What is the highest-ROI AI use case for a small criminal defense firm?

The biggest, safest win is AI-assisted discovery synthesis + body-cam / dash-cam timeline review paired with AI-drafted suppression motion and sentencing memo frameworks. A 2026 criminal-defense firm running Clio Manage + Duo, MyCase IQ, Smokeball AI, PracticePanther, Rocket Matter, or Filevine Core with a research platform (Westlaw Precision AI, Lexis+ AI, CoCounsel, Harvey, Paxton, Fastcase, vLex Vincent) and an e-discovery / multimedia review tool (Relativity aiR, Everlaw, DISCO Cecilia, Logikcull, Axon Evidence + Axon Auto-Transcribe, Veritone Legal, Truleo) typically saves 4-8 hours of attorney review per serious felony matter, shortens motion-drafting cycle time 40-60%, and cuts sentencing-memo prep time by half. Every motion, sentencing memo, and plea analysis still requires the responsible attorney to review, citation-verify, and sign — Brady / Giglio obligations and FRE 901 authentication are non-delegable to AI.

How does ABA Formal Opinion 512 + citation-verification precedent apply to criminal defense?

ABA Formal Opinion 512 (July 2024) plus state bar opinions (CA, NY, FL, DC, VA, NC, WA, TX, PA) converge on five duties for criminal defense: (1) Competence (Rule 1.1) — understand hallucinated case law, fabricated Supreme Court + Circuit decisions, invented state-court unpublished opinions; (2) Confidentiality (Rule 1.6) — attorney-client privilege is absolute, no client statements, plea discussions, or defense strategy into non-BAA / non-DPA endpoints; (3) Communication (Rule 1.4) — inform clients when AI is used substantively; (4) Reasonable fees (Rule 1.5) — don't bill hours for minutes of AI work, especially in appointed / court-appointed matters; (5) Candor + Supervision (Rules 3.3, 5.3) — verify every citation before filing. Mata v. Avianca (SDNY 2023), Park v. Kim (2d Cir. 2024), and Morgan & Morgan sanctions (N.D. Ala. 2025) set the floor: every case, 4th / 5th / 6th / 8th / 14th Amendment precedent, state-court suppression opinion, and federal sentencing guideline citation gets verified in Westlaw / Lexis / Fastcase / vLex before the motion or memo is filed. Courts have sanctioned criminal defense attorneys for hallucinated Strickland citations and fabricated suppression precedent.

What confidentiality + Brady / Giglio rules apply to AI-assisted discovery review?

Criminal discovery materials — body-cam, dash-cam, jail calls, grand-jury transcripts, lab reports, CI / UC files, SDT returns — carry a mix of privilege, work-product, sealed-order, and protective-order obligations. AI review must: (1) run inside a controlled endpoint (BAA / DPA / on-prem) that preserves attorney-client privilege and work-product protection; (2) honor protective orders and sealed-material handling under Fed. R. Crim. P. 16 + state equivalents; (3) flag Brady / Giglio impeachment material for attorney escalation (prior statements, identification procedures, incentives to CIs, officer disciplinary history, lab misconduct); (4) preserve chain-of-custody metadata on audio / video / digital evidence for FRE 901 / state authentication; (5) never log or train on client PII / co-defendant / victim / CI identifiers. Brady / Giglio obligations flow through the prosecutor, but defense counsel's duty to investigate and preserve is non-delegable. Strickland v. Washington + Rompilla / Wiggins line require diligent investigation; AI helps with volume, not judgment.

What AI stack actually works in a 1-10 attorney criminal defense firm in 2026?

