The difference is in the exclusions — and they are thirty pages back
Two quotes with the same sum insured and a similar premium can be entirely different contracts. What separates them sits in the exclusions, the deductibles and the conditions precedent, distributed through the wording where no comparison table reaches. We make those terms comparable and put the clause reference behind every line. The recommendation stays with you, because advising is your regulated act.
Why wordings resist comparison
Four properties, and they share one cause: the terms that decide a claim are not written where comparison happens.
The decisive terms are not where the comparison happens
Sum insured and premium sit at the front and tabulate easily. Exclusions, conditions precedent and notification deadlines sit in the back as prose — and those are what decide a claim. A comparison built from the front pages compares the part that rarely matters.
Wordings are written to be read in sequence, not compared in parallel
The same restriction appears as a named exclusion in one wording, as a condition precedent in another, and as a definition that narrows a covered term in a third. All three have the same effect and none of them look alike.
Endorsements move the answer without moving the base wording
A schedule, an endorsement and a side letter can each override the policy form. Reading the form alone produces a confident answer to the wrong document — the failure mode that surfaces at claim time rather than at placement.
The work is repeated per renewal and rarely reused
The analysis done for one client in March is redone in September for a comparable risk, because nothing was captured in a form that survives the placement. The cost is not one comparison; it is the same comparison many times.
What gets extracted
- Exclusions, with the clause they come from
- Deductibles and self-insured retentions per coverage section
- Conditions precedent and notification deadlines
- Definitions that narrow an apparently covered term
- Sub-limits that sit below the headline sum insured
- Territorial and temporal scope, including claims-made triggers
What stays your work
- The recommendation — advising is your regulated act, not ours
- Suitability for the client's actual exposure, which no document states
- Market judgement about which carrier pays without a fight
- The conversation in which a client is told what is not covered
Where the value actually sits
Not primarily in the time saved on one comparison, though that is real. It sits in the two places where the current process fails quietly.
The first is the endorsement nobody read. A comparison built from the base wording is confident and wrong when a schedule modifies it. A system that ingests the whole file — form, schedule, endorsements — and reports which document a term came from removes a category of error rather than reducing its frequency.
The second is what survives the placement.The analysis done for one client is currently a document in an email thread. When each extracted term carries its clause reference, the same comparison is reusable at renewal and for a comparable risk — and the compounding effect of that is larger than the hours saved the first time.
There is a third effect worth naming because it is uncomfortable: a structured comparison makes visible how often the currently placed cover differs from what the client believes they bought. That is a conversation, not a feature, and it is the reason some brokers want this and others do not.
Frequently asked
Does this recommend a product?+
No, and that boundary is deliberate rather than cautious. Advising on insurance is a regulated activity, and the recommendation carries your liability. What the system produces is the comparison with a clause reference behind every line, so the recommendation you give rests on something you can show. Where a system suggests a product, the person who accepted the suggestion owns the consequence — which is a bad arrangement for a broker.
How is this different from a comparison portal?+
A portal compares the fields a carrier supplied in a structured feed: sum insured, premium, a handful of flags. That covers the front of the wording. This works on the wording itself, including the endorsements, and surfaces the exclusions and conditions the feed does not contain. The two are not substitutes — one narrows the field, the other tells you what you narrowed it to.
What happens when two wordings genuinely disagree?+
It is reported as a disagreement rather than resolved. If one wording excludes a peril outright and another admits it subject to a condition precedent, the output says so and cites both. Smoothing that into a single sentence is what produces confident comparisons that do not survive a claim.
Do you need client data to start?+
Not for the first use case. Comparing two wordings requires the wordings, not the client file — which means the first evaluation runs without personal data at all and the data protection conversation can happen after you have seen whether it works, not before.
What accuracy can you promise?+
None in advance, and a number offered before seeing your wordings comes from someone else's documents. What we can say is how the system behaves when it is unsure: the passage goes to a review list rather than into the comparison. On commercial wordings a share of clauses will always need a human, and a system that pretends otherwise is producing the errors that show up later.
Where is the processing done?+
On our own hardware inside the EEA, with a processor agreement and named processing locations. Prompts and completions are not written to durable storage. For brokers outside the EEA that is a neutral fact rather than an advantage — we mention it because it is the first question in our home market and we would rather state it than have it asked.
How long before it is usable?+
The comparison of two wordings is a days-to-weeks exercise, because there is no system integration involved. Putting it into the placement workflow — pulling wordings automatically, writing results back to the broking system — is the part that takes a project, and it is worth deciding whether you need it after the first version has been used for a quarter.
Two wordings are enough to judge
Send two commercial wordings you have compared by hand and the answer you reached. If the output matches, you have a baseline; if it differs, the interesting part is where — and that conversation is more useful than any demonstration. No client data is involved.
Send two wordings