// What we build
A submission is a set of documents that disagree.
ACORD 125, 126 and 140, the SOV, the loss runs, and the material fact that exists only in the third paragraph of a broker's email. Read each document independently and you get a clean record that is confidently wrong. The work is reconciliation.
The strongest genuine return case in commercial insurance, precisely because the work being automated is genuinely clerical. Nothing here judges a risk.
// The situation
Three use cases have a production record. This is one.
Document analysis and summarization, case analysis and submission ingestion are the three use cases with a real production record behind them. Everything else in the category is thinner than the marketing around it, and we will say which is which before you fund anything.
So the first week is spent counting rather than building: touches per submission, days from receipt to first underwriter decision, rework rate, and how many inbound submissions are the same risk twice. Those four numbers decide the scope.
// The stack you actually run
The artifacts, and the mailbox they arrive in.
Two things that separate people who have run an intake queue from people who have demoed one.
First, the documents contradict each other routinely. The SOV total and the ACORD 140 total insured value disagree. The driver count on the 127 does not match the driver schedule. When the disagreement is material, the output names the conflict, shows both values, and cites the document and page each came from.
Second, the same risk arrives more than once. One retail agent, three wholesalers, six markets each, and the insured name is spelled a little differently in each submission. Every duplicate counts as an additional quote in the denominator, so hit ratio is understated by an amount nobody has quantified.
// The build
Four builds, and what each one is measured on.
Cross-document reconciliation with conflict surfacing
One submission record assembled from every document in the set, with each value carrying the document and page it came from. Where two sources disagree by a material amount, both values are retained and the conflict is raised as a field rather than resolved quietly.
Surface: the submission mailbox and ACORD AL3 into the submission record in your underwriting workstation.
Duplicate and clearance detection
Fuzzy matching on insured name, FEIN, address, effective date and exposure fingerprint, so the same risk arriving from three wholesalers resolves to one clearance record with three submission channels attached.
Surface: the clearance record, with the duplicate group written back as a linked set rather than a deletion.
Appetite and authority screening at intake
Class, limit, geography, attachment and premium-size boundaries checked before the file reaches a desk, with a citation to the governing guideline or binding-authority clause. Out-of-appetite risks are stopped with a reason.
Surface: the guideline and contract corpus turned into an executable check, versioned alongside your underwriting manual.
The email body treated as a document
The mid-term acquisition, the building added last month, the reason the admitted market passed, the sublimit the broker wants matched — extracted, tagged and cited like every other source.
Surface: the intake mailbox into the submission record as a tagged extract with the message retained as the citation target.
What we measure
Touch count per submission, counted by observation rather than estimate. Time from receipt to first underwriter decision, on the same cohort definition on both sides of the build. Rework rate, meaning how often a file goes back to the broker for something that was already in the set. Duplicate rate, reported as a correction to hit ratio rather than as a time saving. Share of fields extracted, inferred and defaulted, reported per submission instead of in aggregate. Definitions and baselines, set in the first week.
“Intake is a production system with a clock on it, which is why the read starts from your process documentation and a sample of the real files rather than from a process diagram.”
// Where this comes from// What's hard about this
Two limits, and what we do about each.
Screening at intake requires the guideline to be expressible.
A large part of real appetite is not written down anywhere: the class you write only for an incumbent broker, the geography you are quietly walking away from this quarter. A screening layer built on the written guideline will confidently stop risks you want.
So the check routes rather than decides wherever the guideline runs out. Everything expressible becomes an executable rule citing the clause it came from. Everything else becomes a referral with the reason attached and a named authority on it, and the rules stay editable by your own team rather than living inside a model.
Triage is where adverse selection enters.
When an underwriter can only work a fraction of inbound, the files that get worked are the ones easiest to read, not the ones most likely to be profitable. That selection effect is invisible in every metric an underwriting operation keeps, and it worsens as volume rises.
The refutation is to separate the two things triage conflates. Readability is scored as completeness and remediated automatically. Appetite fit is scored independently, on the risk. The queue is ordered on the second, and the gap between the two orderings is reported as a number.
// What ships with it
The governance file, scoped to intake decisions.
A model inventory entry for the extraction models and, separately, for anything that scores appetite, since those carry different risk tiers. Data lineage from source document and page to submission field. Pre-deployment testing results, drift thresholds with remediation triggers, and a named human decision-maker specification for every path that can stop a submission. Roughly half the states have adopted the NAIC AI Model Bulletin, and a declination at intake is an adverse action in several of them.
Bring us one week of inbound.
We will tell you what your intake queue costs per submission, and where the touches go. How the read works →