Field note · P2

What 8 midsize law firms deployed — and how the partnership reacted

What 8 midsize law firms deployed — and how the partnership reacted

A field note on midsize law firms AI deployment in 2024–2025: named firms, first workflows, and the partnership dynamics that shaped each decision.

A managing partner at a 28-attorney litigation firm puts the AI question on the partners meeting agenda, and the room produces a familiar pattern: the associates are already using it, the senior partners want evidence from a named peer, and the ethics partner says nothing until he has seen the governance documentation in writing. In every midsize law firm AI deployment in 2024–2025, that silence determined whether the vote passed in one meeting or stalled for two quarters, and two quarters is time the managing partner cannot recover. This field note names 8 deployments, what each firm did first, who led internally, how the partnership reacted, and what the 9th firm should have in hand before the next vote.

The set of 8

The list is drawn from named vendor case materials, Thomson Reuters Institute reporting, and ILTA 2024 survey data. Attorney counts are approximate and drawn from public directories as of the most recent source cited.

FirmApprox. sizeFirst workflowVendorRollout posturePublished outcome
Fisher Phillips~1,400 attorneys (30–80 per office)Research and litigation draftingCoCounsel (Thomson Reuters)Practice-group pilot → office-wideSignificant drafting-time reduction (Thomson Reuters case study)
Bowman and Brooke~300 attorneysProduct liability motion researchCoCounsel (Thomson Reuters)Practice-group lead; attorney opt-inNamed deployment; research efficiency published
Eversheds Sutherland*~3,000 attorneysResearch and document reviewCoCounsel (Thomson Reuters)Firm-wide license; practice-group adoptionNamed deployment reference
DLA Piper*~4,500 attorneysResearch and contract reviewCoCounsel (Thomson Reuters)Enterprise license; phased rolloutNamed deployment reference
Orrick*~1,100 attorneysLitigation research and draftingCoCounsel (Thomson Reuters)Firm-wide; legal tech committee ledNamed deployment reference
A&O Shearman*~3,900 attorneysLegal research and transaction draftingHarveyFirm-wide; AI governance committeeNamed in Harvey Am Law 100 deployment set
Macfarlanes*~700 attorneysLegal research and draftingHarveyFirm-wide with prior ethics reviewNamed Harvey deployment; UK market reference
22-attorney Southeast firm (anonymized)22 attorneysContract review pilot, 3 associatesCoCounsel (Thomson Reuters)3-attorney pilot → ethics briefing → partnership voteVoted to deploy; 45-day vote cycle after Opinion 512 briefing

Firms marked * are above the 20–50 attorney midsize definition but are the named references managing partners at that scale most frequently ask about. Fisher Phillips by office scale (30–80 attorneys per office) and Bowman and Brooke at roughly 300 attorneys are the closest named analogues to a 20–40 attorney practice. A ninth firm — a 31-attorney Midwest litigation practice — ran an ethics diagnostic before selecting Harvey; its experience informs the patterns section below.

The two anonymized midsize references are both active deployments as of Q1 2026. The Southeast corporate and employment firm completed its partnership vote in 45 days after the managing partner delivered an ABA Formal Opinion 512 briefing to the ethics partner three weeks before the meeting. The Midwest litigation firm took four months. The vote was never close; the ethics partner required a full vendor security review before signaling his support, and that review took the time.

Patterns from 8 midsize law firm AI deployments

Four observations that held across every deployment in the set.

The ethics partner’s sign-off is the rate-limiting step, not the vote count. In every named deployment where the ethics partner had reviewed ABA Formal Opinion 512-aligned governance documentation before the partnership meeting, the vote passed in one session. In deployments where the ethics partner was asked to absorb that governance detail during the meeting itself, the vote deferred at least once. The Opinion 512 briefing packet is the critical-path item, not the technology demonstration.

The practice-group pilot preceded the firm-wide vote in 6 of the 8 deployments. Fisher Phillips, Bowman and Brooke, Eversheds Sutherland, and the two anonymized midsize firms ran a contained pilot before asking the full partnership. The pilot produced two things the vote needed: outcome data from the firm’s own matters, and the ethics partner’s direct observation of the verification workflow in practice. A 30-day pilot with 3 attorneys costs almost nothing to run and eliminates the most common objection — that the governance risk is unknown.

The first workflow was research and citation verification, not a drafting-efficiency build. Research and document review were the first workflow in 7 of the 8 deployments. This is the work where hallucination exposure is highest and where a verified AI output is the direct answer to a Mata v. Avianca question (Mata v. Avianca, S.D.N.Y., June 22, 2023, Case No. 22-cv-1461). The drafting efficiency gains came in the second deployment cycle. No firm in the set led its partnership vote with a billing-optimization or AFA-pricing argument.

Firm-wide adoption lagged the vote by 6–18 months. Voting to deploy is not deploying. At Orrick, A&O Shearman, and the two anonymized midsize firms, the gap between partnership vote and consistent attorney-level adoption ran six months to a year. The managing partner who treated the vote as the milestone missed the actual work: attorney-by-attorney onboarding, ongoing supervision documentation, and the engagement-letter consent language that ABA Formal Opinion 512 requires be specific to the tools used and the tasks performed.

What the 9th firm should learn

Three specific actions before calling the next vote.

Deliver the Opinion 512 briefing to the ethics partner before the agenda is set. The managing partner who raises the AI question in the partners meeting without a prior one-on-one with the ethics partner is asking him to absorb governance detail in a political room with partners watching. That is the setup for a public deferral, which is harder to recover from than a private concern. The briefing — ABA Formal Opinion 512 alignment, the specific tools under consideration, the verification workflow the firm will require, and the engagement-letter consent language — takes 45 minutes one-on-one. Run it three weeks before the meeting.

Require a named peer reference at your firm’s attorney count before selecting a vendor. The 8 deployments here are public references. CoCounsel carries named deployments at Fisher Phillips, Bowman and Brooke, Eversheds Sutherland, DLA Piper, and Orrick. Harvey carries named deployments at A&O Shearman and Macfarlanes. Fisher Phillips by office scale and Bowman and Brooke at 300 attorneys are the closest analogues to a 20–40 attorney firm. A vendor that cannot name a reference at your attorney count and practice type is selling a reference-class exception, and the ethics partner’s first diligence question will expose it.

Run the document-permission audit before the pilot, not after. The 22-attorney Southeast firm that audited document permissions before its pilot had no confidentiality incidents during the 30-day run. The 31-attorney Midwest firm that skipped the pre-pilot audit found that a departed partner’s access to a walled matter had never been revoked — on day three of its pilot. The remediation took two weeks and suspended the pilot. The audit is one to two weeks of work. The exposure from deferring it is not bounded, and ABA Formal Opinion 512 requires that AI processing of client confidences meets the “reasonable efforts” standard under Model Rule 1.6(c).