AI Policy

Clarity uses AI in the Script Builder, practice calls, and some free tools. This page explains what happens in each, and what doesn't.

#What the AI does

When you click Build script, Clarity sends a request to OpenAI's API to draft a cold call script: an opener, value proposition, proof point, meeting ask, and objection responses.

What's sent: the product name, description, pain points, call goal, persona and angle you entered, plus our own writing instructions.

#Practice calls

When you start a practice call, an AI plays a fictional prospect from the practice list (made-up names on reserved 555-01xx test numbers). Each thing you type or say is sent to OpenAI with the script you're practicing, so the prospect can reply. When you hang up, the conversation is sent once more to produce your scorecard.

• Speaking instead of typing uses your browser's built-in speech recognition. In Chrome, that audio is processed by Google; other browsers may handle it on your device. Clarity receives only the text. Read-aloud uses your device's voices.
• Practice conversations and scorecards are saved so you, and your team's admins, can review your progress.
• Any quote in a scorecard is checked against what you actually said, and removed if it doesn't match.
• Practice calls never count toward A/B results.

#Free tools

The free tools at /tools that are marked "AI" send what you type, plus the pages they read for prospect research, to OpenAI to generate a result. In the Call Analyzer:

• Pasted transcripts are analyzed in your browser. Nothing is sent anywhere unless you click Get AI coaching, which sends the transcript once.
• Uploaded recordings are sent once to OpenAI for transcription and speaker labelling. Clarity doesn't store the audio or the transcript.

#AI agents and the API

Transcripts your agents send to log_outcome or analyze_call are measured by Clarity's own rules-based analyzer, not an AI model. Only the measurements (talk share, question counts, call-flow checks) are stored, never the transcript text.

#What's never sent to an AI model

Your real call logs, call notes, imported contacts, A/B results, teammates' details and billing information.

OpenAI states that data sent through its API is not used to train its models by default. See OpenAI's enterprise privacy commitments.

#What the AI does not do

• It doesn't make decisions about people. A/B winners are chosen by a standard statistical test on your logged outcomes, not by AI. Variant assignment follows a simple, transparent rule: each rep gets the variant they've used least.
• It doesn't make calls or talk to real prospects. Your reps do. The only "prospects" it plays are the fictional ones on the practice list.
• It only scores practice, never real work. Practice scorecards are coaching feedback on role-play calls. Leaderboards, A/B results and the Admin Dashboard's real-call numbers are plain counts of logged outcomes, not AI judgments.

#Limits you should know

AI-generated scripts can be wrong, overconfident, or invent facts, such as a statistic or a customer result you don't have. Before anyone uses a script:

• Read it and remove any claim you can't back up
• Make sure it fits your brand and doesn't promise what your product can't deliver
• Treat it as a first draft that your team owns

You're responsible for the content your team uses on calls.

#Calling responsibly

Clarity helps you write and test scripts. It doesn't check who you're allowed to call. Follow the laws that apply to you, such as do-not-call registries, telemarketing rules, and consent requirements, and respect every opt-out request.

#Questions

Email live.script.official@gmail.com. If we add AI to other parts of Clarity, we'll update this page first.