Responsible AI Policy

We are committed to developing and deploying artificial intelligence systems responsibly, ethically, and transparently. Our AI technologies are designed to augment human decision-making in HR and compensation management while maintaining the highest standards of fairness, accountability, and user trust.  

Core Principles:  

• Human-centred design – AI supports and enhances human expertise, not replaces it 

• Transparency – we clearly explain how our AI works and its limitations 

• Fairness – we actively work to prevent bias and ensure equitable outcomes 

• Accountability – we take responsibility for AI-driven decisions 

• Privacy by design – we protect user data from the ground up 

Natural Language Processing (NLP): 

• Analysing job documents (profiles, descriptions, adverts) 
• Evaluating and improving content clarity and completeness 

Content generation: 

• Our AI tools support users in creating job documentation. All outputs are reviewed and approved by humans before being shared.

Job architecture and evaluation: 

• Job comparisons and benchmarking 
• Job levelling (in development) 
• Employee-to-job mapping 

• We use OpenAI (via Microsoft Azure), Claude (Anthropic), and in-house models. All data processing is hosted securely within UK-based AWS and Azure infrastructure. 
• We carry out due diligence on all third-party AI providers and maintain contracts to ensure compliance with UK data protection standards. 

What we process: 

• Job profiles, descriptions, adverts, and job architecture data 
• Minimal employee details: first name, last name, email address 
• Salary range data if supplied 

We explicitly exclude any processing of protected characteristics such as age, gender, ethnicity, or disability. 

We rely on a mix of contractual necessity, legitimate interests, and consent, depending on the context. This aligns with Article 6 of UK GDPR. 

How we protect data: 

• All data processed within UK regions 
• Data anonymised before model training 
• Clear deletion procedures 
• Retained for no more than 12 months after contract end 

Privacy controls: 

• Data minimisation and purpose limitation 
• Consent where appropriate 
• Regular privacy impact assessments 
• No automated decisions with legal or significant effects are made without meaningful human involvement (Article 22 UK GDPR) 

Our approach includes: 

• Excluding protected characteristics from processing 
• Pre-deployment fairness testing 
• Diverse training data to prevent systemic bias 
• Monitoring for fairness and accuracy 

We also keep detailed records of all testing, take corrective action where needed, and publish fairness reports. 

Infrastructure security:  

• UK-based AWS and Azure environments 
• End-to-end encryption 
• Regular security testing 

Model security: 

• Version control and change management 
• Access controls 
• Secure deployment practices 

Data security: 

• Role-based access 
• Full audit logging 
• Anonymisation for training 
• Tested disaster recovery 

Human Oversight & Control 

Humans are always in the loop: 

• All AI-generated content is reviewed and approved by people 
• Clear escalation paths for concerns 
• Users can override or reject AI suggestions 

Responsibility and review: 

• All AI governance and oversight is led by Operations. 
• Technical and Customer Success teams monitor AI behaviour and user feedback. 
• Regular cross-functional reviews of performance and risks are held to support continuous improvement. 

We clearly label AI-generated content, provide simple explanations of how features work, and share updates about improvements. 

We also provide: 

• Decision explanations and confidence scores 
• Documentation of decision logic 
• Clear guidance on when and how to use AI features responsibly 

We comply with: 

• UK GDPR 
• Relevant employment law 
• ICO guidance and emerging UK regulation 

We align with: 

• The UK Government’s transparency and accountability framework for AI 
• Industry best practices and customer-specific compliance requirements 

We maintain full documentation and audit trails and regularly review our approach. 

We:  

• Test all features in development and staging 
• Check for bias and accuracy before release 
• Monitor live performance 
• Update models and retrain regularly 

We collect feedback from users and stakeholders and review this policy annually to reflect new risks and regulations. 

If something goes wrong: 

• Escalation paths are in place 
• A rapid-response team handles incidents 
• Customers are informed 
• Root cause analysis is done for every issue 

We assess and monitor risks regularly and use alerts and logs to spot and resolve potential issues early. 

Structure and ownership:

• Governance of AI at RoleMapper is led by Operations. 
• Technical teams operate and monitor systems. 
• Regular reviews take place across functions to ensure we meet our commitments. 

Policy enforcement: 

• All staff receive training 
• Breaches are dealt with appropriately 
• Compliance is audited regularly 

Stakeholder engagement: 

• Customers are kept informed 
• We’re open about any issues 
• We work with regulators and industry bodies 
• Feedback channels are available and encouraged