours saved is the one estimate on the page: resolutions multiplied by a baseline minutes-per-resolution figure that you set, and it is labeled as an estimate everywhere it appears. The risk score is a weighted signal model rather than a trained churn predictor, so treat a high band as a reason to look, not as a verdict.
Also in Support