For a long time, Kyber has made a simple claim:

Kyber's AI templates write like your best adjuster’s first draft every time.

The purpose of our AI templates, whether a Partial, Denial, Payment, or Reservation of Rights, is not to remove the adjuster from the writing process. Rather, the goal is to give an adjuster a stellar first draft. Things like fraud disclaims, organization approved language, tone, and formatted are taken care of in a matter of seconds.

Not aiming for 100% ready to hit send. But structured, accurate, and defensible enough that the hard work is already done.

The adjuster then has the opportunity to edit and tweak the letter to their liking, before either assigning a manager to review or sending.

Customers have consistently validated this anecdotally. Adjusters tell us they rarely need to rewrite drafts. Reviewers tell us they spend less time correcting letters. Claims leaders tell us cycle times drop dramatically once Kyber is in place.

And almost inevitably, the follow-up question comes:

“How much does an adjuster actually edit?”

Until now, the answer was anedcotal. Now, it is empirical.

Turning Anecdote Into Proof

Kyber represents approximately $7B in gross written premium across personal and commercial P&C. That footprint gives us a unique opportunity to answer the editing question honestly, using real behavior instead of opinions.

We analyzed revision history across all AI-generated letters, spanning carriers, lines of business, and letter types.

The goal was simple: Measure what happens after the first draft.

What the Data Shows

By looking at how many revisions occur between draft generation and review assignments, a clear pattern emerges.

• 55% of letters are sent with zero edits
• Over 65% are sent with two revisions or fewer
• The median letter requires no edits before sending

There is a long tail. Some letters require more work, as expected. When claim notes are poor quality, coverage is highly bespoke, or multiple endorsements interact in unusual ways, adjusters step in to refine the draft. The key signal is the center of the distribution. Most letters cluster tightly around zero to one revision, which is consistent with strong first drafts rather than rough drafts.

How to Interpret This Distribution

In claims operations, revision count is not just a productivity metric. It is a quality signal.

Senior adjusters do not rewrite their own drafts. They skim, adjust tone, trim language, and send. Junior adjusters and weak templates, by contrast, reveal themselves through heavy rewriting and back-and-forth.

Kyber’s revision distribution mirrors the former, not the latter.

When adjusters edit Kyber drafts, they are most often:

• Removing excess policy language included for completeness
• Making minor narrative adjustments for claim-specific nuance
• Confirming figures, recipients, or formatting

They are not rebuilding structure, adding missing policy sections, or correcting hallucinations.

Why This Explains the Cycle-Time Gains

Kyber customers report 5× faster letter cycle times in other case studies. This revision data explains why.

Time is not saved because letters generate quickly. That part is table stakes.

Time is saved because adjusters start from a draft that is already good.

Instead of assembling a letter, adjusters validate one. Instead of writing, they review. Instead of rewriting, they refine.

That shift compounds across thousands of letters.

The Big Takeaway

Kyber has always claimed to write like your best adjuster’s first draft every time. Now, we have the numbers to back the statement.

Across billions in written premium, revision patterns show that most letters require little to no editing, while adjusters spend more time only on genuinely complex cases. Review remains part of the process. The difference is that adjusters start from a solid draft rather than a blank page.

This is what enables meaningful gains in correspondence speed and consistency without compromising accuracy or control.