AI transparency and user control

Reading your recorded evidence and proposing skills; extracting stated requirements from a vacancy advert; comparing the two; drafting CV and letter wording from evidence you have already confirmed. Much of this is rule-based and runs in your browser. The assisted actions you choose to run — improving wording, drafting a letter, asking the adviser — are carried out on Mulyba's servers, which send the selected, confirmed content needed for that one request to an external model provider. Mulyba does not keep the request text or the provider's raw reply; it keeps a record that the action happened, and the finished result where the product is designed to save it for you, such as a CV version, a letter or an adviser report.

• Adding employment, qualifications, dates, employers or achievements you did not record.
• Deciding that a skill is true. You confirm, edit or reject every one.
• Altering the original text of a vacancy advert.
• Scoring your personality, tone, confidence or honesty.
• Predicting whether you will be hired.

A role is labelled ready now, possible with support, or a future target. That label is always accompanied by the count of essential requirements met, the specific evidence that met them, and the requirements that are unevidenced. If we ever cannot explain a result, we will not show it.

Reject any proposed skill and it disappears from matching and from every document. Reword any skill and your wording is used everywhere. Switch off AI assistance in the trust centre and the service becomes entirely manual. Nothing you reject is quietly reinstated.

Automated interpretation can be wrong, and it can be wrong unevenly across people. Mulyba supports human judgement — yours, and that of any adviser or work coach helping you. It never replaces it, and it never guarantees employment.