Which Law Firms Use AI? Case Studies from BigLaw to Solo Practices

Law firms using AI now span every size of practice, from AmLaw 10 firms running firmwide deployments to solo practitioners taking on more matters without hiring. AmLaw rankings order US firms by gross revenue, and the examples below run from solo and boutique practices up to the AmLaw 10.

This guide profiles ten named firms that have publicly documented which platform they deployed and what changed afterward. The sources are firm announcements, vendor case studies and firm-reported figures, and nothing else. Readers comparing options for practices of their own size can measure these accounts against their own caseload.

A note on scope: AI use varies across firms, practice areas and jurisdictions, and public disclosures rarely describe a firm's full technology strategy. Every figure here is attributed to the party that published it.

Key takeaways

1. The law firms using AI profiled here range from solo practices through the AmLaw 10, and every example is documented publicly rather than inferred.
2. The platforms cluster by job: contract drafting and review (Spellbook), legal research and document analysis (CoCounsel, Lexis+ AI, Kira), and firmwide workflow automation (Harvey).
3. Reported outcomes concentrate on capacity rather than headcount reduction. Alturas Capital Partners, for example, shortened commercial lease negotiations from weeks to days and reduced outside counsel spend.
4. Adoption surveys disagree sharply on how many firms have adopted AI, and the gap between adoption and measurable revenue gain is wider than the adoption numbers alone suggest.

[cta-1]

At a glance: the firms in this guide

Ten law firms using AI are profiled below, ordered from the smallest practice to the largest, with the platform each firm deployed and the outcome it reported. Contract work and legal research account for most of the deployments, and a handful of firms built something of their own on top of a vendor platform instead of adopting it as shipped.

How many law firms use AI? The adoption data

How many law firms use AI depends heavily on who was asked, and the major 2026 surveys do not agree. Thomson Reuters reports that 41 percent of law firms say their legal teams are using generative AI, up from 28 percent the prior year, while Clio's 2026 reports put adoption far higher across every size band. The two sets of figures measure different things: whether a firm has any AI use at all, versus whether a firm has adopted it as a system.

1. Thomson Reuters: 41 percent of law firms and 47 percent of corporate legal departments report their legal teams using generative AI
2. Clio's 2026 Legal Trends Report for Mid-Sized Law Firms: 86 percent adoption among mid-sized firms, with 60 percent reporting a formal AI policy
3. Clio's 2026 Legal Trends for Solo and Small Law Firms: 71 percent adoption among solo firms and 75 percent among small firms
4. Thomson Reuters on returns: 53 percent of respondents already seeing a return on AI investment, and 43 percent expecting hourly billing to decline within five years

Clio publishes both of those reports behind a request form rather than on an open page, so those three figures are attributed here without a link.

The adoption-to-revenue gap deserves particular care, because the conventional claim runs the wrong way. Clio titled its 2026 solo and small firm report around the finding that small law firms lag larger peers on using AI to grow revenue. Smaller firms are therefore not a safe bet for faster returns simply because they carry less overhead. They adopt readily, and converting that adoption into billable gain appears to be the harder step.

Cunningham Legal: solo practitioner competing with Spellbook

A lawyer at Cunningham Legal adopted Spellbook to accelerate contract drafting and now handles more client matters without adding staff. Solo practitioners compete against firms with deeper benches while running the practice itself, and before AI tooling the binding constraint was simply available hours.

Risk identification on the first review pass matters as much as the speed does, because an issue caught early is cheaper than the dispute it can become later. The lawyer reviews and finalizes that first pass before the document goes out, which matters more in a practice with no second reader behind it.

1. Reported saving: 15 to 20 minutes per clause, and 1 to 2 hours per day, per Spellbook's published customer figures
2. Constraint addressed: available attorney hours, which previously capped how many time-sensitive matters the practice could accept

Key takeaway

Cunningham Legal's gain came from available attorney hours, the constraint that binds hardest on a practice with no second reader.