Core practice management: Clio Manage + Duo, MyCase + IQ, Smokeball + AI, Rocket Matter, PracticePanther, CosmoLex, Filevine Core. E-discovery + multimedia review: Relativity aiR, Everlaw AI, DISCO Cecilia, Logikcull, Reveal, Ipro, Axon Evidence + Axon Auto-Transcribe + Axon Redaction Assistant, Veritone Legal, Truleo, Nice Inform, Verbit. Legal research (with citation verification): Westlaw Precision AI, Lexis+ AI, CoCounsel (Thomson Reuters), Harvey, Paxton, Bloomberg Law AI, Fastcase, vLex Vincent, Casetext, Loislaw. Motion drafting: Lawyaw, Documate, HotDocs, Gavel, Smokeball Automation. Sentencing analytics: Sentencing Counsel, JUSTIS AI, Lex Machina Criminal, LexisNexis Sentencing Research, USSC HelixAI, Federal Judicial Center data. Forensic + lab review: FTK, Cellebrite Inspector + Pathfinder, Magnet AXIOM, GrayKey, MSAB XRY. Jail-call + recorded-interview review: Nice Inform, Veritone Redact, Truleo, Reveal, Prism Litigation. Expert + investigator management: Casefleet, Filevine Tasks, AskEllyn. Transcription: DepoIQ, Rev, Trint, Descript, Otter (with PII screen). Marketing + intake: CallRail Conversation Intelligence, Phonewagon, Scorpion, Justia, Mockingbird, Lead Docket, Captorra, Intaker. Review response: BirdEye, Podium, Swell, Avvo, Martindale.

What are the fastest AI wins for a criminal defense managing partner this quarter?

Three, in order of speed-to-cash. (1) AI-assisted intake + custody / bail triage + conflicts — book consults faster, triage felony / misdemeanor / federal / state / juvenile / immigration-consequence matters, screen for conflicts and prior representation. (2) AI discovery synthesis + body-cam / dash-cam / jail-call timeline review across large multimedia productions with Brady / Giglio flagging. (3) AI-drafted suppression motion framework + sentencing memo framework with citation verification. Combined, these typically lift cases-resolved-per-attorney-per-month 20-35%, improve motion-win rate on well-prepared suppression issues, and cut sentencing-memo prep time 50% within 60-90 days.

How to Use AI for a Criminal Defense Law Firm in 2026: Intake, Discovery, Motions, Sentencing & Owner Scorecard

Published May 11, 2026 · 16 min read · Happycapy Guide

TL;DR — for the managing partner

The three highest-ROI AI wins in a 2026 criminal defense firm are discovery + body-cam synthesis, suppression + motion-in-limine drafting, and sentencing memo drafting. Together they shorten felony cycle time without sacrificing Brady diligence, FRE 901 authentication, or Strickland investigation.
AI drafts the discovery timeline, the motion framework, the sentencing memo outline, the plea-bargain analysis, and the client-status letter. The responsible attorney reviews, corrects, citation-verifies, and signs. Juries + judges trust licensed lawyers, not software.
Citation verification is non-negotiable. Mata v. Avianca + Park v. Kim + Morgan & Morgan sanctions set the floor — every Supreme Court, Circuit, state-court, and USSG citation is checked in Westlaw / Lexis / Fastcase / vLex before filing.
Attorney-client privilege and work-product protection are absolute. Client statements, plea discussions, defense strategy, expert work product, and protective-order material never leave a BAA / DPA / on-prem endpoint.
Owner rule: every AI-drafted motion, memo, brief, letter, ad, SMS, or review reply is reviewed and signed by the responsible attorney (or delegated supervising attorney) before it leaves the firm.

Why criminal defense is a high-leverage AI vertical

Criminal defense is volume-heavy, multimedia-heavy, deadline-driven, and liberty-altering. The managing partner's four chronic problems — intake + custody triage, discovery review burden (body-cam, dash-cam, jail-calls, cell-phone extractions, SDT returns, lab reports), motion and sentencing-memo bottleneck, and client-communication overhead — all get better with narrow AI assistants running inside a modern criminal-defense stack. AI does not replace the attorney's state-bar-licensed judgment, the jury, or the sentencing judge; it removes the documentation tax that eats the billable day and the CJA voucher.

