All customers

Case study · Insurance

TIE decides transportation insurance endorsements in under a minute.

Native helped TIE, a US transportation insurance MGA, put agents on every submission and endorsement pack. They check each one and either decide it or hand the underwriter a pre-filled decision, and 74% of decisions now need no underwriter.

TIE case study

Between 30 March and August 2026, the system made 4,224 endorsement decisions, 3,130 of them fully automated at 23 to 40 seconds each. 74% of endorsement decisions are now made with no underwriter involved, and queue-to-decision time went from 4 to 8 hours to under a minute.

74%of endorsement decisions with no underwriter
<1 minendorsement queue to decision, from 4 to 8 hours
4,224endorsement decisions in five months

About TIE

The client is a US managing general agent writing transportation insurance, with about $60M of gross written premium and a team of 20 to 25, around 10 of them underwriters. Turnaround drives its revenue, because retail agents send the same file to several markets and the fastest usable quote tends to win.

The challenge

Submissions took 2 to 4 days, and endorsements sat 4 to 8 hours in a queue behind new business. An endorsement wins no new revenue, so it waited.

The work itself was quick. Reading and checking an endorsement took about 5 minutes, across roughly 886 requests a month, and submission entry took about 6 minutes before 20 minutes of underwriting.

The business had already tried automating it in ChatGPT, hit the limits and seen the potential.

What we built

Endorsement reviewerTen endorsement agents approve, decline or refer mid-term policy changes like adding a driver or a truck. Ten change types need no person at all.
Submission reviewerEight parser agents read the whole new-business pack and hand the underwriter a pre-filled decision. The underwriter still makes the call.
Document parsing with fallbacksA classifier works out what each document is from its contents, even when one file mixes several types. Driver licences that arrive as poor phone photos get a second model behind the first.
Carrier safety checkThe agent queries the federal SAFER database mid-process to catch carriers that re-register under a new DOT number to shed a bad safety record.
Missing-document detectionAn incomplete file is flagged the same day, before anyone starts chasing what never arrived.

How it works

  1. 01ArriveThe document pack lands in Surefyre, however messy it is.
  2. 02ExtractEach document is identified and read, with Mistral OCR first and a vision model as the fallback.
  3. 03RouteThe file is sorted by type and sent to one of 8 submission agents or 10 endorsement agents, which run their checks.
  4. 04Decide or referWhere the rules settle it, the agent decides. Anything else goes to an underwriter with the analysis and reasoning already written.

Results

The endorsements that still reach a person arrive pre-analyzed and are worked within the hour. Queue time saved came to about 18,800 hours on the automated decisions and 5,500 more on referrals. Of 2,988 driver-add requests, about 1,989 were approved automatically.

On new business, the system made 2,036 submission decisions. Entry time fell from about 6 minutes to 73 seconds, and underwriting time from 20 minutes to 14.

Put AI to work across your whole business.

Bring the hardest problem on your list. In 30 minutes we'll show you where AI will have the most impact first, and what it takes to get there.