Vocalis

Every call analyzed, not a sample of them

Quality teams listen to a handful of calls a week and hope they were representative. Vocalis analyzes all of them when they end, across thirteen sections, and links every finding to the seconds of audio it came from.

Analysis · CALL-885213 sections
Caller intentInquiry · low
Objectives3 of 3 met
ResolutionFull, no follow-up
SentimentPositive, rising
Communication promisesThu 2 PM
Data collection4 of 4 fields
Information updatesAddress corrected
Human contactReached a person
Do not callNot requested
LegalNo threat detected
CrisisNone detected
IVR navigationNot encountered
CallbackScheduled
Resolution time by agentLast 5 months
OctNovDecJanFeb
Reports · ask your data
Which reasons drove the most repeat calls last month?
Prior authorization status82
Claim denial codes61
Deductible remaining44
ID card requests23
ShareExport CSVSchedule

What comes back from a call

Analysis runs when the call ends rather than in a nightly batch, so a missed promise or a safety flag is in front of someone while it can still be acted on.

Caller intent
Why they rang, classified, with urgency and the keywords that decided it.
Objectives
Whether the call did what it was placed or answered to do, down to sub-objectives, with reasoning.
Resolution
Resolved fully, partly or not at all. Whether it escalated, and whether follow-up is required.
Sentiment
How the member sounded, tracked across the call rather than scored once at the end.
Communication promises
Anything committed to, with the method, the timeframe, the urgency, and whether the contact detail was confirmed.
Callback
A callback that was scheduled, pulled out as its own commitment.
Data collection
Which fields the call was meant to collect, what came back, and how much of it.
Information updates
Details the member corrected, with the old value and the new one.
Human contact
Whether a person answered, whether it reached voicemail, and whether a message was left.
Do not call
A request to stop calling, explicit or implicit, so it can be honoured immediately.
Legal
A threat of a lawsuit, a lawyer or a regulator, with severity.
Crisis
Language that needs a person now, with severity and a recommended action.
IVR navigation
If the call hit a phone menu: the path taken, whether it got stuck, and for how long.
Analysis · CALL-48214m 12s
Objective metSentiment positivePromise: callback Thu 2 PMNo DNC requestNo safety concern
Evidence

“I’ve booked a callback with our appeals team for Thursday at two.”

2:41
One call, as the analysis view shows it.

Every finding carries its evidence

A finding is not a label. It arrives with the lines that produced it, the start and end timestamp of those lines, and a confidence score. Press play and you are at that second, not at the top of a four minute recording.

That is what makes the analysis usable rather than merely present. Disagreeing with it takes about as long as reading it, which is the difference between a team that trusts the output and a team that ignores it.

Communication promisesconfidence 0.94
callbackphonethis weekcontact confirmed

Appeals team will call the member back Thursday at 2 PM.

Evidence · 2:41 to 2:48

“I’ve booked a callback with our appeals team for Thursday at two.”

Jump to the audio

Human, voicemail, or a phone tree

On outbound work the first question is whether anyone actually answered. The analysis distinguishes a live human from a voicemail greeting and records whether a message was left, so a campaign’s reach number means what it says.

If the call landed in somebody else’s phone menu, that is captured too: the path taken, whether it reached the right destination, and whether it got stuck. Which turns a stack of failed calls into a fixable routing problem.

Objectives are yours, not ours

Objectives are defined per campaign, in your words, and the analysis reports each one as met, not met, or unknown, with the reasoning and the evidence behind the call. Sub-objectives break a compound goal into the parts that actually failed.

Unknown is a real answer. A system that only returns true or false has to guess on the calls that ended early, and a guess in this column poisons the trend that gets reported upward.

Ask the record a question

Reports are not a fixed set of dashboards you file a ticket to change. Ask in plain language, get the chart, then share the link, export it, or have it arrive on a schedule.

The questions worth asking are the operational ones. Which reasons drove the most repeat calls last month. Where resolution time is going. Which line is generating callbacks nobody is making.

Reports · ask your data
Which reasons drove the most repeat calls last month?
Prior authorization status82
Claim denial codes61
Deductible remaining44
ID card requests23
ShareExport CSVSchedule

Trends, not just totals

The same data behind the per-call view drives the trend views: resolution time by agent, objectives met by campaign, sentiment by line. Because every call is analyzed rather than sampled, a trend is a trend and not an artefact of which calls somebody happened to pull.

Getting it out

Reports can be shared as a read-only link for people who should not have an account, exported to CSV or Excel, or scheduled so the weekly numbers arrive without anyone opening the product. The same data is on the API if it belongs in a warehouse.

Developer documentation: the analytics endpoints, call records, and webhooks that fire when analysis completes.

Make the front door intelligent