Submission intake that clears itself,
and can show its work.
Submissions arrive as email, PDFs, ACORD forms, and loss runs. Someone reads them, checks them against what you already write, decides whether they fit appetite, and chases the broker for what’s missing. That person is usually an underwriter. Haltere builds the intake line that does the reading and the checking, leaves a decision receipt for every clearance, and puts a named human on every release. You own it when it ships.
The queue is where the program’s margin goes.
Clearance against in-force and prior submissions. Blocked classes. Appetite by state and line. The broker who sent the wrong loss runs. None of it is underwriting, and all of it lands on underwriters, because the alternative is a SaaS product you rent, can’t audit, and can’t change. The tier below the enterprise vendors’ AI factories has been told to wait. We built for it.
Extract. Clear. Route. Respond.
Your rules, encoded and versioned. Your underwriters, deciding instead of triaging. Every step on the record.
From the inbox to a structured record.
Email, PDF, ACORD, supplemental apps, loss runs. Every field carries its confidence and the page it came from. Low-confidence fields go to a person before anything else happens.
The checks you already run, at the gate.
Duplicate and prior-submission clearance, blocked classes, appetite bands, missing documents. The rules are yours, versioned, and named on every decision. A submission outside appetite is flagged at the gate; no one has to notice it later.
To the right desk, with the letter drafted.
Cleared submissions land on the right underwriter’s desk. Acknowledgements, declinations, and requests for what’s missing are drafted for a named person to release. Nothing leaves without a human on the record.
Underwriters underwrite. The line does the reading.
One submission, cleared,
with the record the carrier can audit.
Every clearance decision is a receipt: what was received, what was checked and at which rule version, what was flagged, who approved. Fixture data, real shape.
Fixture data. The 2024 declination surfaced at the gate, not in a renewal review. The missing loss runs were requested by a person, not a bot, and the record says so. The receipt format is public
The record your carrier’s audit
already entitles them to.
Drawn from the line’s receipts. Not assembled by the underwriter the night before.
The SaaS quote is per submission.
The audit trail is theirs.
Someone else’s black box.
Accuracy claims you can’t verify. Your submissions on their servers. Rules you can’t read, priced per seat or per submission, and a record that belongs to the vendor when the carrier asks for it.
Yours, in your environment, receipted.
Built to your appetite and your carriers’ formats, in your boundary, with a decision receipt per clearance. Fixed fee, then governed operations. You walk at renewal with everything.
The governed first build.
One backlog item. Fully governed. Evidence included.
Intake from the inbox to the underwriter’s desk: extraction, clearance, routing, and the outbound letters, for one program or line to start. Scoped together before anything starts.
Scoped to the project and quoted in the first conversation. Agreed before anything starts. No time-and-materials meter running.
From scoping to production-acceptable delivery, live in your environment.
Two deliverables, always paired: the working software, and the evidence pack. Show the board AI progress; show the examiners the receipts.
The process guarantee: your team reviews and approves every change. Nothing merges ungoverned. Nothing leaves for a broker or a carrier without a named human release.
Two ways to run it afterward: we operate the line for teams with no engineering bench, or your team operates it and we train your operator. The software, the map, and the receipts are yours in either mode. You walk at renewal with everything. How an engagement runs, stage by stage
What you’ll want to know.
Does this replace our AMS or policy admin system?
No. Your AMS and policy admin system stay where they are: the line runs upstream of them and writes the cleared record into the one you already run. What gets built is the intake that was never a system, the shared inbox and the clearance spreadsheet. The rule that lives in one underwriter’s head gets written down first, with her, and that is the first part of the work rather than a step we skip. Nothing of yours is switched off by us.
What about scanned or handwritten submissions?
Every extracted field carries a confidence score and a pointer to the page it came from. Fields below the threshold you set are routed to a person before anything is cleared. The record shows which fields a human confirmed.
Will it decline or acknowledge submissions on its own?
Not by default. Declinations, acknowledgements, and requests for missing loss runs are drafted for a named person to release. The gate is yours to loosen later, and every release is on the record either way.
Can our carrier see the record?
When you choose to show it. Every clearance decision is a row you can export: what was received, which rules ran at which version, what was flagged, who approved. That is the delegated-authority audit, answered from a query instead of a reconstruction.
How long, and what does it cost?
Four to eight weeks from scoping to a working line in your environment, for a fixed fee agreed before anything starts. We quote it in the first conversation, not on a page.
Bring us the intake queue.
Tell us the program, the carriers, and the person who reads the inbox today. Scoping is a conversation, the fee is fixed before anything starts, and the first review is on your calendar before we write a line.
Inquiries are answered within one business day, by someone who can answer them.
