You need to annotate, classify or review data with reproducible quality.
HUMAN QUALITYEXPERTISE 04 / 04
Human operations & accessibility
SASECNUM organises secure human operations to train, test and evaluate AI systems, and conducts manual accessibility assessments of digital products and documents.
WHEN WE HELP
An AI system must be tested against real, ambiguous or sensitive cases.
Automated accessibility checks are not enough to understand the user experience.
SERVICE OFFERINGS
Clearly framed engagements.
Tangible outcomes.
Each service can be delivered independently or combined within a progressive programme. Scope, stakeholders and deliverables are agreed before work begins.
01↗
Data annotation & preparation
Design and execute structured human tasks with clear instructions, quality controls and traceability.
- Annotation protocols
- Dual review
- Quality monitoring
02↗
Human evaluation of AI
Compare outputs, score performance, review relevance, detect errors and explore edge cases.
- Evaluation rubrics
- Test sets
- Error reports
03↗
Manual accessibility testing
Assess user journeys, components and content with keyboards, screen readers and applicable criteria.
- Journey audit
- Functional testing
- Barrier prioritisation
04↗
PDF document accessibility
Review structure, reading order, alternatives, forms and assistive-technology rendering.
- Non-conformity report
- Remediation guidance
- Verification testing
POSSIBLE DELIVERABLES
Deliverables designed to be used — not merely archived.
Deliverables are presented to decision-makers, handed over to accountable owners and designed to support effective action.
- 01Execution protocols and quality instructions
- 02Controlled and traceable datasets
- 03Human-evaluation reports
- 04Prioritised accessibility reports
- 05Recommendations and verification testing
HOW WE WORK
A clear path from the first conversation to autonomy.
Frame
Understand the objectives, scope, stakeholders and expected level of ambition.
Assess
Gather evidence, analyse practices and qualify the risks.
Build
Co-design the responses, deliverables and governance mechanisms.
Embed
Enable the teams, monitor progress and adjust over time.
APPLICABLE FRAMEWORKS
Human in the loopQuality assuranceWCAG 2.2RGAA where relevantAssistive-technology testing
Frameworks are selected according to the country, sector, applicable obligations and objectives of the engagement.
Bring the judgement, context and sensitivity that automation alone cannot guarantee.
A FIRST CONVERSATION