PIP2 Form Analysis
Department: DWP
Problem to Solve
Use OCR and AI interpretation to automatically read handwritten PIP2 forms, extract structured information, and map it to the official PIP scoring criteria. The goal is to reduce repetitive manual work, improve consistency, and give caseworkers a clean, traceable summary to review—without changing the human decision-making step.
Proposed Solution
Use OCR and AI to automatically read handwritten PIP2 forms, extract structured information, and generate draft scoring—reducing manual workload while keeping caseworkers in control.
Where This Challenge Fits
This challenge focuses on the form processing and evidence review stage of the PIP assessment journey
Start the claim
Claimant
Claim registration
DWP
PIP2 form issued
DWP
PIP2 form completed
Claimant
Form reviewed
DWP
Health assessment
Claimant/DWP
Decision made
DWP
Decision received
Claimant
Start the claim
Claimant
Claim registration
DWP
PIP2 form issued
DWP
PIP2 form completed
Claimant
Decision received
Claimant
Decision made
DWP
Health assessment
Claimant/DWP
Form reviewed
DWP
See it in Action
Resources
What we did
- ✓Researched what was publicly available about PIP2 forms and their structure
- ✓Reconstructed HTML mockups of limited sections of the form to match the real layout
- ✓Created a script for friends and family to follow when filling out the forms by hand
- ✓Collected diverse handwriting samples across different writing styles and qualities
- ✓Built an AI pipeline to recognise handwriting and extract all key data fields accurately
- ✓Tested extraction accuracy across various handwriting qualities and form completion patterns
Outcome
The system successfully recognised handwriting across diverse writing styles and accurately extracted structured data from handwritten forms, even with challenging handwriting or partially completed sections.
We achieved high accuracy in extracting key information fields, including medical conditions, daily living activities, and mobility descriptions - demonstrating that AI can meaningfully reduce manual data entry work.
This proof-of-concept showed that modern vision AI can handle the variability of real-world handwritten forms, making automated processing of PIP2 forms technically feasible and potentially transformative for processing times.
Technical Insights
Key Technologies
- • Vision AI for document analysis
- • Handwriting recognition (OCR)
- • Form structure understanding
- • Field extraction and validation
- • Data normalisation pipelines
Challenges Solved
- • Variable handwriting quality
- • Form layout recognition
- • Partial completions
- • Multi-line text extraction
- • Checkbox and tick detection
Testing Methodology
By using friends and family to complete the forms following a standardised script, we ensured:
- •Diverse handwriting samples: Different ages, writing styles, and pen types
- •Realistic completion patterns: Natural variations in how people fill out forms
- •Controlled content: Known ground truth for accuracy testing
- •Real-world variability: Testing edge cases like cramped writing and corrections