Atlas SalesAtlas Unchained
Pre-launch · in build

Every outbound call, recorded, transcribed and scored.

Atlas Sales captures your team's sales calls, turns each one into a transcript, and scores it against a rubric you control. Reps get their own calls back with specific feedback. Managers can finally review calls they weren't on.

The Atlas Sales iOS app during a live call: a recording chip, a note that the disclosure was played at three seconds, a call timer, and a full-width Stop recording button.

The problem

A small team makes hundreds of calls a month.
Nobody hears any of them.

Reps dial from personal phones. Managers coach from memory, or from the one call they happened to sit in on. Everyone agrees the calls are where the money is lost, and nobody has a way to look at them.

1

Capture

Calls are placed from the Atlas Sales app on the rep's iPhone, using your company number. Each call is announced, recorded on two separate audio channels, and stored on infrastructure you control.

2

Transcribe & score

Every recording is transcribed and scored against your rubric — the specific things you want reps to do on a call. Each score cites the line of transcript it came from.

3

Give it back

The rep opens their own call, sees what they were marked on and why, and can play the exact second. The manager sees the same thing, plus the team view.

A scorecard in the Atlas Sales app showing a grounded score of four out of ten beside a provisional five out of eight marked not counted, then the first criterion with its evidence quote.

The scorecard

A score a rep can argue with.

Every criterion shows the line from the transcript it was judged on. Nothing is a black-box number — if the rep disagrees, they can see exactly what the scorer heard and take it to their manager.

  • Evidence, not verdictsEach score quotes the moment it came from and links to that second of audio.
  • “No evidence found” is a real answerWhen the call doesn't contain enough to judge, it says so. It never renders an absence as a zero the rep has to defend.
  • Your rubric, versionedYou write the criteria. Every score records which version of the rubric produced it, so a rule change doesn't silently rewrite history.

The principle the product is built on

Two numbers that never add up.

A scoring rubric starts life as a mix of things you know matter and things somebody guessed. Most tools average them together and show you one confident number. We don't, because a guess added to a measurement looks like a measurement and isn't — and that's the number people screenshot.

Grounded
4 / 10
Provisional
5 / 8
Provisional criteria are hatched, labelled, and excluded from every total, average, trend, team figure and export. They become grounded when a named person confirms them — not automatically, and not quietly.

For managers

Review the calls you weren't on.

A web console with the whole team's calls: playback, speaker-labelled transcript, the scorecard, and where calls are losing points. It runs separately from this website, on its own origin, and there is no URL that hands out a recording — every listen is scope-checked and logged.

The Atlas Sales web console showing a call detail page: waveform with markers, the transcript with speaker labels, and the scorecard with grounded and provisional totals.
Call detail. Waveform markers on the moments that were scored, the transcript beside them, and a banner warning when the speaker labels are a guess — with a button to swap them. A score attached to the wrong speaker is worse than no score.
Team overview in the Atlas Sales console: call volume, connect rate, talk time, grounded average, unscored count, and a per-rep table sorted alphabetically.
Team overview. Volume, connect rate, talk time, and where calls lose points. The per-rep table is sorted alphabetically with no rank column and no sortable score — deliberately.
Rubric management in the Atlas Sales console: criteria with confidence flags and a publish preview showing what a change would affect.
Rubric management. Edit the criteria, then see before you commit: what changed, how many scorecards it supersedes, and which comparisons stop being valid.

Product screens from the current build, shown with sample data.

Consent and access screen in the app: the exact disclosure wording that was played, when it was played, and a log of who has listened to this recording.

Consent & access

Recorded out loud, not quietly.

Every call opens with a spoken disclosure before recording starts — not after. The exact wording is stored as an immutable record, so months later you can still answer what a given caller was told.

  • Announcement before recordingThe order is enforced in the call flow, not left to a setting somebody can turn off.
  • The rep sees who listenedThe access log is shown to the rep on their own calls, including when a manager played one. A coaching tool that only logs one direction is surveillance.
  • No public recording linksAudio is only reachable through an authenticated, permission-checked request. There are no signed URLs and no shareable media links.

Who it's for

Small teams where every call matters.

Built for two-to-ten-rep operations in home services, real estate and property management — the size where you can't afford an enterprise call-intelligence platform and can't afford to lose the calls either.

Home services

Estimate calls where the difference between a booked job and a price-shopper is what got asked in the first ninety seconds.

Real estate

Lead follow-up at volume, where the same three questions decide whether a call turns into an appointment.

Property management

Owner inquiries that turn on qualification — the unit, the occupancy, the service area — before anyone quotes a fee.

Where we actually are

Pre-launch, with one customer.

We'd rather tell you this than have you find out on a call. Atlas Sales is in build, running with a single property-management company. Here is the honest state of each piece.

Web consoleRunning

Team overview, call list, call detail with playback and transcript, scorecards, rubric management, settings. Running internally against real call data.

Scoring & transcriptsRunning

Transcription and rubric scoring with per-criterion evidence, versioned rubrics, and the grounded / provisional split.

Call capture & consent flowBuilt, not yet live

The call state machine, the spoken disclosure and dual-channel recording are written and pass an end-to-end replay test that asserts recording never starts before the announcement. They have not yet run against a live carrier.

iOS appIn build

Designed in full and unblocked. Not shipped. Distribution is gated on Apple organisation enrolment.

AI coachIn build

Answers a rep's questions from their own scored calls, with citations — and refuses when the data can't support an answer. Scoping, citation and refusal logic are done; it is not open to customers yet.

Training path, XP, streaksNot building yet

Designed and deliberately shelved. A coaching path with no lessons written is a shell, and you don't get a second first impression on a coaching tool.

What you won't find on this page: customer counts, logos, testimonials, or a number claiming how much reps improve. We have one customer and no outcome data yet. When we have results worth quoting, they'll be here with the name attached.

Pricing

Not published yet — on purpose.

We're setting pricing with our first customers rather than picking a number and defending it later. Tell us how many reps you have and roughly how many calls a week they make, and we'll quote you and explain how we got there.

What it will be based on: number of reps, call volume, and how long you want recordings kept. No per-seat minimum, no annual lock-in at this stage.

See it on your own calls.

A twenty-minute walkthrough of the console, honestly scoped to what works today. If it isn't a fit yet, we'll tell you that.