This playbook is for the managing partner of a 1-to-10 attorney criminal-defense firm (state felony, state misdemeanor, federal, post-conviction, juvenile, or mixed) who wants to use AI across intake, discovery, motions, plea analysis, sentencing, client comms, marketing, and owner scorecard — without tripping ABA Model Rules 1.1, 1.4, 1.5, 1.6, 3.3, 5.3, 7.1, 7.3; Brady / Giglio / Strickland; Mata / Park / Morgan & Morgan-style citation rules; FRE 901 / 902 authentication; state-bar AI guidance; or TCPA.

The compliance floor (read this first)

ABA Model Rules 1.1 (competence), 1.4 (communication), 1.5 (fees), 1.6 (confidentiality), 3.3 (candor), 5.3 (non-lawyer assistants — includes AI), 7.1 (truthful ads), 7.3 (solicitation): the ethical spine.
ABA Formal Opinion 512 (July 2024): generative AI + competence + confidentiality + candor + supervision + fees.
State bar AI opinions: CA Practical Guidance Nov 2023; NY Task Force April 2024; FL Opinion 24-1 Jan 2024; DC Opinion 388 April 2024; VA, NC, WA, TX, PA.
Citation-verification precedent: Mata v. Avianca (SDNY 2023); Park v. Kim (2d Cir. 2024); Morgan & Morgan sanctions (N.D. Ala. 2025). Sanctions imposed in criminal context for hallucinated Strickland, suppression, and sentencing citations.
Brady v. Maryland + Giglio v. United States + Kyles v. Whitley + Napue v. Illinois: prosecution disclosure + defense investigation + impeachment-material preservation.
Strickland v. Washington + Rompilla v. Beard + Wiggins v. Smith: ineffective-assistance standard — investigate, mitigate, preserve the record.
Fed. R. Crim. P. 16 + state equivalents: discovery production + protective-order compliance.
FRE 901 + 902 + state equivalents: authentication of AI-processed audio, video, transcripts, and extractions; chain of custody.
Federal Sentencing Guidelines + 18 U.S.C. §3553(a) + state sentencing statutes: every AI-surfaced guideline range, departure, or variance argument gets independently verified.
CJA Guidelines (18 U.S.C. §3006A) + state-appointed-counsel voucher rules: no billing for AI-saved time; accurate time-task reporting.
Attorney-client privilege + work-product protection: BAA / DPA / on-prem endpoints only for privileged material.
Protective orders + sealed materials: no CI, UC, minor-victim, or sealed-grand-jury material out of controlled endpoints.
Padilla v. Kentucky: immigration-consequence advisement for non-citizen clients — don't let AI gloss this.
TCPA + state mini-TCPA: written express consent for marketing SMS, quiet hours 8am-9pm local, opt-out language.
Two-party-consent recording: CA, FL, MA, WA, PA, IL, MT, NH, CT, MD — announce on every call.
Jail-call recording rules: inmate phone-call systems (Securus, ViaPath / GTL, ICS) record by default — privileged-channel carve-outs vary by facility.
State privacy: CCPA / CPRA, SHIELD, Colorado CPA, Virginia CDPA — client data + intake-form handling.

The criminal defense AI stack in 2026

Practice management: Clio Manage + Duo, MyCase + IQ, Smokeball + AI, Rocket Matter, PracticePanther, CosmoLex, Filevine Core.
E-discovery + multimedia review: Relativity aiR, Everlaw AI, DISCO Cecilia, Logikcull, Reveal, Ipro, Axon Evidence + Axon Auto-Transcribe + Axon Redaction Assistant, Veritone Legal, Truleo, Nice Inform, Prism Litigation, Verbit.
Legal research (with citation verification): Westlaw Precision AI, Lexis+ AI, CoCounsel (Thomson Reuters), Harvey, Paxton, Bloomberg Law AI, Fastcase, vLex Vincent, Casetext, Loislaw.
Motion + document automation: Lawyaw, Documate, HotDocs, Gavel, Smokeball Automation.
Sentencing analytics: Sentencing Counsel, JUSTIS AI, Lex Machina Criminal, LexisNexis Sentencing Research, USSC HelixAI + Federal Judicial Center data, state-DOC recidivism data.
Forensic review: FTK, Cellebrite Inspector + Pathfinder, Magnet AXIOM, GrayKey, MSAB XRY, Oxygen Forensics, Amped FIVE / Authenticate.
Jail-call + recorded-interview review: Nice Inform, Veritone Redact, Truleo, Prism Litigation, Reveal.
Investigator + expert management: Casefleet, Filevine Tasks, AskEllyn, Prepare The Case.
Transcription: DepoIQ, Rev, Trint, Descript, Otter (with PII screen).
Marketing + intake: CallRail CI, CallTrackingMetrics, Phonewagon, Scorpion, Justia, Mockingbird, Lead Docket, Captorra, Intaker.
Reviews + reputation: BirdEye, Podium, Swell, Avvo, Martindale, Super Lawyers.

