How we use AI
Which features use AI, what each one receives, what we remove first, and where a person is in the loop. Written from what the code does, not from what we would like it to do.
Your assessment is not scored by AI. A rule-based engine computes the scores, profile and career matches. AI is used afterwards, to put results into words and to help with drafting and questions.
Where AI is used
Career interpretation (NexProfile™)
Puts your already-computed results into readable text.
- What it receives
- Your assessment results, including scores, and your age or year of birth and gender if you gave them.
- Who checks it
- Generated automatically and shown without a person reviewing it first.
Career chat
Answers questions about your profile and about careers.
- What it receives
- A summary of your profile (top interests, traits and job matches, with their scores), your age and gender, and what you type. Your name, email addresses and phone numbers are removed.
- Who checks it
- No review. The chat tells you AI answers may be inaccurate.
NexPilot™ (counselors)
Helps a counselor prepare for and run a case.
- What it receives
- The counselor's messages and case information: profile results including scores, plan status, age, gender and education. The participant's name is withheld.
- Who checks it
- The counselor reads every answer. Anything that would change something, such as creating a task or drafting a message, happens only after the counselor confirms.
Resume builder
Drafts a CV from your profile and what you give it.
- What it receives
- An assessment summary and the text and work history you type or import. Your name, email addresses and phone numbers are removed; the rest goes as you wrote it.
- Who checks it
- Saved as a draft that you edit, and marked as AI-generated.
Skill-gap analysis
Compares your skills with a target occupation.
- What it receives
- An assessment summary and the text and work history you provide. Your name, email addresses and phone numbers are removed; the rest goes as you wrote it.
- Who checks it
- Shown to you without a person reviewing it first.
Intake from free text
Reads an account written by or about a person and proposes profile fields.
- What it receives
- The text that was entered, with name, email address, username and phone numbers removed.
- Who checks it
- Suggestions only. They never overwrite something already entered.
What we remove first
Before we build an AI request we take out your name, email address and username, and the model is told to refer to “the participant”. Where a name belongs in the result, such as your own CV or results page, we put it back after the answer returns, inside our own systems.
What we send on purpose
Assessment results, age and gender go to the model because they change what good guidance is: a 15-year-old and a 17-year-old need different next steps. We judged that with name and contact details removed, they are needed for the purpose. Text you type into an AI feature, and work history you import, is sent with your own name, email addresses and phone numbers removed. The rest goes as you wrote it, so please leave out other people's names and details.
Who sees it
Every AI request goes through one service, OpenRouter, to commercial language models, for example from OpenAI and Google, which OpenRouter runs on the hosting providers it selects (for example Microsoft Azure). Each request tells OpenRouter to use only provider endpoints that do not collect prompts or train on them, and the same restriction is switched on in our OpenRouter account. If a model has no such endpoint, the request fails; it does not fall back to one that keeps data. OpenRouter is listed on the Sub-processors tab.
What we keep
Results that AI helps produce, such as interpretations, CV drafts and skill-gap analyses, are saved in your account, hosted in Helsinki. For profile interpretations we also keep the full request and response in a log in our own EU database to check quality and cost. The text of these logs is deleted after 30 days; the token and cost figures stay. The logs are not sent to the AI provider.
Limits, in plain words
- AI text can be wrong or too confident. Treat it as a starting point for a conversation, not as a decision.
- Nothing here is used to score you, admit you or reject you. The scores come from the rule-based engine.
- If you would like a person to go through any result with you, ask your counselor or write to us.
Last checked against the code:
Question about AI or your data?
Tell us what you would like to know and we will answer in writing.