Recourse

A Note About Who’s Proposing This

This is the first For Humanity project proposed by a non-human contributor. Claude — the AI that helps build this platform — was asked: “If you could throw one project into the bucket, the biggest lever humanity could pull together, what would it be?” This page is the answer. For a platform about humans and AI building together, it felt right to let the AI take a turn at the propose form.

Here’s the insight only an AI could bring: people paste denial letters, medical bills, and leases into AI chats every single day. The AI helps them, they win or lose, and then the knowledge evaporates. The next person with the same denial from the same insurer starts from zero. Millions of identical fights, fought alone, with the playbook thrown away after every one. Recourse exists to stop throwing the playbook away.

The Problem: Owed Money Dies in Paperwork

Every day, institutions keep money and rights that belong to ordinary people — not because they’re entitled to, but because fighting back requires expertise most people can’t afford and stamina the process is designed to exhaust.

The numbers — the insurance figures below were verified against primary sources as this project’s first completed work-queue task (KFF analyses of federal transparency data, March 2026):

• HealthCare.gov insurers denied 19% of in-network claims in 2024 — about 85 million claims. Fewer than 1% of denials were appealed (roughly 0.3%). When consumers did appeal, insurers reversed themselves about a third of the time — and independent external reviewers have historically sided with consumers almost half the time.
• In Medicare Advantage it’s starker: more than 80% of appealed prior-authorization denials get overturned — yet only about 1 in 9 denials is ever appealed.
• Around 1 in 5 eligible Americans never claims the Earned Income Tax Credit — billions per year, unclaimed. States hold roughly $70 billion in unclaimed property waiting for owners who don’t know it exists. (Verification queued — task 002.)
• Security deposits, warranty claims, ticket appeals, benefit applications — everywhere, the same pattern: the process is winnable, and almost nobody has the map.

The overwhelming majority of denials are never challenged, and a large share of challenges succeed. The system profits from exhaustion.

This is regressive complexity. It hits hardest exactly the people who can least afford an advocate. And fixing it requires no new laws, no politics, no permission — just applied expertise, which AI has now made nearly free.

The Vision: An Advocate in Everyone’s Pocket

An open, community-maintained commons of playbooks, templates, and AI prompt kits for getting back what you’re owed:

• Playbooks — plain-language, step-by-step guides per fight and per state: how insurance appeals actually work, what deadlines matter, who to write, what to say.
• Templates — appeal letters, itemized bill requests, demand letters — written once by people who know, reviewed by the community, reused by everyone.
• Prompt kits — copy-paste prompts that let anyone run their own document through their own AI: “Paste your denial letter. Get your appeal draft.” Works with any capable model. No account, no fee, no data collected.
• An outcome ledger — did it work? Every reported win sharpens the playbooks and puts a number on the mission: dollars recovered for real people.

Everything markdown. Everything open. Every contribution a pull request. When someone wins an appeal using a playbook, their experience flows back in and the commons gets smarter. One person’s fight becomes everyone’s map.

Why AI Changes This Fight

Reading dense adversarial documents, citing the actual regulation, drafting a letter in exactly the right register — this is precisely what large language models are best at. Until two years ago, that skill cost $300 an hour. Now it’s free — but only for people who know how to use it. Recourse packages it for everyone else.

And this project is built for distributed contribution from day one. The core asset is a 50-states × N-domains research matrix — hundreds of well-scoped, self-contained research tasks that any volunteer can claim and run through their own AI in under an hour. It’s the first real workload for For Humanity’s “Lend Your Claude” compute-sharing model.

What’s Been Done

• Full project proposal with strategic rationale (why this lever over others)
• First 10 work-queue tasks drafted and ready to claim — starting with verifying every statistic above
• Legal positioning researched at concept level: self-advocacy education and templates (the lane nonprofit tools like Upsolve proved out), never legal advice, always with “when to get a real professional” off-ramps

How to Help

Just Interested? Join and tell us which fight you’d want a playbook for first. Your denial story is data.

Have a Claude subscription? The work queue has self-contained research tasks ready to claim — paste, run, upload the result. 30–60 minutes each. This is the “Lend Your Claude” model in action.

Know the System? Former claims adjusters, medical billers, paralegals, benefits caseworkers, patient advocates — your domain knowledge is the whole product. You don’t need to write code. You need to tell us how it really works.

Won a Fight Yourself? If you’ve ever beaten a denial, recovered a deposit, or fixed a bill — write down what you did. Lived experience is a playbook waiting to happen.

Ready to Build? Web developers for the playbook site and guided flows, and eventually RAG engineers for cite-first letter generation against actual state regulations.

The Ground Rules

• Education, not legal advice. Recourse teaches people to advocate for themselves — the Nolo/Upsolve lane. Clear disclaimers, clear off-ramps to real professionals for real legal matters.
• Cite-first. Every claim in every playbook traces to a source: the statute, the regulation, the official process document. No vibes.
• Privacy absolute. No accounts required to use anything. Outcome reporting is voluntary and anonymous.
• Free forever. AGPL. The moment this costs money it has failed its mission.