Policy PDF Upload & AI Extraction

Insurance policies are dense, multi-page documents full of coverage limits, exclusions, endorsements, and legal language. Manually reading through a policy to find the numbers you need takes time and introduces the risk of missing something important. Estimatics solves this by letting you upload the policy PDF and having AI extract the structured data automatically.

What policy extraction does

When you upload an insurance policy PDF, Estimatics AI reads the entire document and pulls out the data points that matter for your inspection work. The system parses declarations pages, coverage schedules, endorsements, and exclusion sections to build a structured summary of what the policy covers and what it does not.

How to upload

There are two places to upload a policy document:

During job creation (Step 7) — the final step of the job creation form includes a file upload area where you can drag and drop or browse for the policy PDF.
From the job detail view — after a job is created, navigate to the job detail page and find the Policy section. You can upload or replace a policy document at any time.

Select your PDF file and the upload begins immediately. There is no file size limit for typical policy documents, though shorter processing times come with cleaner, well-formatted files.

What gets extracted

The AI extracts a comprehensive set of fields from the policy document:

Policy Number — the policy identifier from the declarations page.
Insured Name — the name of the policyholder.
Property Address — the insured property location.
Coverage A (Dwelling) — the insured value of the main structure.
Coverage B (Other Structures) — limits for detached structures such as garages, fences, and sheds.
Coverage C (Personal Property) — coverage for the policyholder's belongings.
Coverage D (Loss of Use) — additional living expenses while the home is uninhabitable.
Water Damage Coverage — whether water damage is included and any applicable sub-limits or conditions.
Wind / Hail Coverage — wind and hail provisions, including any separate deductibles.
Exclusions — specific perils or conditions the policy does not cover.
Endorsements — amendments to the base policy that add, remove, or modify coverage.
Appraisal Clause — the rules governing the appraisal process if the policyholder and carrier disagree on the loss amount.

Each extracted value is mapped to the corresponding field on the job record.

Processing time

Policy extraction runs asynchronously. After you upload the PDF, the system queues the document for processing. You can continue working on other tasks while extraction runs. The job detail view shows a status indicator:

Processing — the document is being analyzed.
Complete — extraction is finished and the data is available.

Most policy documents are processed within one to two minutes. Longer or more complex documents may take slightly more time.

Confidence metadata

Every extracted field includes a confidence level that indicates how certain the AI is about the value. High-confidence fields were found in clearly structured sections of the document. Lower-confidence fields may come from ambiguous language or unusual formatting. You can review confidence levels on the job detail view and manually verify any field that the system flags as uncertain.

Contradiction detection

Insurance policies sometimes contain conflicting information — for example, an endorsement that modifies a coverage limit stated on the declarations page, or overlapping exclusion language. The AI flags these contradictions so you can review them and determine which value applies. Flagged contradictions appear as warnings alongside the extracted data.

OCR support

Estimatics handles both digital and scanned policy PDFs. If the uploaded document is a scanned image rather than a text-based PDF, the system runs optical character recognition (OCR) before extraction. OCR converts the scanned pages into machine-readable text, allowing the AI to extract the same fields it would from a digital document. For best results with scanned policies, upload documents that are clearly legible and not heavily skewed or marked up.

How extracted data auto-fills job fields

Once extraction is complete, the extracted values automatically populate the corresponding fields on the job record — policy number, coverage limits, deductible, and more. If you had already entered values manually during job creation, the system presents the extracted values alongside your entries so you can choose which to keep. No data is overwritten without your review.

Tips for best results

Upload the complete policy document. Partial uploads — such as only the declarations page — will produce incomplete extraction. Include all pages, especially endorsements and schedules.
Use clear, legible files. High-resolution scans and native digital PDFs produce the most accurate results.
Review flagged contradictions. When the AI detects conflicting information, take a moment to verify the correct value. This protects you during appraisal or dispute resolution.
Upload early. Running policy extraction before you begin your AI analysis gives the system full context about what the policy covers, which leads to more relevant findings and coverage-aware reports.