Feature Information

How does AI Skills Matching Work?

The AI Skills Matching feature utilizes a large language model (LLM) generative AI system to assess a job applicant’s suitability by comparing their resume or CV against a job ad.

This qualitative evaluation is performed through a three-step process:

1. The AI system analyses job ads to extract relevant skills, assigning importance and proficiency weightings based on AI inference. Users have the option to review and modify the skills inferred from the job ad.
2. The AI system analyses resumes to identify an applicant’s skills and weigh proficiency in identified skills.
3. The AI system then compares the extracted job skills with applicants skills (from the applicant's resume) to evaluate the applicant's fit for the job. The system provides five classifications: Excellent Fit, Good Fit, Potential Fit, Poor Fit and Not a Fit for each applicant.

What are appropriate ways of using AI Skills Matching?

It is important to note that LLMs are fundamentally probabilistic and can produce variable results. The AI Skills Matching feature is intended solely as an evaluative tool to support our customer’s recruiters and hiring managers during the recruitment process. The classifications produced by the AI Skills matching feature are limited to the job skills matched against the resume skills, and do not take into account other factors used in human decision-making.

The AI Skills Matching evaluation should not be used as the sole basis or most important factor for making decisions about a job applicant's recruitment, and should not be used to overrule human decision-making.

LLM Model

The AI Skills Matching feature is built on AWS’s Amazon Bedrock LLM (Anthropic based models). Amazon Bedrock doesn't store or log prompts and completions, and doesn't use prompts and completions to train any AWS models and doesn't distribute them to third parties. AWS is a current PageUp subprocessor.

Australia

Automated Decision Making

From 10 December 2026 entities regulated by the Australian Privacy Act 1988 and the Australian Privacy Principles ("APPs"), must ensure their privacy policies contain information about automated decision making which may significantly affect individuals’ rights or interests. This requirement extends to computer programs which do a thing that is substantially and directly related to making such a decision.

PageUp's AI Skills Matching feature utilizes an applicant's personal information – CVs/resumes – to aid our customers in making hiring or recruitment decisions. Consequently, this process may fall under the scope of APPs 1.7 and 1.8, referred to as the "Automated Decisions APPs".

PageUp offers the following information to assist customers align their privacy policies with the requirements of the Automated Decisions APPs.

Personal Information Used by AI Skills Matching

The AI Skills Matching feature may utilize personal information found in resumes and CVs. This could include:

• Name
• Employment history
• Educational history
• Professional and educational certifications
• Interests and hobbies
• Current and former address
• Email address
• Phone number.

Decisions Supported by AI Skills Matching

Our customers are assisted by the AI Skills Matching feature in making decisions including:

• Recruitment process decisions
• Hiring decisions
• Remuneration decisions
• Promotion decisions.

United States of America

California

From 1 October 2025 new regulations take effect to address employers' lawful use of artificial intelligence and automated-decision systems under California's Fair Employment and Housing Act.

We note that it is unlawful for an employer or other covered entity to use an automated-decision system or selection criteria (including a qualification standard, employment test, or proxy) that discriminates against an applicant or employee or a class of applicants or employees on the basis of a protected characteristic.

PageUp can provide the results of anti-bias testing conducted annually on the AI Skills Matching feature which may be relevant to potential claims or defenses available under the regulations.

Colorado

From 30 June 2026, Senate Bill 24-205, Colorado’s AI Act imposes obligations on "Developers" of "High Risk Artificial Intelligence Systems". As a potential Developer of a high risk AI system, PageUp makes the following documentation available to its customers who may be classified as a "Deployer" of a High Risk AI System.

Statutory Statement on Uses and Harms of the AI Skills Matching Feature

The AI Skills Matching feature is intended solely as an evaluative tool to support our customer’s recruiters and hiring managers during the recruitment process. The classifications produced by the AI Skills Matching feature are limited to the job skills matched against the resume skills, and do not take into account other factors used in human decision-making.

The AI Skills Matching evaluation should not be used as the sole basis or most important factor for making decisions about a job applicant's recruitment, and should not be used to overrule human decision-making.

