We built a submission engine. It shipped.
A commercial insurance submission engine was scoped, built and tested in this studio. It is now a product of its own, called PacketSlip, with its own site, its own pricing and its own readiness statement. This page is what the work looked like from here.
- Status
- Shipped
- Now called
- PacketSlip
- Built for
- US commercial P&C agencies
- Lives at
- packetslip.com
Where it lives now
Everything about buying, using or evaluating the product is on its own site. Nothing on this page is the product.
PacketSlip
Commercial insurance submission software. It reads the packet — ACORD forms, loss runs, statements of value — checks it against itself, flags what conflicts or is missing with a citation back to the page it came from, and drafts the follow-up. Pricing, a demo and the product's own account of what it does are all there.
What the problem was
The reason it was worth building, in the terms the agencies described it.
- The packet arrives as a pile
- An email with attachments: ACORD forms, a loss run, a statement of value, a broker note. Nothing in it is structured, and the useful information is distributed across all of it.
- The same figure appears more than once
- A building value on the SOV and a different one on the ACORD 126. Someone has to notice, and noticing depends on reading both documents carefully on a day when there are eleven other packets.
- What is missing is harder to see than what is wrong
- An absent loss run or an unanswered supplemental question is invisible until an underwriter asks for it, by which point the submission has already been sent.
- The follow-up is written from scratch each time
- The list of what to chase exists only in the head of whoever read the packet, and has to be turned back into an email.
What the review screen looked like here
The interface below is from the demonstration built in this studio, on synthetic documents. The product's own screens are on its site.
| Finding | Status |
|---|---|
| Requested effective date is blank | Needs review |
| Annual payroll differs between the application and the supplement | Needs review |
| Prior carrier named without a policy number | Needs review |
| Insured entity name matches across all three documents | Checked |
| Classification code present and consistent | Checked |
Further reading
What makes document extraction dependable, and the cases where a person should still make the call, is written out in reading documents reliably, and when not to.
What this says about the studio
One finished thing is worth more than a page of principles, so it is worth being precise about what it does and does not prove.
- It proves the studio finishes
- A process was described, scoped, built, tested against documents it had not seen, and released under its own name. That is the whole claim.
- It is not a client result
- PacketSlip is the studio’s own product, not work delivered for a client. No outcome figure is attached to it here, and none will be until one has been measured and published with permission.
- Its readiness is its own to state
- What PacketSlip handles, how it treats data and what it is ready for is set out on its own site rather than summarised here, where it would go stale.
The same work, on your process
If a process in your business looks like that one did — a pile of documents, a person checking them against each other, a follow-up written from scratch — that is the conversation this studio is for.