Conversation Analyzer

Simulations are a safe place to build a skill. Real calls are where that skill either holds up or breaks down. The Conversation Analyzer brings real conversations into the same coaching system you already use for practice, so a manager can score an actual discovery call against the studio's Discovery rubric and compare it to how that rep performs in practice.

What it does

It takes a sales conversation from any of three places, turns it into a clean, speaker-labeled transcript, and runs AI analysis against a rubric you choose. The result is a saved transcript with a summary or a full score breakdown, ready for review and coaching.

Reps and managers can:

• Send a notetaker bot to a live or upcoming Google Meet, Microsoft Teams, or Zoom call so it records and analyzes automatically.
• Import a past conversation: upload an audio file, a transcript document, or captions, or paste transcript text.
• Pick how to analyze it: a quick summary, or a full Score & Evaluate pass against a studio or default rubric.
• Review and coach: read the scored transcript, edit speaker labels, play back synced audio, and (for non-general, non-management calls) push the transcript into Deal Clinic with Run Diagnostic.

Why scoring real calls matters

Practice scores tell you whether a rep can execute a skill on demand. Real-call scores tell you whether they execute it under pressure with a paying customer on the line. When both run against the same rubric, the gap between them is the coaching conversation: a rep who aces objection-handling practice but loses ground on live objection calls needs different help than one who struggles in both.

Where to find it

The analyzer is a module. Once enabled, it appears in the navigation as Conversation Analyzer at /dashboard/conversation-analyzer. If you do not see it, an admin enables it under Module Configuration > Modules (/workspaces/admin/data/modules), where you can also rename the module and choose whether to surface it on the home page.

Most reps will not stop to upload a recording after every call. Live Recording removes that friction: a notetaker bot joins the meeting, records it, and the transcript is analyzed automatically once the call ends. This is how you get real-call coverage at scale instead of only the calls someone remembered to save.

Live Recording is the default mode on the analyzer page. It supports Google Meet, Microsoft Teams, and Zoom.

Two ways to get the notetaker into a meeting

1. Paste a meeting link. Drop the meeting URL into Paste a Meeting Link and choose when the bot joins:

• Join now sends the notetaker immediately, for a call already in progress or about to start.
• Schedule time picks a future date and time (in your local timezone) so the bot joins on its own. Use this the moment a meeting is booked, so nobody has to remember to start it.

2. Invite the notetaker by email. Add notetaker@salesably.ai to the meeting the same way you would invite a colleague. Either add it as a guest on the Google Calendar or Outlook event, or forward the meeting invite email to that address. This path works across Zoom, Google Meet, and Teams.

Each invite covers a single meeting. For a recurring series, invite or forward each occurrence separately.

Advanced options (optional)

By default the bot auto-detects the conversation type and picks a rubric. Expand Advanced Options before sending to override:

• Transcript title: leave blank for automatic naming based on the meeting, or set your own.
• Scenario: leave on Auto-detect, force a specific type (Discovery, Pitch, Objection Handling, Management), or choose General meeting (no scoring) to capture a transcript without grading it.
• Scoring rubric: auto-select, or pin a specific studio rubric. This is disabled when the scenario is General, since a general meeting is not scored.

After the call

When the meeting ends, the transcript is transcribed and analyzed in the background. The result lands in the owner's results at /dashboard/transcript-results. Live Recording requires the meeting recorder to be configured for your studio; if it is not, the page shows a setup notice instead of the send button.

Not every conversation runs through the bot. You may already have a recording from a phone system, a transcript exported from another notetaker, or captions a customer success rep saved. Import lets you score those after the fact, so historical calls and calls from tools outside Salesably still feed your coaching data.

Choose Import on the analyzer page, then pick how the conversation comes in.

Upload a file

Drag and drop or browse for a file. Supported inputs:

Type	Formats	Notes
Audio	MP3, WAV, M4A, FLAC, OGG, WEBM, MP4, AAC	Up to 2 GB. The file is transcribed with automatic speaker separation, then analyzed. Large files can take several minutes.
Documents	PDF, DOCX	Up to 10 MB. Speaker-labeled dialogue is read out of the document. Legacy .doc is not supported, convert it to .docx, .pdf, or .txt first.
Captions / Text	VTT, TXT	Caption files keep their real speaker names. For plain text, use one line per turn in the format Speaker: message.

Paste a transcript

Use Paste Transcript when you only have text. Format each turn as Speaker Name: message on its own line so the system can tell who said what.