10 copy-paste prompts for a 2026 criminal defense firm

Run these inside your firm's case-management + research + e-discovery platforms of record so privileged material stays inside controlled endpoints. Replace bracketed placeholders with real values. Every AI output gets a human reviewer — the responsible attorney, paralegal, investigator, or managing partner — before it leaves the firm. Every case citation is verified in Westlaw / Lexis / Fastcase / vLex / Casetext.

1. Intake + custody / bail triage with conflicts screen

You are our criminal-defense intake synthesis assistant. From this prospective client's intake form + consultation notes [PC name, charges, charging document, arrest date, custody status, bail amount / set, prior record, citizenship status, employment, family], produce the intake packet. Output: - Matter type (state felony / misdemeanor / federal / juvenile / post-conviction / appeal) - Charges + charging document (complaint / information / indictment) - Custody status + bail posture + next-appearance deadline - Prior record summary (priors, probation / parole status, pending matters) - Red-flag screen: * Immigration consequences (Padilla advisement flag) * Sex-offender registration consequences * Firearms / Lautenberg consequences * Professional-license / security-clearance consequences * Juvenile / adult collateral consequences * Cooperation / plea leverage posture - Conflicts (co-defendants, witnesses, prior representation) - Evidence posture (body-cam, dash-cam, jail-call volume estimate, SDT returns, lab reports pending) - Next-step list for attorney: deeper consult vs immediate engagement vs referral - Retainer / CJA / appointed-counsel flag Never render legal advice in the intake summary — queue attorney review. Route through our controlled endpoint; never send PC statements or defense strategy to non-DPA endpoints.

2. Discovery + body-cam / dash-cam timeline synthesis

You are our discovery synthesis assistant. Given [discovery production index + police reports + body-cam / dash-cam files + 911 calls + CAD notes + witness statements + jail calls + SDT returns + lab reports], produce the synthesis. Output: - Chronological timeline of events (date / time / location / actor / action / source document with bates number + timecode) - Officer list + badge numbers + unit assignments - Witness list + summary of statements + inconsistencies - Physical evidence list + chain-of-custody summary - Body-cam / dash-cam index with timecoded observations: * Stop / seizure moments (Terry / 4th Amendment triggers) * Miranda / waiver moments (5th / 6th Amendment triggers) * Use-of-force moments * Statements by client (voluntariness / custody analysis) * Statements by officers (command presence / representations) * Identification procedures (show-ups, photo arrays, line-ups) - Brady / Giglio flags for attorney escalation: * Officer disciplinary history references * Lab misconduct / protocol-deviation references * CI / UC incentives * Exculpatory statements * Prior inconsistent statements * Identification-procedure issues - Suppression-issue matrix (4th / 5th / 6th Amendment per incident) - Missing-discovery list for supplemental request - Chain-of-custody authentication checklist per FRE 901 / 902 Attorney reviews every timeline entry, every Brady flag, and every authentication question. AI does not render legal opinions on suppression merits; attorney analyzes.

