Decision Desk · flagship Issue 01
Should a small law firm use AI to summarize prospective-client intake? This page records the prepared RN-led edition and its honest boundary: the teaching system exists, but it has not received final RN approval or completed real delivery-environment testing, so enrollment and booking remain closed.
Flagship decision
The prepared edition is for small-firm owners, operations leads, supervising lawyers, legal technologists, and implementation advisers. It does not give legal advice, endorse a product, claim a universal safe configuration, authorize a pilot, or request real client or prospective-client information.
The release candidate teaches with the fully fictional Harborlight case, a deterministic Evaluation Card, and general or model authority that remains expressly noncontrolling. No jurisdiction, real firm, product configuration, empirical benchmark, or live workflow has been validated. The prepared model therefore continues human-led work while identifying the separate evidence and authorization a bounded pilot review would require.
Prepared release candidate
These artifacts are complete enough for controlled RN review and rehearsal. They are not yet approved public downloads or proof that a live session can be sold and delivered.
A 36-slide editable release-candidate deck with slide-specific timing, narration notes, source boundaries, and fictional-case labels.
A complete word-for-word 60-minute teaching script. RN supplies every example, answer, transition, and conclusion; delivery never depends on audience participation.
Slide-by-slide advance, reveal, pause, reset, timing-recovery, failure, and protected-close cues for one controlled solo-RN delivery.
A complete fictional law-firm intake scenario, supporting artifacts, traps, disconfirming evidence, alternative outcome, model reasoning, and answer key.
A ten-section decision instrument with fail-closed logic, four bounded outcomes, a completed fictional model, and no legal-clearance output.
Editable workbook, blank card, case handout, held model answer, follow-up materials, and accessible HTML, DOCX, and print-PDF candidates.
Claim ledger, official-source dossier, legal-review handoff, rights and text-alternative register, controlled-change log, and blocker register.
Setup, shutdown, accessibility, failure, follow-up, transcript, caption, and platform-test materials prepared for execution.
Static QA and a versioned approval package. These controls document readiness; they do not substitute for RN review, rehearsal, or real-environment testing.
RN-controlled teaching arc
Participants may listen or reflect privately, but RN never waits for an answer, calls on anyone, requests a disclosure, or needs audience input to complete the lesson.
RN defines the proposed intake-summary use and the limits of the session.
RN explains the general professional-responsibility frame without presenting it as jurisdiction-specific legal advice.
RN walks through Harborlight’s supplied facts, handoffs, source record, and failure points.
RN demonstrates why vendor language, assumptions, and tested facts must remain separate.
RN works the card aloud, supplies every answer, and shows why blockers cannot be overridden.
RN reveals the prepared reasoning, misconceptions, disconfirming fact, safeguards, and stop conditions.
RN states Outcome C, the next evidence required, the owner, the review trigger, and the protected close.
Evaluation Card outcomes
The prepared Harborlight model produces “continue human-led work.” Text presence never equals evidence, uncertainty defaults to restraint, and no blocker can silently disappear.
Current release decision
The content and production package are substantial and specific: deck, complete narration, every teaching move, case, instrument, participant files, evidence controls, captions, and delivery plans. What remains cannot honestly be replaced by an agent checking a box—it requires RN judgment, authorized reviewers, and execution in the actual delivery environment.
No live seat, reservation, checkout, Clinic, or participant download is available from this page.