Skip to main content

This is a prototype vision of how a future government service could work. It's not a real service yet, but we're exploring what it could look like. Your feedback will help shape the real service.

Step 3: Upload & Process Document - Planning AI Walkthrough

Upload your planning application and watch AI extract information in real-time

Walkthrough progress

Step 3 of 4 • 2 minutes

Step 3 2 minutes

Upload & Process Document

Upload the planning application PDF and watch AI automatically extract applicant information in real-time.

Find your Planning AI application URL in CloudFormation Outputs
The application is ready to process planning documents

Expected outcome

  • Document uploads successfully
  • Processing progress indicator shows real-time status
  • Extraction completes in 8-20 seconds
  • Extracted fields appear automatically

Upload your document

  1. Locate the upload area

    On the document processing interface, find the upload section. It should say "Drag and drop your planning application here, or click to browse".

  2. Upload the PDF

    You can either:

    • Drag and drop: Drag the PDF file from your downloads folder and drop it on the upload area
    • Click to browse: Click the upload area, then select the PDF from your file browser
  3. Watch the upload progress

    You'll see a progress bar as the file uploads to AWS S3. For a 1-4 MB file, this typically takes 1-3 seconds.

  4. Wait for AI processing

    After upload completes, AI extraction begins automatically. You'll see:

    • Status changes to "Processing with Amazon Textract..."
    • Animated spinner or progress indicator
    • Estimated time remaining (8-20 seconds depending on document)
  5. Extraction completes

    When done, you'll see "Extraction complete!" and the extracted fields appear on screen.

What's happening behind the scenes:

  1. Your PDF uploads to a secure S3 bucket
  2. Lambda function triggers Amazon Textract API
  3. Textract performs OCR and identifies text regions
  4. AI models extract key-value pairs (applicant name, address, etc.)
  5. Results return to your browser in structured JSON

What to watch for

Pay attention to these impressive capabilities during processing:

Multi-page processing

AI reads all pages of the PDF, not just page 1. It finds information wherever it appears in the document.

UK address recognition

Accurately identifies UK postcodes, street addresses, and location formats. Handles "Land adjacent to..." descriptions.

Context understanding

Distinguishes between applicant name, agent name, site address, and billing address even when multiple names/addresses appear.

Speed at scale

Processing 200 applications/year takes 50 hours with AI vs 150 hours manually. No degradation in speed regardless of volume.

Processing time expectations

Different document types have different processing times:

Document Type File Size Processing Time
House Extension 1.2 MB 8-12 seconds
Change of Use 2.8 MB 12-18 seconds
New Dwelling 3.5 MB 15-20 seconds

Troubleshooting

File too large error (max 10MB)

If you see "File too large" error:

  • Check file size in file properties - must be under 10 MB
  • Use a PDF compression tool to reduce file size
  • Try splitting very large multi-page applications
  • Remove high-resolution images or photos if possible
  • Export at lower quality from your planning software

Our sample documents are all under 4 MB and should work without issues.

Wrong format error (PDF, JPEG, PNG only)

If you see "Unsupported file format" error:

  • Only PDF, JPEG, and PNG files are supported
  • Convert Word documents (.doc, .docx) to PDF first
  • Check the file extension is .pdf, .jpg, .jpeg, or .png
  • Don't upload compressed files (.zip, .rar)
  • Ensure you downloaded the PDF, not just opened it in browser
Processing timeout (over 30 seconds)

If processing takes longer than 30 seconds or stalls:

  • Refresh the page and try uploading again
  • Try a different sample document to rule out document corruption
  • Check AWS CloudWatch logs for Lambda timeout errors
  • Verify Textract API quotas haven't been exceeded
  • Wait 2-3 minutes then retry - may be temporary API slowness

Normal processing is 8-20 seconds. Anything over 30 seconds indicates an issue.

API errors (500, 503 responses)

If you see server error messages:

  • Wait 1-2 minutes and retry - may be temporary service issue
  • Check AWS Service Health Dashboard
  • Verify Lambda function has IAM permissions for Textract
  • Check Lambda execution role includes textract:AnalyzeDocument
  • Monitor CloudWatch Logs for detailed error messages
  • Contact support if errors persist
Upload succeeds but no extracted fields

If upload completes but extraction returns empty:

  • Check the document contains machine-readable text (not scanned handwriting)
  • Verify PDF isn't password-protected or encrypted
  • Try opening the PDF locally to confirm it's not corrupted
  • Check CloudWatch logs for Textract response errors
  • Try uploading one of our sample documents to rule out service issues
Note Screenshot placeholder: In production, this page would include screenshots showing the upload interface, processing progress animation, and the moment when extracted fields appear.

Need help?

If you're stuck, check the troubleshooting section or contact us.