AI scoring checks each application against the criteria you set for a vacancy and gives it an AI-generated label with the reasons behind it. It helps recruiters decide which applications to read first. The label is advisory, so a recruiter must review the application before making any decision about a candidate.
📌 If you don't see the AI scoring panel when editing a vacancy, contact us at support@peopleforce.io to request access.
Set up AI scoring for a vacancy
AI scoring is set up separately for each vacancy and is off by default.
Open the vacancy and go to the Job description step.
Below the description editor, find the AI scoring panel and turn on Enable AI scoring.
Add your criteria in three lists, or let AI draft them for you by selecting Generate with AI. You need at least one criterion in any of the three lists. Otherwise, the vacancy can't be saved with AI scoring switched on:
Must have: what is essential for the role, such as education, experience or qualifications. For example, "6+ years of backend development experience".
Nice to have: skills or experience that are a plus but not essential. For example, "Experience working with SaaS companies".
Red flags: something in the application that a recruiter should check before moving the candidate forward. For example, “Experience limited to frontend development”. A red flag weighs against the overall label but doesn’t disqualify the candidate.
🤖 AI creates criteria from the vacancy title and job description, so the job description must be filled in first. Generating replaces any criteria already in the lists. Review and adjust the AI suggestions before saving.
Write criteria that can be checked against a CV. Specific criteria, such as “Professional working proficiency in English” or “Experience with PostgreSQL”, give clearer results than general ones like “Good team player”. Base criteria on job requirements only. Don’t write criteria about age, sex or gender, nationality, ethnic origin, religion, disability, health, pregnancy, family status or sexual orientation. Avoid proxies for them as well. For example, write “Degree in accounting or equivalent experience” instead of “Recent graduate”, and “3+ years of experience in B2B sales” instead of “No career gaps”.
Save the vacancy.
📌 PeopleForce doesn’t show candidates a notice that their application is assessed with AI.
How applications are assessed
Once AI scoring is on, every new application to the vacancy that has a CV is assessed automatically. It doesn’t matter where the candidate comes from: the career site, a job board, an import, the API, or adding them to the vacancy by hand.
The AI receives only:
the text of the candidate’s CV;
their answers to the vacancy’s application form, including the cover letter. If an answer on the form contradicts the CV, the form answer is used;
the vacancy’s criteria.
Each application then receives one of four labels:
Label | Meaning |
Brilliant | Evidence for all must-haves and most nice-to-haves, no red flags found |
Good | Evidence for all must-haves and some nice-to-haves |
OK | Evidence for the must-haves, but with gaps or a possible red flag worth checking |
No match | No evidence found for at least one must-have |
The AI assesses each criterion separately and then gives one overall label. There is no numeric score, formula or fixed weighting behind it. Each application is assessed on its own against the criteria, not compared with other candidates.
No match doesn’t reject, hide or move the application. The candidate may meet the requirement but not have mentioned it, or may have described it in words the AI didn’t recognise. Read the reason before deciding not to move a candidate forward.
Read the results
Candidate labels appear:
in the Vacancies section of the candidate's profile;
on candidate cards in the vacancy's pipeline;
in the AI match column of the vacancy's table view, which you can sort from the best match to the weakest;
on the candidate's print or PDF version.
Use the AI match filter on the vacancy, or on the Candidates page, to show only candidates with a specific label.
See the reasons
Select a candidate's label to open CV insights.
It shows:
the overall label and the reasoning behind it;
one line per criterion under Must haves, Nice to have and Potential red flags, each with its own explanation.
A tick always means good news: the AI found evidence of a must-have or nice-to-have, or found no evidence of a red flag.
Insights are written in your company’s default language, so every recruiter reads the same result. Criteria and CVs can be in different languages. The AI is instructed to match them by meaning and not to treat the language of the CV as a factor, unless a criterion is about language skills.
Assess candidates who are already on the vacancy
Switching AI scoring on doesn’t assess applications received before. To assess them:
On the vacancy's board, select the candidates.
In the bottom bar, open the ⋯ menu and choose Run prescreening.
Candidates without a CV, or with a CV that can’t be read, are skipped. Results appear on the cards as soon as they’re ready, and you don’t need to reload the page.
To assess a single candidate, open their profile and select Generate insights next to the vacancy.
Update screening results
To assess an application again, for example after changing the criteria, uploading a new CV or updating the candidate’s form answers, open CV insights and select Regenerate. The new result can differ from the previous one.
When candidates are not scored automatically
In these cases, use Run prescreening, Generate insights or Regenerate:
You switched AI scoring on after candidates had already applied.
You changed the criteria. Existing results aren’t updated automatically.
The candidate’s CV or form answers changed after they were added to the vacancy, for example, a new CV was uploaded.
What the AI doesn't see or do
Personal and sensitive profile fields aren’t sent. These include the candidate’s name, email, phone, Telegram username, personal links stored as contact details, desired salary, gender, date of birth, the assigned recruiter and record creator, fields marked as sensitive, and fields not shown on the application page. The AI also doesn’t use interview scorecards, tests, the pipeline stage, recruiter notes or other applications.
The CV and form answers are sent as the candidate wrote them. PeopleForce doesn’t remove anything from them. Most CVs start with a name and contact details, and some include details such as age, health, family situation, nationality or religion. The AI is instructed to assess only evidence relevant to your criteria and to disregard personal characteristics, but we can’t guarantee that this information has no effect on the result.
The AI doesn’t reject, disqualify or move candidates, contact them or make hiring decisions. It only adds an AI-generated label and insights. Red flags and unmet must-haves weigh against the label but never reject a candidate automatically.
Only one CV per candidate is read.
Each new run replaces the previous result. No history of earlier results is kept. If a run fails, the previous result is kept.
Results can differ between runs, and the AI can be wrong. Treat the label as a starting point for your review, not a decision. A recruiter must read the application before deciding anything about a candidate.
AI scoring is separate from AI CV parsing, the option that fills in a candidate’s profile from their CV.
Use the results through the API
The AI scoring result is included in candidate data in the PeopleForce API as read-only data: the label, the reasoning and the result for each criterion. You can use it in your own integrations and automations. For example, you can move candidates labelled Brilliant or Good to a review stage so a recruiter reads their applications sooner.
Automations you build on the API belong to your company, not to PeopleForce, and you’re responsible for human oversight of their results. Don’t use AI insights as the only basis for an automated decision to reject, exclude or hold back a candidate. More details in our developer documentation.