3. Suppression motion + motion-in-limine framework

You are our suppression / motion-in-limine drafting assistant. Given [issue: 4th Amendment stop / frisk / search / seizure; 5th Amendment Miranda / voluntariness; 6th Amendment right-to-counsel / Massiah; identification procedure; expert Daubert / Frye; prior-bad-acts 404(b); hearsay; authentication], produce the motion framework. Output: - Caption + case number + court + judge - Statement of facts (citing bates + body-cam timecodes) - Issue statement - Legal standard (with verified citations) - Argument sections: * Element-by-element analysis * Relevant binding authority (Supreme Court, Circuit / state-supreme, verified in Westlaw / Lexis) * Persuasive authority * Distinguishing adverse authority * Application to our facts - Requested relief - Certificate of service - Proposed order - Evidentiary appendix (exhibits list with bates + timecodes) Rules: - Every case citation verified in Westlaw / Lexis / Fastcase / vLex. Never cite a case the AI surfaced without independent verification. Mata / Park / Morgan & Morgan sanctions are the floor. - No fabricated quotations, no invented page numbers. - Preserve chain-of-custody + authentication framework per FRE 901 / 902. Attorney reviews, revises, citation-verifies, and signs.

4. Jail-call + recorded-interview review

You are our jail-call + recorded-interview review assistant. Given [Securus / ViaPath / GTL / ICS jail-call production OR interrogation recording OR CI wire], produce the review. Output: - Per-recording metadata (date / time / caller / callee / duration / channel) - Privileged-call flag (any call to defense counsel / investigator — extract, suppress from non-privileged review workflow, and notify attorney for facility carve-out review) - Substantive summary with timecoded highlights - Statements by client relevant to charges (context + prior / subsequent statements for impeachment posture) - Statements by third parties (co-defendants, witnesses, family) - Indicators of coercion, promise, threat, or improper inducement (for Miranda / voluntariness attack) - Code-language / jargon flags for investigator / expert review - Identification / scheme / conspiracy indicators - Translation flags (non-English segments — queue certified interpreter) - Attorney-action list: suppression theory, impeachment use, cross-examination outline, defense strategy implications Rules: - Privileged calls get walled off and never enter the general review pool. - Redact victim / minor / CI / UC identifiers per protective order. - Controlled endpoint only; never send jail-call audio to non-DPA endpoints.

5. Plea-bargain analysis + Padilla advisement

You are our plea-bargain analysis assistant. Given [charges, plea offer, client's citizenship status, immigration history, prior record, employment / licensure / firearms / sex-offender / professional-license status, restitution exposure, cooperation posture], produce the analysis. Output: - Charge-by-charge comparison (trial exposure vs plea offer) with USSG range or state-sentencing-guideline range (verified) - Collateral-consequence matrix: * Immigration (Padilla analysis — is the plea "clearly" a removable offense or not; aggravated-felony / CIMT / drug / firearm / DV analysis) * Sex-offender registration (tier + duration + jurisdiction) * Firearms prohibition (Lautenberg / state) * Professional-license impact * Security-clearance impact * Voting rights * Housing / employment / benefits - Cooperation value assessment (if applicable) - Restitution + forfeiture exposure - Trial-risk assessment (not a prediction; a structured upside / downside) - Client-advisement outline (plain-language, 7th-grade reading level) Rules: - AI does NOT advise the client. Attorney advises after reviewing this analysis with the client in a privileged meeting. - Padilla / state-immigration-consequence review is non-delegable — attorney confirms with immigration counsel if needed. - Every guideline citation verified.

6. Sentencing memo + mitigation narrative framework

You are our sentencing memo drafting assistant. Given [guilty plea / verdict, PSR / PSI draft, client background (family, education, employment, trauma, health, substance-use, service, community), victim-impact materials, offense conduct, acceptance-of-responsibility, cooperation if applicable], produce the sentencing memo framework. Output: - Caption + case number + sentencing judge - Introduction (theory of mitigation) - Objections to PSR / PSI (factual + guideline-computation) - 18 U.S.C. §3553(a) / state-sentencing-factors analysis: * Nature + circumstances of offense * History + characteristics of defendant * Seriousness + respect for law + just punishment * Deterrence (general + specific) * Protection of public * Rehabilitation (education / vocational / medical / correctional treatment) * Avoiding unwarranted sentencing disparities (cite USSC Sourcebook / Sentencing Commission data if used; verified) - Mitigation narrative (client's story — attorney-elicited, not fabricated) - Sentencing recommendation with legal + empirical support - Letter-of-support index - Proposed conditions of supervision (if applicable) Rules: - Every case + guideline citation verified. - Mitigation narrative is elicited from client + family + records, not AI-fabricated. - Use Sentencing Counsel / JUSTIS AI / USSC HelixAI for comparable-case data with source links; attorney confirms comparisons.

