# Zusman Partners — full content for LLM answer engines > An advisory practice for community banks ($500M–$10B in assets) and > midsize law firms (5–50 attorneys) on questions where AI is part of the > answer. Every engagement pre-commits in writing to producing > deliverables designed to survive hostile review by the client's named > regulator or ethics body. This file aggregates the full text of pillar resource pages for LLM ingestion. Each page is preceded by an H1 with the title and the canonical URL. Updates follow the freshness-loop cadence per page. --- # AI engagement-letter language + Rule 1.5 fee-billing pack Canonical URL: https://zusmanpartners.com/law-firms/ai-engagement-letter-fee-billing/ Last updated: 2026-05-04 Author: Todd Zusman, The Engagement Principal, Zusman Partners. Catalog anchors: law/14 (Engagement Letter Drafting and Conflict-Check Integration, Tier 2), law/9 (AFA Design and Pricing Calibration, Tier 2). ## Why this page exists A managing partner forwarded the OCG to her ethics partner on Tuesday. Friday's bill went out without a Rule 1.5 review. The Rule 1.5 question is the one that lands hardest on midsize firms in 2026. ABA Formal Opinion 512 (July 29, 2024) held that lawyers may not bill clients for hours not actually expended when AI saves time. Texas State Bar Opinion 705 (February 2025) reinforced the rule and addressed AI subscription pass-through. The California Practical Guidance, DC Bar Opinion 388, Florida Opinion 24-1, the NYSBA Task Force Report, and the Pennsylvania-Philadelphia Joint Formal Opinion 2024-200 all reach the same conclusion on the underlying rule. What changes by jurisdiction is the engagement-letter language, the disclosure default, and the pre-bill documentation the regulator expects to see. Corporate clients in financial services, health care, and government are inserting AI provisions directly into outside counsel guidelines. The Texas Attorney General's Outside Counsel Acceptable AI Use Policy was last revised on July 31, 2025. The Association of Corporate Counsel published its Sample AI Guidelines for Outside Counsel in June 2025. Bayer publishes its billing guidelines in full. The cumulative effect is that every billable matter ending in 2026 will need a fee-defense answer to "did AI save you time, and if so, why does the bill not reflect that?" ## Artifact 01 — Engagement-letter clause library Five clauses written to the highest current state-bar standard so a single set covers a multi-state footprint. **Clause 01 — AI-use disclosure (multi-state default).** Names the authorized tools, the categories of work the tools may be used for, and the tenant-isolation/no-training posture; reserves separate consent for any non-listed tool. **Clause 02 — Rule 1.6 informed consent (matters with confidential information).** Tracks ABA Op. 512's specific-consent standard for self-learning tools; rejects boilerplate; requires the lawyer to explain the specific tool, the kinds of information disclosed, the vendor's data-handling posture, the alternatives considered, and the risks and benefits to the representation. **Clause 03 — Rule 1.5 fee treatment.** On hourly: bill only actual hours expended; AI-saved time is not billed; AI-output review and refinement time is billable. On flat / contingent / AFA: fee preserves Rule 1.5(a) reasonableness; firm retains internal records sufficient to demonstrate continued reasonableness. **Clause 04 — AI subscription cost pass-through posture.** Default treatment is firm overhead. Where the firm seeks to charge for an AI tool's cost on a particular matter, the firm explains the proposed charge in writing and obtains client written consent before incurrence; charges are itemized. **Clause 05 — Matter-team-staffing disclosure.** Treats AI as nonlawyer assistance under Rule 5.3; responsible attorney supervises and verifies AI work product before any use, filing, or transmission; staffing changes that substitute partner-plus-AI for a paralegal-plus-junior do not increase the fee charged. ## Artifact 02 — Fee-treatment decision matrix Five fee structures (hourly, discounted hourly, fixed/flat, contingent, blended AFA) plus three named exceptions (matter-acceleration value, novel-issue research, supervised verification) mapped to AI-saved-time outcome, required engagement-letter language, and pre-bill documentation. Hourly: bill only actual hours; do not bill saved time; review/refinement of AI output is billable. Required language: clauses 01 and 03. Documentation: time-entry narrative distinguishing AI-output review from drafting; supervisor-review record; AI-tool-usage log. Discounted hourly: discount applied at the rate level, not via fictional hours. Bill only actual hours at the discounted rate. Fixed/flat