Jason Wiener P.C.: social enterprise law firm scaling with Spellbook

Jason Wiener P.C. is a Colorado firm specializing in social enterprises and employee ownership, and it uses Spellbook for contract drafting and review across its transactional practice. The firm needed to operate as efficiently as larger competitors while holding its depth in narrow areas such as benefit corporation governance.

The Review and Draft features take the first pass, and a lawyer reviews and edits the output before it leaves the firm. A boutique can therefore absorb more volume without hiring into areas where its value is the depth of its judgment.

1. Document types: benefit corporation governance documents, impact investment agreements and stakeholder-focused contracts
2. Where it runs: inside Microsoft Word, on the documents the practice already works in

Key takeaway

Boutique and specialized practices can use AI to compete on turnaround while keeping the expertise that differentiates them.

[cta-2]

Alturas Capital Partners: investment firm legal team using Spellbook

The in-house legal team at Alturas Capital Partners shortened commercial lease negotiations from weeks to days with Spellbook and reduced its outside counsel spend by keeping complex drafting in house. In-house teams face a distinct constraint: portfolios grow, and hiring more lawyers or sending more work out are both expensive answers.

The Associate agent surfaces where a lease portfolio differs from a single agreement, reconciling many documents against each other, and the in-house lawyer reviews that reconciliation before it reaches the negotiation. This is the pattern that recurs wherever software supports an in-house legal team: the aim is holding complex work internal, not reducing the number of lawyers doing it.

1. Work covered: transaction documents, lease and investment clause customization, and redlines finalized during live negotiation
2. Multi-document step: identifying inconsistencies between individual leases and master agreements

Key takeaway

Alturas kept complex drafting in-house instead of expanding outside counsel spend, keeping a lean legal team able to absorb more deal volume.

Rupp Pfalzgraf: building caseload capacity with Lexis+ AI

Mid-sized firm Rupp Pfalzgraf reached an 86 percent usage rate among its attorneys after 18 months of day-to-day use of Lexis+ AI, according to the case study LexisNexis published on its own community site. That page is not reachable to automated link checks, so the figure is attributed here without a link.

The before-and-after comparison in that account is worth more than the usage rate, because a usage rate measures adoption while the comparison measures effect. Research time is also the easiest gain for a litigation practice to evidence, since the old number is usually already on a timesheet.

1. Reported comparison: a precedent surfaced in 30 minutes that a team of clerks and associates had not located in 100 hours
2. Secondary uses: answering questions during client calls, and working in less familiar practice areas

Key takeaway

The Rupp Pfalzgraf account gives one of the clearest before-and-after figures available, because research time was already tracked before AI and after it.

Gowling WLG: analyzing compliance guidelines with Kira

Gowling WLG uses Litera's Kira to read outside counsel guidelines in minutes rather than hours. Outside counsel guidelines, or OCGs, are the billing and conduct rules corporate clients impose on the firms they instruct, and they are long enough that close reading is often deferred.

Gowling WLG trained custom Kira models against its own risk criteria, then built a database that grew to more than 1,100 OCGs totaling roughly 40,000 pages. The effect is to turn a manual reading exercise into a reference a lawyer can query at the point the question actually arises, though a guideline the firm has not yet processed still needs a first manual read before it joins the database.

1. Feature used: the Quick Study feature, to train models against the firm's own risk criteria rather than generic ones
2. What the platform returns: the relevant passages extracted, with the compliance obligations flagged

Key takeaway

Compliance review suits a first AI use case, because the source documents are repetitive, high-volume and a known drain on partner time.

Macfarlanes: the client-facing Amplify platform

Macfarlanes launched Amplify, a generative AI platform built for its clients to use rather than for internal work alone.

A client using Amplify does not have to implement generative AI across its own organization to benefit from it, so a firm-built product can succeed in situations where a platform rollout would stall. The firm positions it as an extension of its existing client offerings rather than as an internal efficiency tool, though Amplify is built specifically around Macfarlanes' own client base and methodology, so a firm without an equivalent Lawtech team could not replicate it directly.

