From OCR to AI Evaluation: How a UK Awarding Body Transformed Handwritten Script Processing

The Client 

A UK Awarding Body offers professional qualifications across business, management, and related vocational areas through a network of approved centres. As an Ofqual-recognised UK awarding organisation, it conducts handwritten assessments that require the evaluation of candidate answer scripts and consistent scoring at scale.

Business Needs

The awarding body needed a scalable way to evaluate candidate scripts from scanned handwritten responses, while reducing three interconnected risks:

1. Handwriting variability & scan quality: Variations in legibility, overwriting, faint text, and scan artefacts lead to inconsistent interpretation, which basic OCR cannot handle.
2. Candidate intent, response mapping, and auditability: Crossed-out answers, multiple rewrites, and inconsistent question numbering make it difficult to identify the final response.
3. Variance in marking standards: Variations in how evaluators interpret responses can lead to differences in scoring, including instances where weaker answers receive higher marks than expected.

The awarding body needed something fundamentally more capable than traditional OCR — a system that could not only transcribe handwriting accurately, but make every interpretive decision visible, auditable, and consistent.

Solution Highlights

1. AI transcription with analytical validation: Accurately transcribes handwritten scripts across various response types, handling the complexity of real exam scenarios.
2. Handling candidate intent and mapping: Identifies final intended answers despite multiple edits, cross-outs, and rewrites, and reliably maps responses to the correct questions.
3. Standardized, transparent output: Provides a clean, neutral, and consistently formatted transcript for examiners, with full visibility into how the transcript was generated.