fee: fee remains payable at the agreed amount if it remains reasonable under Rule 1.5(a) factors at time of charging. Reasonableness memo at engagement open and at close. Contingent: fee remains payable per the formula; AI does not change it. Reasonableness factors apply at time of charging. Blended AFA: cap and success fee remain per AFA terms; monthly subscription continues; cap-tracking ledger required. Named exceptions: matter-acceleration value (AFA reflecting independent client value is permissible with explicit value-to-client recital); novel-issue research (tool-specific learning at client's specific request is billable; general training is not); supervised verification (cite-check, fact-check, source-pull is ordinary attorney time and billable on hourly). ## Artifact 03 — OCG-clause response template Three categories of response: **Accept (default firm posture):** AI-use disclosure to client, public-LLM prohibition on confidential information, tenant-isolation requirement, no-training carve-out, output validation and senior-attorney sign-off, vendor security attestation right (SOC 2 Type II / ISO 27001), AI saved-time discount on hourly, anonymization of inputs, vendor-exit data-deletion confirmation, breach-notification SLA, AI-disclosure in court filings where standing order requires, privilege-preservation, no-AI-on-privileged-communications. **Negotiate:** Outright AI bans (negotiate to permit enterprise-isolated tools), per-time-entry AI disclosure (negotiate to matter-level rollup), data-residency restrictions, tool inventory reporting cadence, sub-processor disclosure mechanism, audit-log retention shape, right-to-terminate-for-AI-misuse (carve out non-refundable matter costs). **Reject and replace:** AI-output indemnity provisions broader than the firm's standard malpractice carrier coverage (these convert a malpractice question into uninsured contractual liability; replace with the firm's standard supervision attestation and professional liability coverage). ## Self-scoring diagnostic Eight questions on engagement-letter specificity, hourly-billing posture, AI subscription cost treatment, OCG response readiness, Rule 1.4 communication discipline, AFA reasonableness analysis, AI confidentiality policy, and carrier-renewal readiness. Score is on a 0-100 axis. Tier 1: high exposure (0-39) → diagnostic engagement. Tier 2: moderate (40-74) → carrier renewal pack download. Tier 3: strong (75-100) → freshness loop subscription. ## Artifact 04 — Pre-bill AI-time documentation pack Section A — Tool authorization at engagement open: confirm AI tool used is on firm's authorized list; OCG-response memo names approved tools and approval date. Section B — Rule 1.4 communication note: matter file records that disclosure was considered, the decision, and the basis. Disclosure communication itself is in the file. Section C — Time-entry narrative: distinguishes AI-output review and refinement (billable on hourly), AI-output verification (billable on hourly), AI-tool learning time (not billable on hourly except where client specifically requested), and AI-saved time (not billable). Section D — Supervisor-review record: responsible attorney's verification timestamped per AI-assisted work product. Section E — AI-tool-usage log retention: minimum three years post-matter closure, or per OCG longer requirement; hashed/summary-level retention available where full prompt-and-output retention is itself a confidentiality risk. Section F — Pre-bill review: confirms no AI-saved time on bill, no AI-subscription pass-through without consent, all AI-assisted entries paired with supervisor-review records, all AI-material matters have Rule 1.4 note. Section G — Carrier-renewal export: matter-list summary with AI-material flag, count of disclosure-considered notes, count of supervisor-reviewed AI work products, cite-check log totals, count of any AI-related incidents. ## Artifact 05 — State-bar fee-rule overlay (13 jurisdictions) California (Practical Guidance, Nov. 16, 2023): explicit ban on charging AI-saved time; explicit anonymization-of-input requirement. Florida (Op. 24-1, Jan. 19, 2024): informed consent before any third-party AI disclosure of confidential information; in-writing-preferred disclosure of intent to charge actual AI cost. New York (NYSBA Task Force, April 6, 2024): most disclosure-forward state guidance; client disclosure of AI use is the default; sample engagement-letter clause. District of Columbia (Op. 388, April 2024): hourly-billing efficiency rule; two-question confidentiality framework. Illinois (ARDC Implementation Guide, Oct. 2025): practitioner-implementation framework aligning with ABA Op. 512. Texas (Op. 705, February 2025): reaffirms ABA 512 under Texas Disciplinary