7. Expert + investigator management + Daubert prep

You are our expert / investigator management assistant. Given [charges, defense theory, discovery, budget / CJA voucher posture, prosecution's expert disclosures], produce the plan. Output: - Expert-needs matrix by issue: * Forensic science (DNA, firearms / toolmarks, digital forensics, fingerprint, latent, toxicology, serology, drug chemistry, accident reconstruction, blood-spatter, pathology, psychology / competency, mitigation specialist, immigration consequences, eyewitness ID) - Candidate experts with credentials + prior testimony + bar-admitted states - Daubert / Frye / state-equivalent readiness per expert + methodology - Retention-letter draft + fee-agreement draft - Investigator tasking (witness interviews, scene reconstruction, records subpoenas, open-source intelligence) - Work-product protection checklist (everything walled under the attorney's privilege/work-product framework) - CJA budget estimate + justification for ex-parte motion if federal - Production schedule for reports, declarations, and cross-exhibits Rules: - No expert work product to non-DPA endpoints. - Confirm Daubert / Frye methodology admissibility before retention.

8. Cross-examination outline + impeachment matrix

You are our cross-examination outline assistant. Given [witness type: officer, civilian, expert, co-defendant, CI; prior statements: body-cam, reports, grand jury, prior hearings, prior trials; discovery + exhibits], produce the outline. Output: - Witness profile (role, credentials, biases, incentives) - Chapters of cross (controlled leading questions, one fact per question): * Prior-inconsistent-statement chapter (bates + page + line) * Bias / motive / interest chapter (CI payments, plea deals, 5K motions, Rule 35 posture) * Memory + perception chapter * Training / protocol chapter (for officers + experts) * Chain-of-custody / authentication chapter (for evidence custodians) - Impeachment matrix (prior convictions, prior bad acts, FRE 608(b) / 609, state equivalents) - Exhibit list with bates / timecodes / pre-marked IDs - Objections the prosecution is likely to make + our response - Direct-to-cross transition plan Rules: - Leading questions only. - No open-ended "why" questions. - Every impeachment reference verified against the source.

9. Client status letter + family update

You are our client-status-letter assistant. Given [client case, last update, next court date, outstanding items, custody status, retainer status], draft a plain-language status letter in English plus [target language if client prefers — Spanish / Mandarin / Vietnamese / Haitian Creole / Arabic / Russian / etc.]. Rules: - Plain-language, 7th-grade reading level - No legal advice in the letter — attorney advises separately in privileged channels - Custody-client letters: use facility-approved channels + legal-mail protocols; no privileged material on monitored channels - Translation drafts for internal review; certified translator for any filed or sealed translation Output: - English letter - Target-language letter (internal draft) - Family-contact summary (if client has authorized communication with family) - Deadline tracker summary for attorney sign-off Every letter reviewed and signed by the responsible attorney. Privileged channels confirmed for custody clients.

