Heron launches 100% coverage for parsing loss runs

For insurance companies like carriers and MGAs, parsing loss runs has long been a critical yet manual and time-consuming process. These documents hold essential data about claims, payments, and reserves — information vital for assessing risk and generating quotes.

Today, we’re launching 100% coverage loss run parsing — the ability to reliably parse all information from any loss run. This works by combining automated systems with domain experts where needed to enable straight-through processing for real-time insurance quoting.

Why automate loss run parsing?

Loss run reports often come in varied formats and with inconsistent data quality and levels of detail. Parsing these documents manually can be a slow and error-prone task, delaying underwriting decisions and quotes. By automating this process, we aim to:

• Handle the majority of cases with quickly and accurately with AI & LLMs
• Identify cases where automated parsing has failed to parse the document accurately
• Review in realtime the documents where data doesn’t reconcile
• Help underwriters focus on assessing risk rather than fixing data inconsistencies

Reconciliation - knowing when automation has failed

Reconciliation ensures that financial figures in loss runs align both within individual claims and across all claims. For instance:

• Within a single claim: Paid + Reserved + Expensed - Recovered = Incurred.
• Across all claims: Summed totals should match provided aggregates.

Discrepancies often arise due to incomplete data, formatting differences, or manual errors during document preparation. These discrepancies should halt automated systems, where our domain experts will intervene.

How it works

1. Automated Parsing: Our system uses the latest in AI / LLMs / NLP, trained on diverse loss run formats, to extract and validate key data.‍
2. Reconciliation Checks: Totals are verified for internal consistency. Discrepancies trigger an escalation process.‍
3. Human Review: Experts resolve inconsistencies, ensuring all data is complete and accurate before it’s passed on for quoting.

90% of loss runs are processed automatically and completed within minutes, whereas 10% of loss runs requiring human review are typically resolved within the hour

‍