Extract data from certificates of insurance

A Certificate of Insurance is the document a vendor sends a counterparty to prove that the vendor carries insurance coverage. The form most US companies use is ACORD 25 (the standardized layout maintained by ACORD, the insurance-industry standards body), though variations exist for marine, aviation, and international coverage. The certificate names the insured (the vendor), the certificate holder (the counterparty receiving the proof), the producer (the insurance broker), and the insurers (often three or four separate carriers, each covering a different policy line such as General Liability, Auto Liability, Workers Compensation, and Umbrella Excess). Each policy line carries its own policy number, effective date, expiration date, coverage limits, and an indicator for whether the certificate holder is named as an Additional Insured. Procurement, risk management, and vendor compliance teams collect, review, and renew these certificates on a continuous cycle to keep the vendor base compliant under master service agreements and contractor agreements. The hard parts are the policy table. ACORD 25 packs four to seven policy lines into a small table on a single page, each row with abbreviated coverage type (GL, AL, WC, UMB), policy number, effective/expiration dates, and a limits column that mixes per-occurrence and aggregate values stacked vertically. NAIC codes appear as small integers next to each insurer's name. Description of Operations is a free-text block that often contains the specific contract reference, the named additional insureds, and any waiver of subrogation language. Schedule of forms and endorsements (the small print on the back) modifies what each policy actually covers and which entities are protected. Certificate holders track all of this for every active vendor, and the renewal cycle drives the workload because expiration dates are rolling. Talonic extracts the full certificate structure: every policy line as a row, every limit as a typed numeric value, every insurer with its NAIC code, the producer and insured, the certificate holder, the description of operations block, and any waiver indicators. Effective and expiration dates surface as ISO 8601 dates so a renewal calendar can be built directly from the data. Every cell carries a confidence score and a pixel-region pointer so the compliance officer can verify a limit or an additional-insured flag against the source ACORD 25 before approving the vendor.

What gets extracted from certificates of insurance

How extraction works for certificates of insurance

ACORD 25 is the most common Certificate of Insurance form, but each carrier and broker renders it slightly differently. Talonic classifies the certificate (ACORD 25, marine COI, aviation COI) and runs it through the CoI schema in the Field Registry, which encodes the producer, insured, certificate holder, insurer table, policy line table, and description of operations. Coverage types are normalized to a canonical set (GL, AL, WC, EXCESS, PROFESSIONAL, CYBER). NAIC codes are validated against the NAIC publicly available carrier list. Limit values are typed numerically and labeled (per occurrence, aggregate, products/completed operations). Effective and expiration dates are returned as ISO 8601 so a renewal alerting workflow can fire automatically as expirations approach. Per-cell confidence and pixel-region provenance follow DIN SPEC 91491 conformity, so a compliance officer can audit any policy line against the source ACORD 25 before approving the vendor for active engagement.

Sample extraction

A standard ACORD 25 with General Liability, Auto Liability, and Workers Comp coverage

{
  "producer_name": "Marsh & McLennan Agency LLC",
  "insured_name": "Goodwin Retail Co.",
  "insured_address": "88 Commerce Way, Portland, ME 04101",
  "certificate_holder_name": "1200 Market Holdings LLC",
  "certificate_holder_address": "1200 Market Street, San Francisco, CA 94102",
  "issue_date": "2026-01-08",
  "insurers": [
    {
      "letter": "A",
      "name": "Chubb Insurance Co. of NJ",
      "naic_code": "22667"
    },
    {
      "letter": "B",
      "name": "Liberty Mutual Insurance",
      "naic_code": "23043"
    }
  ],
  "policies": [
    {
      "insurer_letter": "A",
      "coverage_type": "General Liability",
      "policy_number": "GL-2026-44892-NA",
      "effective_date": "2026-01-01",
      "expiration_date": "2027-01-01",
      "limits": [
        {
          "limit_type": "Per Occurrence",
          "limit_amount": 2000000
        },
        {
          "limit_type": "General Aggregate",
          "limit_amount": 4000000
        }
      ]
    },
    {
      "insurer_letter": "A",
      "coverage_type": "Auto Liability",
      "policy_number": "AL-2026-11203-NA",
      "effective_date": "2026-01-01",
      "expiration_date": "2027-01-01",
      "limits": [
        {
          "limit_type": "Combined Single Limit",
          "limit_amount": 1000000
        }
      ]
    },
    {
      "insurer_letter": "B",
      "coverage_type": "Workers Compensation",
      "policy_number": "WC-2026-77891-ME",
      "effective_date": "2026-01-01",
      "expiration_date": "2027-01-01",
      "limits": [
        {
          "limit_type": "Each Accident",
          "limit_amount": 1000000
        }
      ]
    }
  ],
  "description_of_operations": "1200 Market Holdings LLC is named Additional Insured per attached CG 20 10."
}

Frequently asked