How Hyrpilot uses AI

An AI model is involved in the parts that need to read or write language. Everything it produces is shown to you before it goes anywhere.

The Match Score on a job card is not an AI opinion. It is a fixed calculation with weights we can write down: skills 50, location 20, remote preference 15, work type 15. The same profile and the same job always produce the same number, and the page can show you which part of it came from where.

When a posting does not carry enough information to score, Hyrpilot shows nothing rather than a number. A confident-looking percentage built on a blank posting is worse than an honest gap.

Job search, saved jobs, the application tracker, reminders, and everything about your account and billing run without a model.

By default, Hyrpilot pauses and asks you to approve the final submit for your first five applications. After five have actually gone through, the default switches to hands-off, and later applications are submitted without stopping to ask.

You can change this at any time in your settings:

• Guided: always pause and ask before submitting, no matter how many you have sent.
• Ramp (the default): guided for the first five, hands-off after that.
• Hands-off: submit without asking, from the first application.

Three things stay with you in every mode. A captcha is handed back to you. A login is handed back to you. And fields that look sensitive, or that look like a trap left for bots, are never filled in automatically.

Everything the agent did on a run is written down and readable afterwards, including the steps it took on your behalf and the moments it approved something itself.

To produce the text above, the relevant parts of your profile and the job description are sent to an AI model provider. That is the part that leaves our servers.

• We do not send your uploaded CV file. We send the text we extracted from it, and only the parts a given task needs.
• We do not send your password, your session, or any payment details.
• We do not send screenshots.

Our servers and databases run in Indonesia. The AI providers we call process data outside Indonesia, so a request may be handled in another country under the commitments described in our privacy policy.

Hyrpilot does not train, fine-tune, or improve any AI model using your data. Your CV is not training material, and neither are the applications you send.

We use commercial model providers through their paid interfaces. What they do on their side is governed by their own terms, which is why we keep the amount we send to what a task actually needs.

This is the part most worth reading. Language models produce text that reads as confident whether or not it is correct. In this product that shows up as:

• A cover letter that states something about you that is close to your CV but not quite what it says.
• A tailored CV that smooths a date or a title into something you did not write.
• A form answer that fits the question but not your situation.
• A fit analysis that reads more of a match into a vague posting than is there.

Read anything before it goes out with your name on it. That is the entire reason drafts are drafts, and the reason the first five applications pause and wait for you.

Every plan has a monthly ceiling on how much AI work it can use, and every single run has its own smaller ceiling on top of that. The ceilings keep one runaway job from consuming a whole month, and they are what makes the plan prices possible.

If you reach a limit, Hyrpilot tells you which limit you reached and what it would take to lift it. It does not quietly produce something worse.

AI features are opt-in per category, and you can withdraw any of them. Withdrawing consent for applying on your behalf stops it immediately, including runs already queued.

You can export everything we hold about you, or delete your account and the data with it. Both are self-serve.

Nobody at Hyrpilot reads your CV, your applications, or your conversations as a matter of course. A person may look at specific data only with your permission, where the law requires it, or to investigate abuse or a security incident.