1. Built by: the firm's own Lawtech team, with Harvey supplying the underlying generative AI
2. Audience: client in-house teams, interrogating their own documents and contractual data

Key takeaway

Macfarlanes turned its Harvey deployment into a service offering, which moves the economics from cost saved to revenue earned.

Orrick, Herrington & Sutcliffe: early CoCounsel adopter

Orrick was a beta customer of CoCounsel before Thomson Reuters announced its acquisition of Casetext, the company that built CoCounsel, on June 27, 2023 for 650 million dollars; CoCounsel is now sold within the Thomson Reuters product line. The firm had also acted as counsel to Casetext since 2014, which gave it an unusually early view of the product.

Testing a product before general release means absorbing the rough edges in exchange for shaping what the tool becomes, which is a different calculation from buying a finished one. Firms weighing the research platforms against each other can compare CoCounsel and Harvey directly.

1. Timing: in production testing against live research and transactional work before general release
2. Public posture: AI governance thinking published through the firm's podcast and innovation programming

Key takeaway

Beta participation gave Orrick production experience with CoCounsel before general release.

A&O Shearman: the first BigLaw firmwide Harvey deployment

A&O Shearman, then Allen & Overy, became the first large firm to deploy Harvey firmwide, in a 2023 rollout the firm said would reach more than 3,500 lawyers across 43 offices. The Harvey deployment carried forward through the 2024 merger that produced the combined firm.

A&O Shearman pairs the Harvey rollout with an accountability framework in which lawyers review AI output rather than forwarding it. That is the governance half of a firmwide deployment, and it is the half that decides whether scale produces leverage or exposure.

1. Reported scale: a rollout the firm said would reach more than 3,500 lawyers across 43 offices
2. Reported use cases: contract analysis, multilingual drafting and regulatory horizon-scanning
3. Governance pairing: lawyers review output before it moves forward, a stated condition of the rollout

Key takeaway

A firmwide rollout depends on a review framework as well as the licenses, and A&O Shearman published both together.

Paul Weiss: custom Harvey workflows

Paul Weiss became the first firm to launch custom workflows on Harvey's Workflow Builder, having served as the core design partner that helped define and test the product's capabilities. The AmLaw 50 firm shaped the tool from its earliest stages.

Building Harvey workflows in house is faster for Paul Weiss than commissioning each one from the vendor, and the approach spreads across practice groups instead of stopping at the first pilot. It also keeps the firm's method inside the firm.

1. Role: core design partner, helping define and test Workflow Builder before its launch
2. What the workflows hold: the firm's own expertise, guardrails and preferred practices, built into Harvey

Key takeaway

Paul Weiss kept its own methodology as the differentiator by configuring Workflow Builder instead of adopting Harvey's defaults.

Latham & Watkins: a first-of-its-kind AI academy

Latham & Watkins launched AI Academy, which the firm describes as a first-of-its-kind program in the legal industry, to prepare its lawyers for the changes AI brings to the clients they advise. The program runs as multiple events and training tracks across the year, aimed at different groups.

A standing task force, rather than a one-off rollout, lets the curriculum get revised as the tools change, which separates the program from a vendor demonstration. Aiming the first event at junior associates also targets the cohort whose work the technology reaches first.

1. Opening event: a two-day in-person training for first through fourth-year associates
2. Built by: the firm's AI Task Force, its training committee and its technology department jointly

Key takeaway

Firms that pair licenses with a standing training program build capability across a cohort rather than relying on individual attorneys.

Which law firms use Harvey AI?

Three firms profiled above have publicly documented Harvey deployments. The three sit at different points on a scale running from consuming the platform to building on top of it, and each added something Harvey does not supply.

1. Deploy broadly, govern centrally: A&O Shearman paired breadth with a firmwide review framework, so the governance scaled with the licenses
2. Configure the platform: Paul Weiss built its own workflow library on Workflow Builder, keeping firm method in house instead of commissioning it
3. Resell the capability: Macfarlanes put Harvey underneath a product its clients use directly, which moves the deployment onto the revenue side
4. Shared pattern: all three built something of their own, governance, workflows or a client interface, on top of the licensed platform

Harvey supplies the model and workflow infrastructure; none of the three firms' governance framework, workflow library or client product transfers automatically to a firm that simply licenses the platform. Firms evaluating the platform alongside other options tend to reach the commercial question quickly, and what Harvey costs a firm is covered separately, since the vendor does not publish a public rate card and the figure moves with firm size and deployment scope.

