01 the business
Independent physician group
60 physicians, billing team of 8
Denials arrive as codes on remittance files. The billing team works the largest ones and writes off the rest. The same denial reasons, such as missing authorisation numbers, repeat month after month.
02 before
The problem
Denials arrive as codes on remittance files that someone has to decode. Many are fixable, such as a missing modifier or authorisation number, but the team only has time for the largest ones. The rest are written off, and the same mistakes repeat.
- Small denials are written off without review.
- Appeal letters are written from scratch.
- Nobody tracks why denials happen.
03 the build
What the AI does
- Step 1: Reads remittance and denial files and groups denials by reason, payer and amount.
- Step 2: Separates simple fixes, such as missing data, from denials that need an appeal.
- Step 3: Prepares corrected claims, or drafts appeal letters citing the payer's policy and the evidence in the record.
- Step 4: Queues the work by value and deadline so the biggest and most urgent go first.
- Step 5: Reports the top denial causes each month so the front end can fix them at source.
paused: waiting for a person to approve
How it was put together
- Grouped denials by reason, payer and amount from remittance files.
- Separated simple fixes from denials needing an appeal.
- Drafted corrected claims and appeal letters citing payer policy and the record.
- Reported the top denial causes each month for front-desk and coding fixes.
04 human in the loop
Where a person stays in charge
Billing staff approve every corrected claim and appeal before it is sent. Clinical arguments in appeals are checked by a clinician or clinical reviewer.
- Billing staff approve every corrected claim and appeal.
- Clinical arguments are checked by a clinician.
- Appeals only cite evidence that exists in the record.
05 rollout
From kickoff to live in 5 weeks
- Weeks 1–2Load six months of denials and agree the work queue rules.
- Weeks 3–4Work the backlog with drafts, billing staff approving each one.
- Week 5 onwardsDaily queue plus a monthly root-cause report.
06 after
What changes
More denials are worked before deadlines pass, and recoverable revenue is not written off by default. The same denial reasons stop repeating because their causes are fixed.
Assume $1.2 million of denied charges a year, half currently written off, and a quarter of those recovered with timely work.
$1,200,000 × 50% = $600,000 written off. Recovering 25% of that brings back $150,000 a year.
What to measure
- Denials worked before their deadline.
- Dollars recovered from appeals.
- Denial rate by reason, month on month.
- Staff time per appeal.
07 evidence
What the research says
HealthCare.gov insurers denied 20% of in-network claims in 2023, and fewer than 1% of denied claims were appealed.
KFF, Claims denials and appeals in ACA Marketplace plans in 2023 (opens in a new tab) · 2025
Moving US healthcare admin transactions from manual to automated could save $20 billion a year; a prior authorisation takes providers about 24 minutes by phone, fax or email and 16 minutes through a portal.
08 lessons
Mistakes to avoid
- Chasing only big claims. Volume of small fixable ones adds up.
- Appealing without evidence. It wastes everyone's time.
- Never fixing causes. Prevention beats appeals.
09 questions
Common questions
- How common are denials?
- KFF found HealthCare.gov insurers denied 20% of in-network claims in 2023, and fewer than 1% of denials were appealed.
- Which billing systems work?
- Epic, athenahealth and clearinghouses such as Waystar and Availity.