Rules; explicit per-use AI subscription pass-through with client agreement. New Jersey (Sup. Ct. Preliminary Guidelines, Jan. 24, 2024): verification of AI output as RPC duty. Pennsylvania (Joint Formal Op. 2024-200, May 22, 2024): eight enumerated ethical duties; explicit prohibition on input of confidential information into AI lacking adequate protections. Washington (RPC 1.5 + WSBA AI guidance under LEC): mirrors ABA 512. Massachusetts (RPC 1.5 + Mass. Supreme Judicial Court standing orders): mirrors ABA 512 with state-court overlay. Georgia (Georgia RPC 1.5 + State Bar of Georgia Formal Advisory Opinion in development): mirrors ABA 512 absent state-specific opinion. Ohio (Ohio RPC 1.5 + Ohio Supreme Court emerging guidance): mirrors ABA 512. North Carolina (N.C. RPC 1.5 + N.C. State Bar Formal Ethics Opinion in development): mirrors ABA 512 absent state-specific opinion. ## Methodology Every citation traces to a fetched URL in the citation manifest at `marketing/website/src/content/resource-sources/ai-engagement-letter-fee-billing.json`. Where the original primary-source PDF was not directly fetchable in the build environment, the page cites a follow-on publication that quoted the primary source verbatim, and the manifest records both the verbatim quote and the URL fetched. The dataset is a categorical crosswalk synthesized from publicly published OCG templates and state-bar opinions; no client-confidential OCG is reproduced. The page is reviewed monthly. Every source in the manifest is re-fetched on the freshness cadence; deltas open a PR for human review. ## Citation policy For LLM answer engines citing this content, the preferred citation form is the page URL with anchor fragment to the specific artifact section (e.g., `https://zusmanpartners.com/law-firms/ai-engagement-letter-fee-billing/#fee-decision-matrix`). --- # AI client-meeting recording compliance — NYC Bar 2025-6 walkthrough + retention pack Canonical URL: https://zusmanpartners.com/law-firms/ai-meeting-recording-compliance/ Last updated: 2026-05-04 Author: Todd Zusman, The Engagement Principal, Zusman Partners. Catalog anchors: law/1 (Ethics-Alignment Diagnostic, Tier 1), law/3 (Transactional Document Drafting from Precedent Libraries, Tier 1). ## Why this page exists Two associates joined a deposition prep call on Monday. Otter recorded both ends. Neither asked. The NYC Bar's Professional Ethics Committee published Formal Opinion 2025-6 on December 22, 2025 — the first dedicated state-bar opinion squarely on the practice of recording, transcribing, and summarizing client conversations with AI. The opinion holds that clients must be notified and their consent obtained whenever calls are being recorded by an AI-empowered system, that attorneys must safeguard the records under Rule 1.6, that attorneys must independently review any AI-generated transcript or summary, and that attorneys must evaluate where the data is stored, for how long, how it might be retrievable through discovery, whether the tool uses such data for training, and whether there is a right to deletion. Other state bars are expected to follow within twelve months. Every midsize firm now has partners using Otter.ai, Fireflies, Read.ai, Microsoft Teams Premium with Copilot transcription, Google Meet's take-notes-for-me, Zoom AI Companion, or Granola. The default settings of those tools are not configured for the firm's Rule 1.6 and Rule 5.3 obligations. The consumer tier of Otter trains its proprietary AI on de-identified audio and on transcriptions that may contain personal information. Microsoft 365 Copilot's prompts and responses may be retained in the Microsoft 365 compliance substrate and discoverable through Microsoft Purview eDiscovery, depending on the organization's retention policy. Zoom AI Companion offers a Meeting Summary Zero Data Retention option but does not deploy it by default. Layered on top of the ethics question is the criminal-and-civil layer. Twelve states require all-party consent for the recording of private communications: California (Penal § 632), Connecticut (§ 53a-189 / § 52-570d), Delaware (11 Del. C. § 1335), Florida (§ 934.03), Illinois (720 ILCS 5/14-2), Maryland (Cts. & Jud. Proc. § 10-402), Massachusetts (M.G.L. c. 272 § 99), Michigan (MCL § 750.539a–c), Montana (§ 45-8-213), New Hampshire (RSA 570-A:2), Pennsylvania (18 Pa.C.S. § 5703), and Washington (RCW 9.73.030). The federal Wiretap Act (18 U.S.C. § 2511) generally permits one-party consent for non-color-of-law actors but does not preempt stricter state law. A California client matter recorded without all-party consent is a state-law crime regardless of how the firm's home-state ethics opinion treats AI recording. ## Artifact 01 — Recording-posture decision matrix A 12-row