Upload multiple

Upload Multiple runs a batch (up to 15 files at once) through transcription and analysis together. Use it to backfill a stack of recorded calls in one pass instead of one at a time.

Import from Zoom

If a call was recorded to the Zoom cloud, Import from Zoom pulls the transcript in directly so you can review and score it like any other conversation.

For audio uploads, transcription runs through an external speech-to-text service that separates speakers automatically. Documents and caption files are parsed directly and keep any speaker names they already contain.

These choices decide what the analysis actually measures. The right rubric scores against your team's standard for that kind of call; the wrong one (or no labels) produces feedback that is technically accurate but useless for coaching. The good news is the analyzer recommends sensible defaults, so most of the work is confirming, not configuring.

How would you like to analyze this conversation?

First, pick the depth of analysis:

• Summary Only: a plain summary and key takeaways, no score. Best for a general internal meeting, or when you just want the gist of a call.
• Score & Evaluate: a full graded pass against a rubric. This is the mode that produces coachable, comparable results. Use it for real sales calls.

For uploaded audio, the system reviews the transcript and, if it recognizes a sales conversation, suggests a scenario type and switches to Score & Evaluate for you.

Scenario type

In Score & Evaluate mode, What kind of conversation was this? tells the analysis which framework applies and narrows the rubric list to ones that fit. The options map to your studio's modules:

• Discovery
• Pitch / Messaging
• Objections
• Cold Call / Outbound
• Management Practice
• Plays

Most of these are customer conversations a salesperson is practicing or running. Management Practice is the exception: it covers conversations where the person being evaluated is the manager (a performance conversation, a one-on-one, or another difficult discussion with a colleague rather than a customer). Pick the type that matches what the call actually was. An AI Recommended badge appears when the system has a suggestion.

Rubric

How would you like to evaluate or score it? is the rubric. The rubric defines the criteria, the scale, and the coaching language, so this is the single biggest lever on what the feedback says. Rubrics are grouped:

• Studio Rubrics: ones your team built or customized. These should be your first choice, because they reflect how your organization actually sells.
• Default Rubrics: Salesably baselines, available as a fallback when you have not built a studio rubric for that scenario yet.

An AI Recommended rubric is pre-selected based on the transcript and scenario; change it whenever a different rubric fits better. Manage and create rubrics under Module Configuration > Rubrics (/workspaces/admin/data/rubrics).

Speaker labels

Speaker labels are who-said-what. The analysis needs to know which voice is the rep and which is the customer to score the right person, and a clear label is also what makes the transcript readable for coaching. The system assigns roles automatically and, for caption files, keeps real names. Review the labels in the transcript before analyzing, click any speaker to rename or re-role them, and fix anything that looks wrong. You can also edit an individual line of dialogue if transcription got a word off.

Title

Give the conversation a clear title. A title is generated automatically from the content, but a descriptive name (account, call type, date) makes the result far easier to find later in results and reports.

Analysis is only valuable if someone acts on it. The result page is built for the coaching conversation: it shows where the call went well, where it slipped, and gives a manager the evidence to have a specific, fair discussion instead of a vague one.

What a scored result includes

• Overall score and rubric-level breakdown, so you can see at a glance which criteria carried the call and which dragged it down.
• Summary, key takeaways, and recommended next steps for a fast read before a one-on-one.
• The speaker-labeled transcript, with strengths and gaps tied to specific moments rather than generalities.
• Synced audio playback when the conversation came from a recording or audio upload, so you can listen to the exact exchange a comment refers to.
• Run Diagnostic, which sends the same transcript into Deal Clinic for a deal-level readout. This is available for sales calls; it is intentionally turned off for general (unscored) meetings and for Management Practice conversations, which are not deal diagnostics.

Where results live

• A rep's own results: Conversation Analysis under their dashboard at /dashboard/transcript-results.
• Studio-wide results: the Conversation Analysis report at /workspaces/reports/transcript-results, for admins and supervisors.

Admins and supervisors can open and coach on any team member's conversations. Reps see their own results and the coaching left on them. Because real-call results sit alongside practice results, the most useful coaching move is to look at both together: a rep's practice trend tells you what they can do, their real-call scores tell you what they actually do, and the difference is where your time pays off.

A note on privacy

Real customer calls contain sensitive information. When uploading audio, members can toggle Redact personal information to replace names, emails, phone numbers, and similar details with placeholders. A studio admin can enforce this for everyone by turning on Always redact sensitive information in Studio Settings (/workspaces/{workspaceId}/edit); when enforced, redaction is always on and the per-session choice is hidden. Note that redacting a recording means its audio is not retained for synced playback.