No. 202 / 339

AI drafts a passable will in seconds — is the value the document, or the state-law validity and accountability when it's contested after death?

The shift

Generating a plausible, correctly-formatted will or trust drops from scarce drafting labor — attorney or paralegal hours — to near-instant, near-free output any layperson can produce. What stays scarce is state-specific execution validity, judgment on contested or complex estates, and a professional who is answerable when the plan fails — a failure that, uniquely here, only surfaces after the client is dead and can no longer fix it.

The axioms

  1. Drafting a will or trust is skilled legal labor, so the document is the thing the client pays for.
  2. Laypeople can't produce a competent estate-planning document themselves, so they hire someone who can.
  3. A will is legally operative only if executed under the right state's formalities — signature, witnesses, sometimes notarization or a self-proving affidavit.
  4. Complex or contested estates — blended families, business interests, tax exposure, disputed capacity — require judgment, not a template.
  5. A licensed, insured professional stands behind the plan and is accountable if it's wrong.
  6. The instrument is tested at the worst possible moment: after the client is dead, in probate, when no one can amend it.

Invalid axioms

  1. Drafting the document is skilled labor, so the document is what the client pays for. Producing a clean, well-structured will or trust is now abundant and near-free. The habit-trap: estate planners still price and package the engagement around the artifact — a flat fee "for the will," a tiered "trust package" — billing for the drafting as if generating the pages were the scarce, hourly-metered good, when the drafting is the part AI already does.
  2. Laypeople can't produce a competent estate-planning document themselves. Translating intent into legally-styled clauses is now free for anyone with a chat window, and consumer will-generators predate LLMs. The habit-trap: the low end of the market — simple wills, basic revocable trusts, powers of attorney — was priced on the client being unable to draft it themselves. That gate is gone even where the document still needs valid execution to work.

Unchanged axioms

  1. A will is operative only if executed under the right state's formalities. This rests on statute, not on drafting skill — witness requirements, notarization, self-proving affidavits, and holographic-will rules vary by state and don't relax because the text was generated cheaply. A perfectly-worded document that isn't executed correctly is void, and the model producing the words is the part that was never the hard constraint. (Moving-part flag: a handful of states now recognize electronic and remote-witnessed wills, and that map is shifting — the formality survives, but its specific form is in flux state by state.)
  2. Complex and contested estates require judgment, not a template. Blended families, business succession, estate-tax exposure, special-needs trusts, and anticipated will contests turn on facts with no clean pattern to match — and on adversarial pressure from heirs with money at stake. This is judgment under novel, high-stakes ambiguity, not clause assembly.
  3. A licensed, insured professional stands behind the plan and is accountable when it's wrong. Malpractice liability, E&O coverage, and bar or notary discipline attach to a person. A model can't be sued, disciplined, or made to answer an angry beneficiary — and because the client is dead when the defect surfaces, the accountable party is the only one left to hold responsible. The suability is the product.
  4. The instrument is tested after the client is dead, when no one can fix it. This is the structural feature that makes estate documents unlike almost any other AI-drafted artifact: there is no feedback loop, no chance to correct, no live client to clarify intent. Getting it right the first time, verified against the governing state's law, is the whole job — and verification against ground truth is exactly what the model doesn't guarantee.

New axioms

  1. A flood of plausible-but-invalid DIY wills that only break in probate. Confidently-formatted output is the default; a document that reads like a valid will but was signed with one witness in a two-witness state, or wasn't notarized where required, looks fine until the person is dead and a court refuses to admit it. The failure is silent for years and then irreversible — nobody has built the mechanism that catches an invalid instrument while the client is still alive to re-execute it.
  2. Who bears the risk when a free AI will fails. The drafting tool disclaims everything, the deceased can't be made whole, and the loss lands on beneficiaries who didn't choose the tool — or on the estate via intestacy overriding what the person actually wanted. Liability for a defective self-generated instrument has no clear owner, unlike a defective lawyer-drafted one.
  3. Unauthorized-practice and notarial-fraud boundaries tested at consumer scale. AI tools now hand the public documents that need witnessing and notarization to become valid, faster than bar associations and notary regulators can rule on what counts as practicing law or on how remote/AI-mediated execution should be policed — the same regulatory lag the broader legal field faces, sharpened by the death-triggered stakes.
  4. Pricing once the document is free. When the artifact costs nothing to produce, estate planners have to charge for validity assurance, judgment, and standing-behind-it — but no settled model exists for billing "I verified this will actually work in your state and I'm liable if it doesn't" instead of billing for the pages.

Where it breaks

Planners still sell the will or trust as a priced artifact (invalid), while the thing a client actually needs — a document that survives probate in their state — depends on execution validity and a professional who's liable if it doesn't, verified before a death that ends any chance to fix it (new). The market is charging for the part AI made free and under-pricing the part that only matters after the client can no longer complain.

A second collision: the consumer-DIY moat is gone, so people generate their own wills for nothing (invalid), while the failure mode of those documents — silent invalidity surfacing only in probate — is exactly the case with no feedback loop and no accountable party (new). The cheapest path produces the artifact most likely to fail at the one moment it can't be corrected, and no one has decided who owns that risk.

Related axioms

Other axioms