ACORD forms and loss runs, read the way underwriters read them.
The two document classes that decide commercial submissions - ACORD applications and carrier loss runs - are parsed natively. Field-level accuracy, structured claim histories and clean handoffs to PolicyCenter, Duck Creek or your MGA platform. No re-keying, no spreadsheet glue.
Every ACORD field, mapped to your submission schema.
Brokers send the same business in a dozen formats - typed PDFs, scans, emailed snippets, partials. We treat each ACORD form as a structured contract, not an image, and extract it accordingly.
Forms supported out of the box
Other commercial and personal-lines ACORD forms can be added without retraining the base model.
The pipeline, step by step
Loss runs, turned into a clean claim ledger.
Loss runs are how prior carriers describe risk - and they are almost never consistent. We parse them into a structured ledger that underwriters and actuaries can actually price on.
The realities of carrier loss runs
- Every carrier and TPA uses a different layout
- Open vs closed status conventions vary
- Reserves and paid amounts split across columns
- Multi-year runs stitched into one PDF
- Aggregate rows mixed with claim-level rows
- Currency, dates and cause-of-loss codes inconsistent
Structured claim ledger
One row per claim with date of loss, status, cause, paid, reserves, recoveries and adjuster notes - across all years and carriers in the submission.
Frequency and severity curves
Loss counts and severity bands per year, per coverage, per location - ready for actuarial review and pricing models.
Trend and development view
Loss development factors, large-loss flags and year-on-year frequency shifts surfaced without manual triangulation.
Risk signals
Repeat causes, single-location concentration, open large claims and adverse development flagged for the underwriter.
Where the structured data lands.
Document Intelligence is not a parser in isolation. The output drops directly into the agents and systems that do the next bit of work.
- Underwriting triageAppetite scoring, referral routing and risk-band assignment from the clean ACORD record.
- Pricing and actuarialLoss ledger and severity curves feed reserving models and pricing notebooks without re-keying.
- Submission clearanceDuplicate broker submissions, prior declines and conflicts surfaced before an underwriter opens the file.
- Core system writebackMapped records pushed to Guidewire PolicyCenter, Duck Creek Policy, Majesco or your MGA platform.
- Audit and evidenceEvery field links to its source page and bounding box - regulator and reinsurer reviews stay defensible.
Defensible from the first extraction.
Field-level confidence
Every extracted field carries a confidence score. Anything below your threshold routes to human review.
Source linkage
Each value links back to the page, line and bounding box in the original document. Audit-ready by default.
Versioned schemas
Form layouts and your downstream schemas are versioned. Re-process historic submissions whenever logic changes.
In-region processing
Documents stay in your tenant - cloud, on-premise or air-gapped. No data leaves the perimeter your regulator approved.
ACORD forms, loss runs and submission AI - common questions.
Which ACORD forms does Yukthi Labs process out of the box?
ACORD 125 (Commercial Insurance Application), 126 (General Liability), 127 (Business Auto), 130 (Workers Compensation), 140 (Property) and 141 (Property Schedule). Other commercial and personal-lines ACORD forms can be added without retraining the base model.
How accurate is ACORD form extraction?
Field-level extraction averages 97% accuracy across typed, scanned and partially completed ACORD forms. Every field carries a confidence score, and anything below your threshold routes to human review.
How are carrier loss run reports structured?
Loss runs from any carrier or TPA layout - including scanned and multi-year stitched PDFs - are parsed into a normalized claim ledger with date of loss, status, cause, paid, reserves and recoveries. Outputs include frequency and severity curves, loss development views and large-loss flags ready for underwriting and actuarial review.
Does the structured output integrate with Guidewire, Duck Creek and MGA platforms?
Yes. Extracted ACORD records and loss run ledgers are mapped to your submission schema and written back to Guidewire PolicyCenter, Duck Creek Policy, Majesco or your MGA platform - no re-keying required.
Where is document processing hosted?
Documents stay in your tenant - cloud, on-premise or air-gapped - so submissions never leave the perimeter your regulator approved. Field-level source linkage keeps every extraction audit-ready for regulators and reinsurers.
Tell us the workflow that's costing you the most.
In time. In headcount. In customer experience. We'll tell you whether we can move it in 6–8 weeks - or we'll tell you we can't.