Upload the policy. Let AI read it for you.
Estimatics Team
Estimatics
The policy document problem
An insurance policy is 40 to 120 pages. It contains the coverage limits, exclusions, endorsements, and conditions that determine what gets paid and what doesn't. Every claim depends on it.
And yet, most property professionals don't read it until there's a dispute.
Not because they're careless. Because reading a policy takes time. Extracting the relevant numbers takes expertise. Cross-referencing endorsements against base coverage takes attention that's better spent in the field.
So the policy sits in an email attachment. The inspector estimates based on experience, not coverage limits. The scope goes to the carrier without reference to what the policy actually says. And when the carrier comes back with "not covered under endorsement 47B," the scramble begins.
What Estimatics does with your policy PDF
During job creation, you can upload the insurance policy PDF. Estimatics sends it through an AI extraction pipeline that reads the entire document and returns structured data.
What gets extracted
Insurance information: company name, policy number, insured name and address, policy effective and expiry dates, assigned adjusters.
Coverage limits: Dwelling (Coverage A), Other Structures (Coverage B), Personal Property (Coverage C), Loss of Use (Coverage D) — each with specific dollar limits.
Water and wind coverage: sublimits, exclusions, deductible structures for water damage and wind/hurricane events.
Appraisal and mediation rules: whether the policy requires appraisal before litigation, mediation provisions, umpire selection procedures.
Mitigation and repair programs: managed repair networks, preferred vendor requirements, mitigation coverage limits.
Endorsements: every endorsement on the policy, flagged as controlling or supplementary.
Contradictions: if the AI finds conflicting information between different sections of the policy, it flags the contradiction with both sources and a resolution note.
Confidence metadata
Every extracted field comes with a confidence level. The AI tells you how certain it is about each number, each date, each coverage determination. High confidence means the text was clear and unambiguous. Lower confidence means you should verify manually.
OCR support
The extraction works on both digital and scanned PDFs. For scanned documents, the AI performs OCR first, then extraction. The system flags the document quality so you know if the source material was clean or degraded.
What it means for your workflow
When you open a job after the policy has been processed, the insurance information fields are already populated. Coverage limits are filled in. Deductibles are recorded. Endorsements are flagged.
You didn't type any of it. You didn't read 80 pages of legal text. You uploaded a PDF and the AI did the data entry.
More importantly: when you build your scope, you know the coverage limits before you price a single line item. When the carrier questions a line item, you can reference the specific policy section that covers it. When an endorsement changes the coverage, you know about it before the carrier tells you.
The compound value
Policy extraction isn't just convenience. It's intelligence.
The AI that analyzes your photos for damage also knows the policy coverage. When it generates findings, it can flag whether each damage type is covered, not covered, or requires review under the policy terms.
A finding that says "wind damage to roof shingles — covered under Coverage A, $350,000 limit, $5,000 deductible" is fundamentally different from a finding that says "wind damage to roof shingles." The first is actionable. The second is a starting point.
Available during job creation
Policy PDF upload is available during step 7 of job creation on all plans. The extraction runs asynchronously — you can continue working while it processes. Status indicators show when extraction is complete.
For best results, upload complete, legible policy documents. The AI handles multi-page policies, but clearer source material produces higher-confidence extractions.