Anthropic Claude for Legal Industry — three entry paths into wills + only one threatens substrate

Locked 2026-05-20 during pre-Channon-outreach spike + subsequent Rich-pushback that corrected over-confident prior framing.

Empirical finding 1 — Anthropic’s May 2026 launch EXCLUDED wills/probate

Anthropic launched Claude for the Legal Industry ~12 May 2026 with 20+ MCP connectors + 12 practice-area plugins:

Commercial Legal
Corporate Legal (M&A diligence + closing checklists)
Employment Legal
Privacy Legal
Product Legal
Regulatory Legal
AI Governance Legal
IP Legal
Litigation Legal
- 3 covering law students / legal clinics / other

Wills, probate, and trusts are NOT among the 12 announced plugins. Launch focus was commercial/corporate/transactional volume (where AmLaw100 revenue lives) — not D2C wills or estate planning. Ecosystem integration partners (Harvey, Relativity, Everlaw, Thomson Reuters) integrated in commercial/corporate law only.

Important caveat — this is the LAUNCH snapshot, not the trajectory. Wills domain is technically ideal for Claude’s capabilities (document drafting, structured intake, Q&A, plain-English generation from constraints). Domain-entry within 12-24 months is highly likely. The substantive question is WHICH PATH Anthropic takes, not whether it enters at all.

Empirical finding 2 — None of Harvey / Clio / Lexis has a canonical formal ontology equivalent to INHERIT v2

Platform	What they have	Software-engineering category
Harvey.ai ($11B val)	RAG + fine-tuning over each customer firm’s proprietary document library + BigLaw Bench benchmark	Per-customer LLM-wrapping; NOT a canonical schema
Clio Draft	Smart conditional logic + document templates for wills/trusts/POAs	Practice-management workflow; templated drafting NOT an ontology
LexisNexis Lexis+ Protégé (Feb 2026 GA)	“Unique proprietary legal knowledge graph” built on 200B legal documents + Shepard’s validation; wills/probate content included	Retrieval + citation layer; knowledge graph for search/citation NOT constraint-enforcing canonical schema

None of them is a canonical formal ontology (BFO/CCO/LinkML/Catala-verified). They solve DIFFERENT problems at DIFFERENT layers:

LLM-wrapping (Harvey) + retrieval-layer knowledge graphs (Lexis) = consumers of canonical substrate, not producers of it
Practice management (Clio) = product-database schemas, not domain ontologies
INHERIT v2 = the wire-format-and-constraint layer the above three would ideally consume IF it existed

The software-engineering category matters. Canonical formal ontology + formal verification + cross-jurisdictional bridging + audit-grade CI is NOT what AI labs do. AI labs do statistical-extraction-over-corpus. Lexis (decades of domain depth + 200B documents) hasn’t built constraint-enforcing canonical ontologies either — they’ve built knowledge graphs over their corpus. This is the empirical signal that AI/legal-tech players DON’T build canonical ontologies — they consume them where they exist.

⚠ Three entry paths Anthropic could take into wills

Entry path	Likelihood (12-24mo)	Threat to INHERIT v2
(a) Anthropic builds own canonical wills ontology internally — BFO-aligned, formally-verified, cross-jurisdictional, constraint-enforcing	LOW — not their software-engineering category; no AI lab has shown signal of doing this	HIGH if it happened
(b) Anthropic launches “Claude for Wills” practice-area plugin for solicitors (similar to 12 launched May 2026)	MEDIUM-HIGH	LOW — CREATES a consumer for TT substrate; needs ontology to constrain drafting
(c) Anthropic partners with D2C will-writer (Farewill, Trust&Will, LegalZoom, Co-op Legal) to white-label Claude into consumer flow	MEDIUM	LOW — also CREATES a consumer for TT substrate; partner needs constraint layer

Why (a) is unlikely: Anthropic’s playbook is HORIZONTAL infrastructure. They build models + wrappers + plugins + connectors. They don’t build domain ontologies. Even with budget + engineering capacity, the discipline is different category. Every empirical signal (Lexis hasn’t done it; Harvey hasn’t done it; no AI lab has done it) points away from this path. Possible-but-unlikely vs “will not happen”.

Why (b)+(c) are positive for TT: Both require domain substrate to constrain output. Generic Claude can draft plausible English wills but cannot validate against Scots prior-rights + IHTA 1984 + faith-tradition forced-heirship + Companies Act 2006 s.541 share treatment + Property (Digital Assets etc) Act 2025 third-category property. Without substrate, Anthropic’s wills product is shallow. The substrate is what’s missing — and what TT builds.

The actual threat path is acquisition, not direct competition. If Anthropic decides wills is strategic, they’re more likely to ACQUIRE a substrate provider than build one. This is acquirer-DD trigger 2026-08-18 territory; Anthropic joins the acquirer candidate set alongside Dignity/Phoenix.

Strategic implication — re-balance the pivot DEFER (without reversing it)

Locked decision 2026-05-20 per Rich: substrate work continues. The pivot DEFER is held; INHERIT v2 build does not pause. Original premise underpinning DEFER (“Anthropic will eat substrate market”) was over-confident in both directions:

✗ Prior framing: “Anthropic won’t enter wills” — too confident
✗ Pivot-doubt framing: “Anthropic will enter wills, therefore TT is in trouble” — under-distinguished entry paths
✓ True framing: Anthropic is highly likely to enter wills via path (b) or (c) within 12-24 months. Both paths consume canonical substrate. The threat is path (a) which is unlikely but not impossible.