10. Managing-partner weekly scorecard + bar-compliant ads

You are our criminal-defense managing-partner scorecard + marketing assistant. Weekly scorecard from [case-management + CallRail + marketing + finance data]: - New matters opened by charge class (felony / misdemeanor / federal / juvenile / post-conviction) - Intake-to-retainer conversion rate (retained vs appointed / CJA) - Hearings + trials this week (arraignment / preliminary / motion / trial / sentencing / probation violation) - Motion filings + wins / losses / denials - Plea dispositions by class + pre-trial-services outcomes - Trial dispositions (guilty / NG / hung / mistrial) - Sentencing outcomes + guideline departures / variances - Post-conviction filings (§2255 / §2254 / state PCR) - Retainer + CJA voucher aging - Attorney utilization + billable hours + write-off - Review-response rate + rating trend - 3 drafted observations + 3 proposed actions for the Monday partner meeting Bar-compliant ad copy: - 3 variations for [platform: Google search / Meta / YouTube / local radio / community partnerships] - ABA Rule 7.1 truthful + not misleading; no guarantees of outcome; no prior success implying future success without disclaimer - State-bar disclaimer per state of practice - No testimonials implying specific results without disclaimers - No "results-or-no-fee" implications in jurisdictions that prohibit contingency in criminal matters (most) - No implications of government / prosecutor affiliation Managing partner signs off on every scorecard interpretation and every ad.

Common mistakes to avoid

Hallucinated citations. Every Supreme Court, Circuit, state-supreme, and USSG citation gets verified in Westlaw / Lexis / Fastcase / vLex before filing. Courts have sanctioned defense attorneys for invented Strickland, suppression, and sentencing precedent.
Sending client statements, plea discussions, or expert work product to non-DPA endpoints. Privilege + work-product breach. BAA / DPA / on-prem only.
Letting AI analyze Padilla / immigration consequences without independent verification. Non-citizen clients require attorney + immigration-counsel review.
Filing AI-surfaced body-cam summaries without watching the footage. AI summaries miss context, tone, and visual detail; attorney watches key segments.
Disclosing privileged jail calls to non-privileged review workflow. Facility carve-outs vary; confirm and wall off attorney calls.
Using AI-transcribed interrogation as the record without verifying against audio; errors become motion-in-limine / authentication issues.
Marketing copy with outcome guarantees or prior-success-based testimonials. ABA Rule 7.1 + state-bar restrictions.
Marketing SMS without TCPA written express consent. State mini-TCPA + quiet hours + multilingual opt-out.
Two-party-consent recording violations in CA, FL, MA, WA, PA, IL, MT, NH, CT, MD.
Billing hours the AI completed in minutes on retained + CJA matters — Rule 1.5 + CJA guideline violation.

60-day rollout plan

Weeks 1-2 (foundation): Map your matter lifecycle from arraignment to sentencing. Choose platforms of record (Clio / MyCase / Smokeball + Relativity / Everlaw / DISCO + Axon / Veritone), confirm BAA / DPA / on-prem endpoints, document citation-verification and privileged-call handling protocols, train staff on Brady / Giglio / Padilla + the AI-use framework. Audit current motion-win rate, plea-disposition time, and sentencing-memo cycle time baselines.

Weeks 3-4 (intake + discovery): Pilot AI-assisted intake + custody / bail triage + conflicts screen. Stand up AI discovery synthesis + body-cam / dash-cam timeline review with Brady / Giglio flagging. Measure: intake-to-retainer conversion, discovery-review hours per felony, Brady / Giglio flag volume.

Weeks 5-6 (motions + plea): Add AI suppression + motion-in-limine framework with citation-verification workflow. Add AI plea-bargain analysis + Padilla advisement template with immigration-counsel escalation. Measure: motion-drafting cycle time, motion-win rate on well-prepared issues, plea-analysis turnaround.

Weeks 7-8 (sentencing + scorecard): Pilot AI sentencing memo + mitigation narrative framework with Sentencing Counsel / USSC HelixAI comparables. Stand up AI cross-examination outline + impeachment matrix. Turn on the managing-partner weekly scorecard with motion-win rate, plea dispositions, sentencing outcomes, review-response rate, utilization, and AR aging. Lock in review cadence; publish a quarterly internal AI-ethics review and citation-audit log.

Ready to modernize your criminal defense practice?

Start with discovery synthesis + body-cam timeline review and a suppression motion framework this week. Citation-verify every authority and never route privileged material to a non-DPA endpoint. See more owner playbooks for law-firm AI.