Appeal
Decode the denial. Win the appeal.
Decode the denial. Win the appeal.
- 01Show your work before building on it
- 02Decode the real reason — the empathy moment
- 03A calibrated verdict, with citations
A 78-screen iOS system on one design system, built on traceable citations and honest, calibrated confidence
Fewer than 1% of denied health-insurance claims are ever appealed — yet 40–80% of the appeals people do file succeed. The gap isn’t the merits, it’s the giving up: denials are written to be opaque and exhausting. Appeal closes that gap. Photograph a denial letter and it decodes the real reason behind the boilerplate, surfaces the specific grounds that win — each cited back to your own documents — and drafts an appeal you edit, own, and send. It’s a calm, evidence-led decoder that never pretends to be a lawyer.
About one in five health-insurance claims is denied, but under 1% are appealed — not from laziness, but by design. The deterrent is administrative burden: the learning, paperwork and emotional cost of fighting back. 69% of denied patients don’t know they can appeal and 85% never file. And the denials people surrender to are mostly beatable — the largest category is ‘other’ (no stated reason) followed by administrative errors; genuine medical-necessity calls are a small minority. Letter-writing is already commoditised; what no one owns is the real bottleneck — the moment at the kitchen table of ‘what does this even mean, do I have a case, what do I say?’ The cautionary tale is DoNotPay, whose FTC settlement reads like a spec of what not to do: never tested against lawyers, hallucinated documents, faked endorsements.
Giving up is the design problem
The barrier to appealing isn’t the merits — most filed appeals win. It’s administrative burden: not knowing you can, and the exhaustion of trying.
Opacity is leverage, not a dead end
The biggest denial category has no stated reason. That looks like a wall, but a missing rationale is itself a weakness you can appeal on.
Trust is the product, not the model
Letter-writing is commoditised. What wins is traceable citations, calibrated confidence, and a user who is always the author — never a tool pretending to be a lawyer.
A denial is written to be opaque — dense, procedural, easy to give up on.
Appeal highlights the decisive clause and translates it, cited back to your own letter.
From a confusing letter to a sent appeal you own — understanding first, then action.
Show your work before building on it
After reading the letter, Appeal never silently proceeds. It shows the fields it extracted over a thumbnail of your actual letter; tap any field to see the exact source region. High-confidence reads are quiet, low-confidence reads get an amber ‘tap to confirm’, and nothing downstream runs on an unconfirmed read. That’s the foundation that makes every later citation trustworthy.
Decode the real reason — the empathy moment
The hero screen is a split: your scanned letter with the decisive clause highlighted in brick, and a plain-language ‘what this actually means’ beside it. The 36%-of-denials ‘they didn’t actually say why’ case isn’t a failure state — it’s reframed as leverage: ‘that’s common, and it’s a weakness we can use.’
A calibrated verdict, with citations
The verdict pairs a strength meter with an honest base rate (‘appeals like yours succeed about X% of the time’), and every ground carries a citation chip that traces to the exact line in your own document or a named regulation — never an unverifiable claim. Weak cases get an honest ‘this is a harder case’ path; genuinely complex ones are routed to a lawyer or advocate.
A draft the user owns and sends
The appeal is presented as a finished letter you can edit anywhere; AI-authored runs carry a subtle marigold rule so authorship is always legible, and refine suggestions come as accept/reject diffs. The flow forces a ‘review before you send’ checklist and an explicit ‘I’m sending my appeal’ — deliberate friction that is both a trust feature and the legal firewall: the user is always the author and the sender.
Never paywall the decode
Understanding your denial and seeing your grounds is always free; payment is asked only at send, with a hardship path that grants full access. The ethical north star is to not monetise desperation — the inverse of the failure mode the product was designed against.
- ✕Boilerplate you can’t decode
- ✕A letter with unverifiable claims
- ✕Confidence with no basis
- ✕Tools that imply they’re lawyers
- ✓The real reason, in plain language
- ✓Grounds cited to your own documents
- ✓An honest strength meter and base rate
- ✓You’re always the author and sender
The journey at a glance — decode, argue, send.
The Decode — the boilerplate reason translated into plain language, the decisive clause highlighted in your own letter.
The Verdict — a calibrated strength meter and an honest base rate, with each ground cited back to your documents.
The Draft — the appeal in your name, AI-authored runs marked, leverage points pinned to your evidence.
The case timeline — filed, sent, under review, outcome — so you always know what happens next.
Trust as a feature — local-first, HIPAA-grade handling, surfaced before anything is asked of you.
Appeal is a concept and I’ve framed it as one: no real users, no launch, no shipped metrics. What it shows is a complete, coherent system — 78 screens across 19 flows covering capture, show-your-work decode, adaptive intake, the cited verdict, the user-owned draft editor, send-and-track, the evidence locker, deadlines, learn, escalation, account, security and app-lock, privacy controls, the ethical paywall, and the full edge-and-error set — all on one design system, with traceable citations and calibrated confidence as the spine. The honest test is an expert review: would a patient advocate accept that this helps a layperson understand and contest a denial without ever overclaiming or posing as a lawyer? That’s the bar I designed to.
- Designing against an anti-pattern (DoNotPay’s FTC settlement) was clarifying — every trust mechanic is a direct answer to a documented failure.
- Traceability is the whole product. A citation you can tap back to your own letter is what separates this from a confident letter-generator.
- Honesty scales down as well as up — the weak-case path and the ‘see a lawyer’ route earn the trust that the strong-case path spends.
- A 78-screen system forced real systems discipline — one verdict component, one citation chip, one strength meter, reused everywhere.
