Your supervisor reviews a handful of calls a week because that is all one person can get through. QA Batch Scorer reads all of them, scores each against a rubric you wrote, and quotes the transcript line that justifies every score.
| Agent | Greeting | Understood the issue | Accurate information |
Verified the customer * | Tone and control | Clear next step |
|---|
Sampling was never a quality strategy. It was a capacity limit that everyone agreed to stop questioning.
What one supervisor can physically listen to in a week.
Where the missed disclosures and the coachable moments actually live.
Every call, same rubric, same standard, every time.
One HTML file. Double-click and it opens in your browser. Nothing to install, no account to create.
Pick an AI provider and paste your key. A step-by-step guide is in the download. You do this once.
CSV, TSV, JSON or plain text from any system. It reads your columns and lets you correct it if it guesses wrong.
Every call scored against your rubric, with evidence quotes, a per-agent rollup, and three CSV exports.
Platform pricing is per agent, per month, and usually billed six or twelve months up front.
Plus your own AI provider cost, which runs roughly $2 to $6 per thousand calls depending on length. The tool estimates it before you press Score. Platform figures are published or third-party-reported entry rates for a 25-seat team; enterprise pricing is quoted, not listed.
The tool is a file on your computer. When you press Score, your transcripts go from your machine to the AI provider you chose, using your key, under your agreement with them. There is no account here, no upload, and no infrastructure of ours in the path.
If you need live agent assist and a coaching workflow, buy a platform. If you need to know what is actually happening on your calls before you spend that money, this is four minutes away.
No. If you can open a file and copy and paste a key, you can run it. The setup guide walks through getting an API key with screenshots. The whole thing takes about four minutes and you only do it once.
It is how the tool talks to the AI that reads the calls. You get one from a provider like OpenRouter — sign up, add ten dollars with a card, copy the key. That ten dollars scores a few thousand calls. Doing it this way is why your transcripts never touch anyone's server but your chosen provider's, and why this costs $99 instead of four figures a month.
You pay your AI provider directly. A thousand calls runs roughly two to six dollars in credit depending on how long the calls are. The tool shows you an estimate before you press Score.
Any system that exports transcripts as a file. Five9, Genesys, Talkdesk, RingCentral, Dialpad, Zoom, or a spreadsheet someone maintains by hand. It reads CSV, TSV, JSON and plain text, guesses which column holds the transcript, and lets you correct it.
Thousands. It processes several at a time, shows progress, and can be stopped mid-run. Very long transcripts are trimmed to roughly the first fourteen thousand characters.
Check them rather than trusting them — that is why every criterion comes with the quote from the transcript it was based on. Before any score reaches a coaching conversation, score twenty calls, have your best QA person score the same twenty by hand, and compare. Where you disagree it is almost always the rubric wording, and fixing it costs pennies. The guide covers this.
No. It hands them the ninety-odd percent of calls they never had time to open, so their judgment goes where it matters instead of into pressing play.
Thirty day refund, no questions asked. Email the address on your receipt.
Includes the tool, three rubric templates, a setup and calibration guide, and a sample transcript file so you can watch it work before pointing it at real calls.