Reinsurance submissions are increasingly facing tech scrutiny before they reach a person. Triage tools, exposure benchmarking systems and automated scoring models sit ahead of the underwriter’s desk, sorting incoming business into “proceed,” “query” or “decline” before anyone has read the story behind the risk.
For clean, standard business, this has the benefit of speed and efficiency. But for complex risks – the ones with a difficult loss history, an unusual peril combination, or exposure that doesn’t map neatly onto a standard class code – it really matters. A risk that would interest an experienced underwriter can be filtered out before it ever reaches them, not because it’s a bad risk, but because it was presented in a way the system couldn’t read.
At renewal, when volume is high and timelines are short, this filtering effect is at its sharpest. Getting a complex risk in front of a human underwriter increasingly depends on how the submission is built, not just on the merits of the risk itself.
What underwriters, and the systems trained to think like them, look for
Automated triage tools are built to replicate the first pass an underwriter would make:
- Is the data complete?
- Is it internally consistent?
- Does it fall within recognised parameters?
What looks like a “computer says no” moment is usually a proxy for the same questions a person would ask, just asked faster and with less patience for ambiguity or nuance. A submission that would survive a first read from an experienced underwriter will generally survive the algorithm too. The reverse is also true.
- Lead with the data, not the narrative
Underwriters value a good risk story, but the story only gets read once the data has cleared the first hurdle. Submissions that open with context and background before getting to exposure figures, loss history and terms often lose the algorithm’s attention before the narrative has a chance to begin. Put the numbers first, complete and in a recognised format. Save the framing for the space after the facts, not instead of them.
- Don’t bury the mitigation
A poor loss history combined with strong, recent mitigation can still be an attractive risk, but only if the mitigation is presented as data, not prose. “The insured has since installed a modern sprinkler system” reads as a sentence a model can’t score. A specification, an install date and a certification reference read as facts a model can weigh against the loss that it’s flagging. If mitigation is going to change the outcome, it needs to be structured well enough to change a score, not just soften a paragraph.
- Anticipate the outlier flags before they’re raised
Every complex risk has at least one figure that will sit outside the normal range, such as a limit that looks high for the class, a rate that looks low for the peril, an occupancy that doesn’t fit a standard code. Left unexplained, that outlier is what triggers a decline. Addressed directly and early, in plain terms, the same figure becomes a point of interest rather than a red flag. If a number is going to raise a question, anticipate the question and answer it.
- Make the complexity legible, not hidden
There’s often an instinct to smooth over the parts of a risk that don’t fit a standard template, in the hope that a cleaner looking submission will move through more easily. It tends to have the opposite effect. Systems are generally better at handling risk that’s clearly labelled as complex than risk that’s been simplified quietly and later doesn’t reconcile. A well-structured submission that says plainly “this is a multi-peril exposure across three jurisdictions” will fare better than one that’s been trimmed to pose as a single-line domestic risk.
- Give the underwriter a reason to pick up the phone
The goal of a submission isn’t only to pass a filter, it’s to generate a conversation. Even the best structured data package benefits from a short, specific note flagging what makes the risk worth a second look, and inviting a call before any decision is finalised. Automated tools are increasingly good at sorting, but they’re not a substitute for a broker and an underwriter talking through the part of a risk that doesn’t fit neatly into a field.
The point of getting this right
None of this is about gaming a system. It’s about recognising that the first read of most submissions is no longer human and adjusting how complex risk is presented accordingly. The underlying discipline is the same one that’s always mattered in reinsurance: complete data, clear reasoning, and mitigation that’s demonstrated rather than described.
At NPRe, we still believe complex risk deserves a human conversation, and we’d rather have that conversation early than lose a good risk to a filter it was never built to pass. Brokers who structure their submissions with an automated read from the start tend to find their business gets to the right desk and stays there long enough to be evaluated thoroughly.