Statutory Disclosure on Training Data, Limitations, Purpose, Intended Uses of the AI Skills Matching Feature

The AI Skills Matching feature is built using pre-trained Large Language Models via AWS Bedrock; PageUp does not use customer data to train or fine-tune these models. Customer data processed by the feature, such as information from resumes or job descriptions, is not stored, logged, or used by AWS to train its underlying models.

The AI Skills Matching feature works with objective, job-related characteristics, such as relevant career experience, skills and qualifications to generate an qualitative evaluation which aims to predict how a given applicant fits the job skills criteria. This helps recruiters and hiring managers to triage applicant profiles when deciding whether or not to move a candidate forward in a recruitment process. When used correctly, the AI Skills Matching feature can increase recruitment process efficiency and reduce recruitment costs for employers. The AI Skills Matching feature does not replace human decision making and cannot make informed hiring decisions.

The AI Skills Matching evaluation is applied to all candidates equally, regardless of their demographic or other protected characteristics; no candidate demographic or biometric information is used by the feature to generate its evaluation.

Performance Testing & Intended Outputs

PageUp conducts internal testing using synthetic data and external reviewers to validate the AI Skills Matching feature’s accuracy and reliability.

PageUp uses expert external recruiters to rate the accuracy of the AI Skills Matching feature’s skills extraction:

PageUp also conducts quantitative testing using sets of synthetic resumes which expert recruiters have previously rated with respect to a synthetic job ad: Excellent Fit, Good Fit, Potential Fit, Poor Fit and Not a Fit, using the following classification thresholds:

Precision assesses how many of the positive predictions made by a model were actually correct, while Recall measures how many of the true positive cases in the dataset were successfully identified by the model.

Recall (or "true positive rate") of the AI Skills Matching in meeting the classification thresholds, i.e., as applied to the AI Skills Matching feature, this test will reveal whether the feature has misclassified the fit of a candidate’s resume to a job description by giving a worse fit classification – (a false negative):

Precision of the AI Skills Matching feature in meeting the classification thresholds, i.e., as applied to the the AI Skills Matching feature, this test will reveal whether the feature has misclassified the fit of a candidate's resume to a job description by giving a better fit classification – (a false positive):

Discrimination Testing and Mitigation

Prior to release, PageUp undertook a comprehensive research and testing phase to evaluate the AI Skills Matching feature. This process benchmarked the system's accuracy and its alignment with how a recruiter would manually match skills between a job advertisement and a candidate's resume. The next phase included an early adopter program with five customers, allowing for testing with real-world job data.

Our team also conducted internal testing for bias across protected classes, including gender, country of education (as a proxy for nationality), and year of education completed (as a proxy for age), with no issues found. Further independent testing will be completed with a third-party bias auditor in Q1 2026.

To mitigate known and foreseeable risks of algorithmic discrimination, PageUp has implemented several key measures as part of the system's core design:

Focus on Skills: The AI is specifically instructed to focus on matching objective, job-related skills, experience, and qualifications rather than personal characteristics.

• No Training on Historical Data: The AI is not trained on any historical hiring data from PageUp customers. This is an intentional choice to ensure its evaluations are free from any biases that may exist in previous hiring decisions. The model also does not learn from your team's ongoing decisions, ensuring a consistent evaluation standard.
• Human-in-the-Loop Control: The feature is designed to assist, not replace, human judgment. Recruiters are empowered to review and modify the skills the AI extracts from a job description before any applicant matching occurs. Users can also directly override the AI’s final candidate classification with their own evaluation.
• Individual Evaluation: The AI evaluates each applicant individually against the job requirements on their own merits. It does not rank or compare candidates against each other, which helps avoid common human biases such as contrast and confirmation bias.

New York City

New York City Local Law 144 (LL 144) prohibits employers and employment agencies from using an automated employment decision tool (AEDT) to substantially assist decision making unless conditions are met. To "substantially assist" decision making is defined by the regulation to mean the AEDT is used as the sole, overriding or most important factor in decision making.

PageUp recommends against using the AI Skills Matching feature as the sole, or most important factor when making recruitment decisions, or using the AI Skills Match to overrule human decision making.