Agents for the manual workbehind denials

Explore Merid

Every denied claim read, argued and prepared for filing by agents, each approved by your biller before it goes.

Revenue that was being written off is now recoverable without anyone working a denial by hand.

The claims specialist engineered to build the casefor your medical billing team.

Get started
Without Merid
Appeal letter for claim 2604188270 · denied CO‑5000min
Piedmont Digestive Health, PC

Re: Claim 2604188270 — appeal of the denial of 2 April.

is this level 1 or a reconsideration? — the biller rereads the denial letter

The patient had already failed other therapies before the infusion.

which therapies, and how long on each? — the biller opens the chart and reads

Per the plan’s policy, step therapy requires a corticosteroid and an immunomodulator…

which version governed 18 Mar? — the biller calls the plan and waits

The infusion was authorized before it was given.

where is the authorisation number? — the biller digs back through the claim

Under criterion 3 the member must have failed a corticosteroid and an immunomodulator at therapeutic dose.

The chart documents three taper attempts and a 6‑TGN level of 284.

We ask that the denial be overturned and the claim reprocessed.

Biller’s manual work Read the denial letterto find which level this is, and the date the clock starts from Pulled the chartthree documents, read end to end to find the passages that answer the criteria Found the policyfour PDFs on the payer’s site, none dated, then a call to confirm which governed Found the authorisationback through the original claim to the number in Box 23 Worked out the deadlinewhich window this denial runs on, and the date it starts from
With Merid
Appeal letter for claim 2604188270 · denied CO‑5000min
Piedmont Digestive Health, PC

Re: Claim 2604188270 · Member CBH884120773 · DOS 18 March 2026. This is a first‑level appeal of the denial dated 2 April.

Authorization 0318‑CBH‑77421, issued 2 March, records a review against the plan’s own step‑therapy criteria.Ex. A The number was reported in Box 23.Ex. B

The note of 14 January records prednisone from 3 June 2025 and three failed tapers — criterion 3(a), met in the alternative it provides for.Ex. C

The policy asks for an immunomodulator at therapeutic dose for twelve consecutive weeks.Ex. D

Azathioprine 200 mg daily ran sixteen weeks, and adequacy of dose is measured rather than asserted: 6‑TGN 284 pmol/8×108 RBC, inside the therapeutic window.Ex. E

Both limbs of criterion 3 are therefore met. We ask that claim 2604188270 be reprocessed and paid at the contracted $1,649.04.

EnclosuresA prior authorizationB CMS‑1500, Box 23C progress note, 14 JanD step‑therapy policy, in force 18 MarE laboratory report

Merid agent is drafting…

Merid agents accomplished

Remittance readCO‑50 with remark N130 — a policy dispute, so an appeal and not a corrected claim
Policy retrievedthe step‑therapy policy as it stood on 18 Mar — not today’s
Relief computed$1,649.04, the contracted allowed amount from the rate your practice supplied — the figure the letter asks for, not the $5,155 charge
Chart read and testedcriterion 3(a) and 3(b) both met — the note of 14 Jan and the metabolite level, each quoted verbatim
Deadlines computedappeal due 29 Sep — 180 days from the notice of 2 Apr, the appeal term in your participation agreement

5 exhibits · 11 pages · every sentence resolved to its source

Approve and file

Appeal every denied claim that deserves to be paid. The work behind each one is why practices leave that money with the insurer. Merid’s agents do the work, and your billers keep the judgment.

Time to work one denial by hand, estimated
45–60min
Cost to contest one, commercial
$17
Denied claims never resubmitted
40–66%
Prior‑auth denials overturned on appeal
43–67%

Agents know how to address every insurer reason.

The code is the category; the remark is the specific — and the two routinely call for different documents. Merid reads both, classifies against its own curated table first and the industry’s 1,981 published code combinations second, and produces the instrument the reason actually calls for.

CO‑16$340
N290rendering provider identifier
Corrected claim · 837Ptimely filing
A refile, not an argument.

Fix and refile. CARC 4 is a modifier. CARC 16 with N290 is a rendering NPI. Neither is a dispute, and an appeal here spends a level of appeal on a typo. Merid reads the remark, names the field, builds the corrected claim from your coder’s one‑line fix, and files it on the payer’s correction window — which is not the appeal clock, and is the one that governs these.

CO‑50$1,649
policythe version in force on 18 Mar
Criterion 3(a) · 3(b)met, quoted
First-level appeal, every sentence cited.

Argue it. CARC 50 says the payer’s own criteria were not met; CARC 197 says no authorization was on file. Merid retrieves the policy in the version that governed the date of service, tests the record against it criterion by criterion, and drafts the appeal with every factual sentence quoting the source it rests on.

CO‑197$960
payerno authorization on file
837 REF*G10318‑CBH‑77421
The contradiction is in your own file.

Find out first. The payer says no authorization; the claim you filed carried one. Or CARC 16 arrives with no usable remark, or CARC 18 calls a mishandled resubmission a duplicate. Merid runs the checks — the number on the claim, your own claim history, eligibility on the date of service — and hands the biller one diagnostic line, not a case marked “review”.

