AI Transparency
Last updated: July 1, 2026
SkillTap scores people's work. We think you should know exactly how, including where it falls short.
This page is written for the reps being scored as much as for the people buying SkillTap.
What the AI does
- It plays a prospect. You hold a spoken practice call with an AI persona — a buyer with a role, a personality, and a reason to push back.
- It transcribes the call. Your speech and the persona's are turned into text as you go.
- It grades the transcript. A second AI model reads the transcript against the rubric your organization wrote, and produces a score from 0 to 100 plus written coaching feedback.
Depending on the roles your organization configures, your manager can see that score and feedback.
What it is not
SkillTap is not a lie detector, a personality assessment, or a measure of anyone's competence. It is a practice tool.
It does not know your quota, your deals, your history, or anything about you beyond the transcript of the call you just had.
Limitations
Scores are probabilistic. The AI can misjudge tone, unfamiliar accents, industry jargon, and non-native speech. The same call scored twice may not produce the same result.
Two things follow from that, and we would rather say them here than bury them:
- A low score can be the model's fault, not yours. Transcription errors, an unusual sales motion, or a rubric that does not fit the call all produce bad scores for good calls.
- A high score is not proof of a good call. The model rewards what the rubric describes, which is not always what wins.
The scores are most useful compared against your own over time, on similar calls. They are least useful compared between two different people.
Human review is required
Scores are a coaching signal, not a decision. Your organization has agreed not to use them as the sole basis for employment decisions.
That is a contractual obligation in our Terms of Service, not a suggestion: before anyone acts on a score, a person has to look at it, apply their own judgment, and be free to reach a different conclusion.
Your content is not used for training
We do not use your calls, transcripts, or scores to train AI models, and we do not use them to improve SkillTap.
Our AI inference provider operates under paid-tier terms under which submitted content is not used to improve its products or train its models. Providers may retain content briefly to detect and prevent abuse.
Call audio
We do not retain call audio. It is streamed to an AI inference provider to generate the persona's replies in real time, and is not stored by us. Only the transcript is kept.
See our Privacy Policy for everything we do keep.
If you think a score is wrong
Tell your organization's SkillTap administrator. They control the rubric, the personas, and the data, so they are the ones who can review the call, correct the record, or change how the rubric grades it.
We cannot revise a score on your behalf, because it is your employer's data and their rubric — but if the administrator finds a fault on our side, we want to hear about it at support@skilltap.ai.
Contact
FullBricks LLC 30 N Gould St Ste N Sheridan, WY 82801, USA