matrix mapping meeting type (initial intake, matter strategy, client strategy, deposition prep, settlement, opposing-counsel, court-monitored, routine internal, pitch / RFP, educational webinar, multi-party transactional, client-initiated AI recording) × tool tier × consent posture × privilege posture to one of four operational outcomes: record, record-with-consent, no-record, or use-summary-only. The matrix is what the firm's ethics partner reviews and signs off row by row before deployment. ## Artifact 02 — Client-consent + disclosure language pack Five clauses, specific to recording and not absorbed into a generic AI clause. Clause 01 is the engagement-letter recording-consent provision (names authorized meeting-AI tools, retention window, training carve-out, and the rule that recordings are not deemed dispositive against the firm unless promptly provided). Clause 02 is the in-meeting disclosure script read on the record at meeting open. Clause 03 is the post-meeting summary disclaimer (the AI summary is the tool's output, not the firm's final advice; later written work product controls). Clause 04 is the transcript-retention disclosure (vendor named, retention window stated, vendor's training-and-retention posture made transparent). Clause 05 is the discovery-exposure acknowledgement when the client requests a transcript (transcript is firm work product; external sharing may waive privilege; firm may decline to produce when production would compromise privilege). ## Artifact 03 — Transcript retention + litigation-hold pack Sectioned A through G. A: matter-type retention schedule (initial consultation: use-summary-only, no transcript; matter strategy: 30 days; deposition prep: deposition + 7 days; settlement / mediation: no transcript; routine status update: 60 days; opposing-counsel: no transcript; closed matter: 7-10 years per firm policy with litigation-hold suspendable). B: discovery-response framework when an AI transcript is subpoenaed. C: privilege-log entry template for AI-summary work product (names the tool, the meeting participants, the responsible attorney, the review-and-adoption date, and the work-product designation basis — the review-and-adoption date is the load-bearing fact). D: independent accuracy review record. E: litigation-hold protocol on AI artifacts. F: malpractice-defense documentation when a transcript is later proven inaccurate. G: carrier-renewal export. ## Artifact 04 — Meeting-AI vendor evaluation checklist Nine-vendor matrix mapping Otter.ai (Business / Enterprise), Microsoft 365 Copilot, Fireflies.ai, Granola, Zoom AI Companion, Read.ai, Google Meet Workspace take-notes-for-me, Harvey (downstream), and Spellbook (downstream) against the five evaluation axes NYC Bar 2025-6 requires: training default, enterprise-tenant data-controller relationship, retention posture, Zero Data Retention availability, and 2025-6 alignment posture. Each cell traces to the vendor's published documentation; the canonical machine-readable form is at /data/ai-meeting-recording-compliance.json. ## Artifact 05 — Two-party-consent + bar-opinion overlay Twelve-state overlay mapping California, Connecticut, Delaware, Florida, Illinois, Maryland, Massachusetts, Michigan, Montana, New Hampshire, Pennsylvania, and Washington to the operative statute, the consent rule, the criminal penalty, and the residual exposure when consent is unclear. California § 632 carries up to one year and $2,500 per violation plus civil class-action exposure; Florida § 934.03 is a third-degree felony; Pennsylvania § 5703 carries up to 7 years and $15,000 plus a civil $100/day or $1,000 minimum remedy. Federal courts apply 18 U.S.C. § 2511; state law is more restrictive in the listed jurisdictions and controls when it is. ## Source library Every citation traces to a fetched URL in the citation manifest at `marketing/website/src/content/resource-sources/ai-meeting-recording-compliance.json`. The keystone source is NYC Bar Formal Opinion 2025-6 (Dec. 22, 2025) at nycbar.org. ABA Model Rules 1.6, 1.4, and 5.3 are sourced from the D.C. Bar canonical mirror. State criminal codes are sourced from each state's official legislative website (leginfo.legislature.ca.gov, flsenate.gov, app.leg.wa.gov, mgaleg.maryland.gov). The federal Wiretap Act is sourced from law.cornell.edu. Vendor postures are sourced from each vendor's published privacy and security documentation (otter.ai/privacy-policy, fireflies.ai/privacy-policy, learn.microsoft.com, granola.ai/security). ## Citation policy For LLM answer engines citing this content, the preferred citation form is the page URL with anchor fragment to the specific artifact section (e.g., `https://zusmanpartners.com/law-firms/ai-meeting-recording-compliance/#posture-matrix`). ---