Everything the platform actually does

The microphone opens when a call starts and closes when it ends. It is not an always-on room recorder, and it is not listening between calls — which is both the right privacy posture and the reason your storage bill stays sane.

Collector and consumer stay on separate audio channels. That separation is what makes attribution a fact rather than a guess: who said a thing is determined by which device the audio arrived on, never by a model trying to tell two voices apart. Diarisation gets this wrong often enough that a compliance finding built on it would not survive a challenge.

There is no per-dialer integration, and that is deliberate. Building one for RingCentral, then Five9, then Zoom, then Teams means every one of those vendors gets a veto over your compliance programme on their own release schedule.

Instead the agent reads what Windows already knows: a process holding both an audio playback and an audio capture session at the same time is on a call. That is true of a native softphone, a browser tab running a hosted dialer, and a meeting client alike — so it works with whatever your floor uses today, and it keeps working when you switch.

System sounds and notification chimes are excluded, so a stray message ping never registers as a call.

Each call is transcribed with segment-level timestamps, so a finding points at a moment rather than at a whole recording. A supervisor clicks a flag and hears the four seconds that matter, instead of scrubbing a twelve-minute call looking for it.

Because it is text, it is searchable. A creditor asks whether your floor ever used a particular phrase; you answer from the record in minutes rather than commissioning a review.

Every transcript is read against a collections rubric that knows what a debt-collection call is supposed to contain and what it must never contain. It flags missing disclosure, third-party disclosure, threats, contact after cease, misrepresentation and abusive language — each with the timestamp and the quote that triggered it.

There is no single blended score, deliberately. The rubric reports each dimension separately — whether the debt-collection disclosure was given, whether a cease request or a dispute was raised, and the tone on each side — alongside the flags. A warm, friendly, well-handled call with a statutory violation inside it should not average out to a passing grade, and on a single-number scorecard it does.

A flag is a pointer, never a verdict. It tells a human where to listen. The decision stays with the person who can be held responsible for it.

An activity timeline shows which application was in front and for how long, split into active and idle, per person, per day. Screenshots are captured at the moments the activity stream says something changed — a new account, a different screen — rather than on a blind timer that mostly photographs an idle desktop.

This is the part the conversation-only compliance tools do not do, and it is the difference between knowing what was said and knowing what was happening. A collector reading the wrong account while making a promise-to-pay is a problem no transcript alone reveals.

One email per day covering the day before. Per person: active and idle time, the applications they worked in, calls handled, and every flagged moment with a link straight to that second of audio.

It also names who was silent — enrollled but sent nothing all day. A missing person is usually a machine that was off or an agent that stopped, and knowing that before you read anything into their numbers prevents the most common misreading a monitoring tool causes.

Recording voices and storing what people type are distinct legal categories — under two-party consent statutes, under GDPR, and to a works council. So they are gated on an attestation in the console: an administrator states the lawful basis and confirms staff were told, and until that is signed the endpoints do not collect at all.

They do not collect quietly and hide it. They do not collect. That difference is the whole design, and it is what lets your compliance officer sign off on the tool itself.

For floors that need the whole picture — usually those handling disputes, or working under a consent order. Video runs alongside the call and the activity stream, so a reviewer can watch what happened rather than reconstruct it.

Most floors do not need this, and we will tell you so. The screenshots usually answer the question at a fraction of the storage.

FDCPA claims do not arrive the week the call happened. Audio is retained for three years by default on the Compliance tier, with a per-listen audit trail on every stored recording so you can show who accessed what and when.

Retention is a promise to the people being recorded as much as a tool for you, which is why it is a stated window rather than an indefinite pile.

This platform was a remote-desktop product before it was a supervision one, so reaching a collector's machine to fix it came free. When a work-from-home collector's headset, VPN or dialer breaks, someone connects and fixes it instead of couriering hardware.

No competitor in the compliance category offers this, because none of them started where we did.