In recent weeks, the Strait of Hormuz has forced the marine insurance market into real-time decision making.

War-risk premiums have moved by multiples in days. Notices of cancellation have triggered rapid repricing across portfolios. Underwriters have been asked, almost overnight, to reassess exposure in one of the world’s most critical shipping corridors.

Yet many are still relying on data that reflects a market that no longer exists.

That disconnect became clear as Gulf exposure surfaced in delegated authority portfolios. In one case, a vessel transiting a newly restricted zone appeared in a coverholder’s bordereaux three weeks after it sailed, priced at pre-escalation rates and submitted in a legacy spreadsheet format. By the time it reached the underwriter, the question was no longer pricing. It was whether the coverage would stand.

This is often framed as an AI challenge. It is more fundamentally a data problem.

Over the past several years, the insurance industry has invested heavily in artificial intelligence, particularly in tools that extract structured data from unstructured documents. Those capabilities are real and increasingly effective. But the events in the Gulf highlight a harder constraint. When market conditions shift in days, the limiting factor is not extraction. It is certainty.

In delegated authority, that constraint sits squarely in the bordereaux.

These files underpin how the London market prices risk, manages exposure, and governs portfolios. Yet they still arrive as PDFs, inconsistent spreadsheets, and bespoke templates, varying by coverholder and broker. In stable markets, that creates inefficiency. In volatile markets, it introduces material risk.

When pricing changes tenfold within a reporting cycle, small inconsistencies become significant. A misclassified vessel type, a voyage mapped to the wrong geography, or a premium reflecting outdated conditions are not edge cases. They can lead directly to pricing errors, aggregation blind spots, and disputed claims.

AI can accelerate the processing of this information, but it does not, on its own, resolve the underlying issue. Extracting data is not the same as verifying that it complies with binding authority terms, particularly when those terms are being tested in real time.

That distinction matters. Underwriters are being asked to make decisions on capacity and pricing while notices of cancellation ripple through portfolios and exposures evolve daily. In that environment, delayed or unverified data is not just an operational challenge. It is a threat to underwriting discipline.

From Reporting to Governance

The Strait of Hormuz should be understood as more than a geopolitical flashpoint. It is a stress test of an infrastructure that was not designed for volatility.

If data quality remains a downstream, manual process, the market is effectively underwriting in the dark. A 30-day reporting cycle cannot support decisions in a market where risk and pricing can shift over a weekend. In that context, reporting becomes retrospective. Governance must become real time.

Bridging the gap between data extraction and underwriting certainty requires a shift in approach.

First, validation must move to the point of ingestion. Submissions and bordereaux should not be accepted unless they conform to the constraints of the binding authority.

Second, the market should prioritize standardization at the source over downstream data cleaning. Using AI to repair inconsistent inputs is inherently reactive.

Third, underwriting discipline depends on visibility. Real-time insight into exposure is no longer optional in fast-moving risk environments.

Until the data foundation is addressed, AI will continue to operate on top of uncertainty, accelerating processes without fully resolving the risks embedded within them.

The signal from the Gulf is clear. The market is moving faster than the bordereaux. The data needs to catch up.