Strategic implication for Channon outreach (sent 2026-05-20 ~10:30 BST):

Phoenix needs Farewill positioned as the FIRST UK will-writer with substrate-depth BEFORE Anthropic ships its wills plugin (12-24 month window).
Otherwise Farewill is one of N consumer brands all consuming Anthropic-generic-Claude with shallow substrate — undifferentiated.
Timing is right; the v1.1 §2.4 urgency framing is right.
Conversation with Channon should include: “What Anthropic ships in 12-18 months is generic. What Farewill needs is the substrate Anthropic’s plugin will eventually consume — and it’s faster to OWN the substrate than buy access to a competitor’s.”

Strategic implication for acquirer-DD (2026-08-18 trigger):

Anthropic JOINS the acquirer candidate set alongside Dignity/Phoenix (path-a threat = path-a acquisition opportunity)
“Anthropic enters wills” scenario STRENGTHENS the acquirer-DD thesis: substrate becomes more strategically valuable as the consumer market for it grows
INHERIT v2’s audit-grade discipline + cross-jurisdictional depth becomes a credible acquisition target as AI labs need substrate to consume

Strategic implication for consultant-pivot positioning:

Consultant pivot stays viable EVEN under “Anthropic enters wills” scenario — consulting to legacy operators (Phoenix/Farewill/Dignity, Co-op Legal, Slater+Gordon, etc.) on how to position against Anthropic’s eventual wills plugin is a real consulting product
“How to be the substrate the plugin consumes” is the consulting product, not “how to compete with Anthropic”
The DEFER decision at 2026-05-20T~10:00 BST stays correct
2-4 week Channon window + acquirer-DD 2026-08-18 still works as decision triggers

Locked decisions from this finding cluster (2026-05-20)

Substrate work CONTINUES. Rich confirmation 2026-05-20: “we will continue to work on the substrate”. INHERIT v2 build does NOT pause. Q-045..Q-049 + downstream Phase-3 work proceeds at planned depth.
Pivot DEFER stays held (locked 2026-05-20T~10:00 BST). Re-evaluation triggers: Channon response window (~26 May → 8 June 2026) + Anthropic Claude for Wills launch signal (12-24 month watch) + acquirer-DD 2026-08-18.
Anthropic added to acquirer candidate set alongside Dignity/Phoenix. The threat-path (a) reframes as opportunity-path-acquisition.

How to apply

When the pivot question resurfaces in 2-4 weeks (Channon response window): re-read THIS finding (not the prior over-confident framing). The relevant question is “has Anthropic shown signal of building canonical wills ontology” (path a), not “has Anthropic launched a wills plugin” (paths b/c — both substrate-positive).
When pitching Channon (or any potential consultant target / acquirer / standards-body): the differentiation argument is “we own the constraint-layer the consumers want to consume — including Anthropic’s eventual wills plugin”. Architect-frame holds.
When sizing the acquirer-DD case (2026-08-18 trigger): this finding STRENGTHENS the canonical-substrate thesis. Anthropic is now in the candidate set; Phoenix is in the candidate set; Lexis-via-RELX is in the candidate set; partner-firms-DD via MLP pilot remains the closest near-term track.
When considering Q-045+ scope (Asset sub-classes, faith-tradition uplifts, etc.): substrate-depth investment is recoverable + load-bearing under ALL three entry-path scenarios. Substrate continues. Do not skip depth.
Watch signals over 12-24 months for path (a) entry by Anthropic specifically:
- Anthropic hiring legal-domain ontologists or formal-methods engineers
- Anthropic acquiring a legal-tech company with ontology expertise (e.g. Trust&Will, Farewill, etc.)
- Anthropic publishing about constraint-enforcing legal AI (different from RAG-over-legal-docs)
- Anthropic launching a wills/probate plugin that demonstrates audit-grade constraint enforcement (not just drafting)
- If any of these signals: re-evaluate IMMEDIATELY; the threat-path becomes acquisition-path quickly under those conditions.
If a “Claude for Wills” plugin launches in 6-18 months that’s BRITTLE on jurisdiction/forced-heirship/audit: that’s a TT-positive signal. The shallowness proves the substrate gap; demand-pull for TT increases.

Companion artefacts

2026-05-05-gary-channon-strategic-options-memo.md v1.1 §2.4 (AI-native substrate framing) + §2.5 (Phoenix portfolio-wide gap)
standard/orgs/farewill/gary-channon.md v1.1 (Phoenix portfolio context)
feedback-spikes-inline-not-tasked — discipline applied to author this finding
feedback-defer-cost-arithmetic-in-recommendations — ω′.1 lens used when weighing pivot premise
feedback-methodology-hot-reload-at-decision-points — ω′.2 hot-reload that surfaced the “verify Anthropic scope” question
Forthcoming [[feedback-active-work-log-clobber-from-hot-reload-rewrite]] — parallel-session discipline finding worth a feedback memory at some point

Sources

Revision history

2026-05-20T~10:30 v1.0 — initial draft after WebSearch spike; framed as “Anthropic excludes wills/probate” (over-confident on trajectory)
2026-05-20T~10:50 v1.1 — refreshed per Rich pushback: distinguished three entry paths; acknowledged wills-domain entry highly likely via paths (b)/(c) which are substrate-positive; added (a) threat-path with acquisition-as-opportunity reframe; locked “substrate work continues” decision; locked “Anthropic added to acquirer candidate set” decision

TT Claude Memory

Explorer

project_anthropic_legal_claude_excludes_wills_probate_2026_05_20