AI governance, human oversight and explainability

1. Purpose of the Zeenna AI system

Zeenna supports the first stage of recruitment. It helps engage candidates, conducts structured pre-screening, generates follow-up questions, structures responses and prepares information for recruiters. Zeenna uses third-party large language models through APIs and does not train a foundation model from scratch.

2. Separation of AI and evaluation logic

A core architectural principle is the separation of conversational AI from the rules used to evaluate candidates. AI supports conversation, follow-up questions and descriptive summaries. Rule-based scoring applies criteria, rubrics and weights defined by the recruiter or customer. An authorised human makes the final candidate decision.

AI-generated summaries are not the source of an autonomous recruitment decision and are kept separate from rule-based scoring.

3. Human-in-the-loop and recruiter control

Zeenna is designed as a decision-support system, not an autonomous hiring tool. A recruiter or other authorised person remains responsible for human oversight.

• Reviews questions, responses and evaluation information.
• Defines or approves evaluation criteria and weights.
• May review, edit or disregard an AI-generated summary.
• May decide differently from the system’s supporting output.
• Makes the final decision on candidate progression, hiring or rejection.

4. Transparency to candidates

Candidates are informed before the interaction begins that they will be interacting with an AI system. Relevant disclosures are part of the product interface.

5. Explainability and auditability

The governance model is based on traceability of material steps. Depending on the configured workflow, explanation and audit may use the relevant input data, AI inputs and outputs needed for operational oversight, recruiter-defined scoring rules and weights, system and audit records, and human interventions.

The objective is to explain how Zeenna was used in a specific process, which rules affected the score and where human oversight was applied, rather than the internal mechanics of a foundation model.

6. AI risk management and impact assessment

Zeenna uses an AI System Impact Assessment process to identify and manage risks associated with AI-enabled workflows.

• Impacts on candidates and other stakeholders
• Fundamental-rights risks
• Fairness and potential bias
• Privacy and data protection
• Transparency and accountability
• Security, technical and operational risks

7. Bias and fairness controls

• Testing AI-generated questions and summaries
• Assessing the impact of significant changes on candidates
• Considering fairness within the AI impact assessment process
• Adjusting prompts or workflows where a material bias risk is identified

8. Roles and responsibilities

• Information Security Officer or CISO oversees AI security and compliance governance and approves relevant risk controls.
• AI Compliance Officer manages AI-specific risk assessments, transparency obligations, impact assessments and fairness controls.
• Engineering or Tech Lead is responsible for AI workflows, LLM API integrations, technical controls and operational monitoring.
• Product Management ensures human-in-the-loop workflows, interface transparency and separation of rule-based scoring from AI summaries.

9. AI change management

Material changes to AI workflows are governed through controlled development and approval processes. This includes prompt templates, model providers or routing, workflow logic, scoring logic and interface changes that may affect human oversight or transparency.

10. Monitoring and continual improvement

AI workflows are monitored on an ongoing basis. Zeenna evaluates operational reliability, workflow behaviour, user feedback and relevant incidents or deviations. Findings inform controlled system improvement.

11. Regulatory context

Zeenna treats AI in recruitment as a highly regulated use case and applies corresponding principles of human oversight, transparency, risk management and documentation. Specific legal obligations depend on the deployment context, the customer’s role and the applicable law.

12. Related documentation

• Zeenna Instructions for Use
• Zeenna Decision Process Architecture and AI Explainability Guide
• AIMS Manual and Scope Document
• AI Acceptable Usage Policy
• AI Security Policy
• AI Roles and Responsibilities
• AI SDLC Policy and Procedure
• AI System Impact Assessment Policy and Procedure
• AI Lifecycle Document