01
Section-aware, not text-aware
The slip is mapped to a section model before comparison, so a change is reported against the clause it belongs to rather than against a line number.
PolicyCompareIQLive
Read every clause.
In two minutes, not four hours
Baseline-versus-renewal slip review as a structured, auditable analysis — financial deltas, coverage drift, missing sections and risk implications, each cited back to the source PDF.
2 minmedian review time
52sections analysed per comparison
11classes of business supported
The reality
What this looks like
in most firms today
- Renewal season means three underwriters, two weeks and a spreadsheet of differences.
- A dropped Side-A extension or a shifted retention is found late, or it is found by a claim.
- Two people reading the same pair of slips produce two different lists of changes.
The work is not difficult. It is long, repetitive and unforgiving of a lapse in attention on page 41 — which is precisely the profile of work that should not be done by a senior underwriter at eleven at night.
Where the existing tools stop
Where tools stop
Generic document comparison tools diff text. They do not know that a change to the definition of Insured Person carries more weight than a change to the broker’s address.
Large language models will summarise a slip willingly, and will just as willingly summarise a section that was not there.
What it takes
What renewal review needs is section-aware comparison with severity attached to every finding and a citation behind it.
How it delivers value
How PolicyCompareIQ earns its place
in the workflow
02
Severity rated, not just listed
Every finding is classified high, medium or low risk, so a 62-item change list becomes a 20-item review list and a 42-item record.
03
Cited back to the source PDF
Each finding carries the page and section it came from in both documents. Nothing has to be taken on trust, which is the only basis on which it gets used.
04
Ordering detected, never assumed
Baseline and renewal are identified from the period-of-insurance dates rather than from the filename, so the comparison never runs the wrong way round.
Capabilities
The specification, without decoration
- Clause-level comparison across 52 sections in a median two minutes
- Eleven classes of business, from D&O and cyber to offshore energy and reinsurance
- Financial delta extraction — premium, limits, retentions and sub-limits
- Missing-section detection, reported explicitly rather than silently omitted
- Versioned prompt templates per class of business, promoted through a production gate
The approach
How it actually works
Written for the technical evaluator rather than the buyer. If you are going to put this in front of an architecture review board, this is the section they will read.
Extract, then map to a section model
Text extraction runs first, then each document is mapped to the section model for its class of business. Comparison happens between mapped sections, which is what makes a finding meaningful rather than positional.
Parallel MAP across sections, then a REDUCE pass
All sections are analysed in parallel, then a reduce step classifies and de-duplicates findings across the whole document pair. Flags raised in the map phase are confirmed or withdrawn in the reduce phase — not before.
Class-specific prompt versioning
Each class of business carries its own versioned prompt template. Versions are promoted explicitly to production, so a change to marine wording analysis cannot silently alter the D&O results.
Streaming job telemetry
Long-running comparisons stream their state — sections completed, elapsed time, signals raised — so the reviewer can start reading findings before the job ends.
Encrypted at rest, on your tenancy
Documents are held encrypted and the deployment runs inside your own Azure tenancy. Slips do not leave your boundary to be compared.
The breakthrough
What changes once it is running
Median and count figures are measured in production. Time-saving figures are expressed as ranges because they depend on your document set.
2 minmedian comparison timeAcross 52 sections, measured in production
52sections per comparisonMapped to a class-specific section model
60–80%less review time per renewalModelled against a two-underwriter manual review
11classes of businessD&O, cyber, marine, aviation, property and more