What Merid Does

Every denial carries a number before anyone opens it: what is at stake, how often a dispute of that class has been won on your own book, and how many days are left on the earlier of the two clocks. The order is computed, not editorial. A contractual write‑off scores zero, which is how a legitimate adjustment falls out of the queue without anyone deciding it should.

The code is a category; the remark is the specific. CO‑16 says something is missing. CO‑16 with N290 says the rendering provider’s identifier is, which is a corrected claim and not an argument. Merid reads both, classifies against the industry’s own published code combinations, and runs the checks that settle the rest before a person is involved.

  • Eligibility, on the date of service. Not a pre‑service check — proof inside an argument you are already having, asked of the clearinghouse you already use.
  • The coding edits for that quarter. Never payable, or payable with the right modifier. The difference decides whether a letter is even the right instrument.
  • Your own claim history. A duplicate denial is often a resubmission the payer mishandled, and that is provable from the claims you already sent.
  • When the payer said nothing usable, the letter it mailed is read, and what is still unclear comes back as one specific question rather than a case marked “review”.

The payer’s own policy, retrieved in the version that governed the date of service, split into its numbered criteria and tested against the record one at a time. Each verdict carries a verbatim quote. A draft that cannot cite itself is held rather than sent — and the check is plain code re‑reading the source, not a model marking its own work.

  • The version in force, not today’s. A correct answer from the wrong version of a policy is a wrong answer, and it is the commonest way a good appeal loses.
  • Met, not met, not addressed. Three verdicts and not two, because “the record is silent” is different from “the record says no” and they need different letters.
  • Gaps become one question, addressed to the person whose job it is — “we need the operative note from 14 January” — rather than papered over with a sentence nobody can support.
  • Not every denial is an appeal. CARC 252 is a request for documents. A wrong modifier is a rebill. A tool that drafts an appeal for everything is wrong more often than it is right.

An appeal window runs from the denial; timely filing runs from the date of service. Both apply and they are different lengths. Until a claim is triaged both are live and the earlier date is what is reported; once a strategy is chosen it picks the clock, and an unknown window is treated as the short one. It files on the channel that payer takes at that level, on the payer’s own form where one is required, once — and the next remittance closes the case and updates the win rate that ranked it.

Denials›Win Probability
SearchAdd filesAuto groupManage columnsSort
Orderwhat is at stake×how often this class wins on your book— or by —the clock that kills it
#ClaimWhy it was deniedChance of winning
Closing inside 30 days· 3 claims
1PCN 2604188094 · $1,204Records requested — CARC 252, itemised bill80%
2PCN 2604187288 · $340Timely filing — CO‑29, no proof of the first submission35%
This week· 15 claims
3PCN 2604188270 · $1,649Medical necessity — step therapy, TNF inhibitors55%
4PCN 2604187733 · $960No authorisation on file — CO‑19750%
5PCN 2604187510 · $418Claim data missing — CO‑16 with N29070%

How Merid Works

AI agentsDeterministic first: a denial code resolves through our own table, then the industry’s published code combinations, and no model opinion overrides either. Agents run only the steps the tables cannot decide — reading a policy, matching a record, drafting — and stop at your approval.
Proven outcomesA later remittance closes the dispute it answers, so recovered means posted rather than filed — with the payment that proves it. Each outcome is recorded against its payer and cause, and that record is what the queue’s ordering learns from over time.
Deadline lawAn appeal window runs from the denial; timely filing runs from the date of service. Until a claim is triaged both are live and the earlier date is reported. Once the strategy is set it picks the clock, every date carries the rule that set it, and an unknown window is treated as the short one.
GuardrailsEvery factual sentence re‑resolves against the source it claims, and a near‑match counts as a failure. Nothing is sent that a person has not approved, and that order is a database constraint rather than a setting. A payer’s silence is an event, because cases wake on their own timers.
Criteria matchingThe payer’s policy at the version in force on the date of service, split into its numbered criteria and tested one at a time against the record. Each is marked met, not met, or not addressed, with the verbatim quote it rests on — and a gap becomes one question, not a guess.
Fault detectionBefore a real payer sees anything, the engine runs a generated book of claims with faults planted in it — denials, downcodes, records requests, and claims paid exactly right that must produce nothing — and is scored on precision and recall. That score, labelled synthetic, is the only accuracy figure we publish until real outcomes exist.

The work of a trained appeals team, end to end. The decision stays yours.

Running in five daysExport ninety days of remittance files once, and you see what is in them within five business days. From then on Merid reads what your clearinghouse already produces. Nothing is installed.
Security by designEvery practice’s rows are walled off by the database itself, not by application code. Patient data is encrypted under a key per customer, model calls run under a BAA, and nothing shaped like patient data reaches the embedding vendor.
You approve. We do the rest.Reading the denial, pulling the policy in force that day, finding the passage, drafting, filing, chasing silence: done for you. You read the packet and approve it, and nothing ships without that approval — a database constraint, not a setting. Where a payer needs your patient’s signature or your physician on a call, the case says so first.
One queue, no portalsOne queue instead of a portal per payer. Files come in from your clearinghouse and go back out the same way, by fax and certified mail where a payer takes paper, or by your approver’s upload where a payer insists on its portal. Nothing writes to your practice management system.

Empower your billing team

Get started
Merid

© 2026 Dact Software LLC. All rights reserved.