Flood-claim audit product

King Tide started as a tool for one flood-adjusting team. Joseph Koenig, a senior adjuster and national flood-insurance trainer, was losing valuable time on every storm to a mechanical chore: hand-checking every claim his trainees sent up for the same small, formatting-level errors before any of it reached the company-side examiner. He commissioned Daedalus to build the audit he was running by eye. It worked well enough that he turned it into a product the whole flood industry can use.

The errors were never the interesting part of the job. Depreciation brackets, overhead and profit, labor minimums, baseboard deductions on rooms with cabinets and vanities, unfilled dropdowns, missing photo captions, prior loss data entry. None of it requires an adjuster’s judgment, and all of it has to be right before a claim leaves the desk. During a hurricane, a trainer spends hours doing this by hand across a stack of trainee files while carrying their own.

Daedalus built King Tide to absorb that work: a self-hosted audit engine that reads every page of a complete claim package, runs mechanical detectors to identify the most common mistakes adjusters make, and returns the file annotated in place with a summary cover page. Then we built the product around it: a multi-tenant dashboard with separate logins for trainers and trainees, billing, and an owned production deployment. With that, the tool Koenig built for his own team became something any trainer, team, or independent adjuster can run.

At a glance

Manual audit time per claim

30 min

King Tide audit time per claim

< 1 min

Pages audited per claim

every page

Claims per storm (team of 5)

~175

Review loop removed per claim

15 min

Time saved per storm

~87.5 hrs

Estimates model one trainer and four trainees through a single hurricane deployment, the catastrophe surge when claim volume peaks and flood work actually happens. They assume 30 minutes to hand-audit a claim for these mechanical errors versus under a minute in King Tide, about 10 minutes saved on a trainer’s own claims (an experienced adjuster catches most of their own), and about 15 minutes of review back-and-forth removed per trainee claim. Because flood work is event-driven, savings are expressed per storm rather than per month.

The problem

The team ran a high-friction review chain where every trainee’s claim had to clear the trainer before it reached the company examiner.

• Every claim a trainee submitted had to be proofread by the trainer before approval was given to send out a proof of loss to the insured.
• The review was mechanical: depreciation, overhead and profit, labor minimums, baseboard deductions, dropdowns, captions. None of it required judgment, and all of it had to be correct.
• During a hurricane the trainer can carry 40–100 of their own claims while reviewing 10–40 from each of several trainees.
• Files moved back and forth over email, Slack, GQueues, or Telegram, with a flag-and-fix loop on every one.
• The mechanical proofreading crowded out the actual job: teaching trainees the right line items, the flow of a claim, how to work with the insured, and how to write a narrative that tells the story.

“Those are mechanical errors, and King Tide catches them before a claim ever reaches me. I’m not a human spell-checker anymore, which means I can actually teach: how to take care of the insured, how to tell a real story in the narrative, how to estimate like a professional. Instead of correcting the same small mistakes every storm, I’m developing a team that can lead the future of flood adjusting in America. That’s how a trainee becomes a great adjuster.”

Joseph Koenig, Senior Adjuster & National Flood Insurance Trainer

That became the shape of the work: take that work off the trainer’s desk, give it to every trainee directly, and turn the review into a learning loop instead of a bottleneck.

What we built

Daedalus took the audit Koenig was running in his head and built it into software: first the engine, then the multi-tenant product around it.

The audit engine

A Python and PyMuPDF engine reads every page of a complete claim package and runs its mechanical detectors over it, the same pass an experienced adjuster runs by eye. It returns the file annotated in place, with a brand-styled cover page that lists every finding by page and category, so the reviewer reads one page and knows exactly where to look. A clean claim comes back marked clean.

• Narrative information entered incorrectly.
• Template data not replaced with claim-specific information.
• Mortgage mismatches.
• Overhead & Profit applied to prohibited items.
• Missing or incomplete photo captions.
• Prior-loss inconsistencies, or missing entirely.
• RC on prohibited items or on tenant properties.
• Reserve inconsistencies.
• Discrepancies across the claim-summary sections.
• Missing depreciation on any material item.
• Baseboard deductions missed where cabinets or vanities displace the wall.
• Labor minimums applied on NFIP claims.

From script to multi-tenant SaaS

The original tool was a single-purpose script that ran on one machine. Daedalus rebuilt it as a multi-tenant application, a Next.js dashboard over the FastAPI engine, so any trainer or adjuster can upload a claim and get it back audited without touching a command line.

The trainer / trainee model

The rebuild added the piece that changes the workflow: separate logins for the trainer and for each trainee. A trainee runs their own claim through King Tide before it ever reaches the trainer, and the trainer sees every seat’s audits from one dashboard. The trainer stops being the linter. The trainee sees their own mistakes, fixes them, and learns to stop making them.

Owned, not rented

King Tide runs on infrastructure DDLS controls: self-hosted database and file storage, deployed behind the standard DDLS edge. It is a product the business owns and can change in days, not a subscription to someone else’s platform.

How a claim moves through the system

1. 01 Upload

   A trainee drops the complete claim package in as a single PDF, before submitting for Proof Out Approval.

2. 02 Audit

   Mechanical detectors scan every page for the errors a trainer would otherwise catch by hand.

3. 03 Return

   King Tide returns an annotated PDF and a cover-page summary in seconds, every finding marked in place.

4. 04 Fix & learn

   The trainee sees each error, fixes it, and learns the pattern, instead of waiting on a round of trainer notes.

5. 05 Oversee

   The trainer reviews from one dashboard and spends the reclaimed time on substance, not formatting.

Where the storm load comes from

On a typical hurricane deployment the trainee workload, not the trainer’s own claims, drives almost all of the savings, because that is where the full thirty-minute hand-audit and the fifteen-minute review loop both fall.

Time saved

Two distinct savings stack up on every storm: the audit a trainer no longer performs, and the review loop the team no longer runs.

The model is deliberately conservative. It counts only clock time on the audit and the review loop. It does not count the errors that would have slipped through to the examiner, the files kicked back for rework, or the compounding effect of a trainee who learns to stop making a mistake. Every one of those is a claim the trainer never has to review again.

Why it worked

King Tide works because it draws a hard line between mechanical and substantive. It does not read the estimate, judge the line items, or write the narrative. Daedalus deliberately kept it out of any work that needs an adjuster’s judgment. It catches the formatting-level errors a human catches by eye, and it catches them every time.

That line is what makes it a teaching tool instead of a crutch. When King Tide handles the mechanical proofreading, a trainer’s time goes back to the work that actually builds an adjuster: making sure trainees deliver real service to the insured, learning to tell a proper story in the narrative, choosing the right line items, and becoming top-notch estimators. Instead of spending a storm correcting simple mistakes, the trainer spends it developing the team. The value scales with how green that team is: the less experienced the trainees, the more a trainer was doing by hand, and the more King Tide gives back.

The result

What started as an internal tool for one adjusting team is now a product any trainer, team, or independent adjuster in the flood industry can use. It runs the same audit Koenig used to run by hand, on every claim, in seconds, and gives a trainer back the better part of a workweek on every storm.

More importantly, it changes what a trainer does during a catastrophe surge. Instead of clearing a backlog of mechanical fixes, the trainer spends that time developing an excellent team: adjusters who serve the insured well, write narratives that tell the story, and estimate with real skill. That is the part of the job that was always theirs to do, and it is how the next generation of flood adjusters in America gets built.