Are there AI-native law firms?

AI-native law firms are practices built around AI from the start rather than firms that added it to an existing structure. The AI Firm Index, a directory launched by Lupl co-founder Matt Pollins that passed 40 listings within months of launch, tracks the category as it grows.

The index defines the category by design rather than by founding date: pricing, intake, delivery and team structure all built with AI execution as the foundation. Most listed firms publish a price list openly, which is itself a departure from how the profession has historically quoted work. The roster changes as new firms launch, and the live directory is the current record of who qualifies.

1. Eudia, Alaro, AgileCounsel, Crosby, ElevateNext and Farringdon are among the listed firms
2. Published pricing is common across the category, instead of quote-on-request
3. Being newly founded is not a requirement, and established firms have applied
4. Intake, delivery and team structure are designed around AI execution, not retrofitted to it

Key takeaway

AI-native firms show what pricing, intake and delivery look like when they are designed around AI rather than retrofitted.

How AI is changing law firm operations and business models

Across the ways firms approach AI, the deployments in this guide share more than their differences suggest. None treats the technology as autonomous or authoritative.

Lawyers decide, and the tooling assists

A&O Shearman's review framework and the checkpoints in the Paul Weiss workflows point the same direction. Firms prefer systems where inputs and outputs stay transparent and reviewable, and they avoid arrangements where a lawyer cannot see how an output was reached.

1. Human review sits on the output path rather than alongside it
2. Tools that fit inside an existing workflow require less change to adopt, which is part of why the deployments profiled here mostly run inside Word or a similar document surface

Capability shifts down the seniority ladder

The Rupp Pfalzgraf account describes a first-year clerk producing work comparable to a more experienced associate on a summary judgment motion involving 15,000 pages. That shift is why the billing conversation follows the technology conversation so closely.

1. Junior lawyers take on work that previously required more experience
2. Thomson Reuters found 43 percent of legal professionals expecting hourly billing to decline within five years

The delivery model follows the tooling

Macfarlanes packaged its deployment as a client product, and the AI-native firms started from published pricing. Both are signals that the change reaches the commercial model, not just the workflow.

1. Firms with an existing structure tend to retrofit the workflow first and the pricing later
2. Firms built around AI start from the pricing and work backward

Why firms choose Spellbook for contract work

Spellbook meets transactional lawyers in the document rather than asking them to work somewhere new. Drafting, review and redlining happen inside Word and Google Docs, and the review output arrives as native Track Changes attributed to the lawyer who accepted it.

The platform grounds every output in the firm's own playbooks and precedents alongside a live pool of market data spanning more than 80 countries and hundreds of industries, so a lawyer can point to where a suggestion came from rather than take it on faith. On the security questions firms raise first, Spellbook publishes SOC 2 Type II, HIPAA and GDPR compliance, and zero data retention agreements with OpenAI and Anthropic, meaning customer data sent to those providers is not persisted.

The platform is a commercial contract specialist tool rather than a legal research platform or a litigation system, and firms whose primary need is case law research are better served by a comparison of the research-focused options.

Frequently asked questions

How long does AI take to show a return in a law firm?

Return depends on rollout structure and firm size rather than on a fixed timeline, and larger firms running structured rollouts can take longer to reach full utilization. Individual productivity gains tend to appear well before firm-level return shows up, and the two are easy to conflate when one is visible and the other is not.

1. Earlier signal: individual productivity gains, visible within the first weeks of use
2. Later signal: firm-level return, which depends on rollout structure and firm size

Does a firm need to be large to justify AI investment?

No. The case studies above include a solo practitioner and a boutique practice alongside AmLaw 10 firms. Clio's 2026 data does complicate the common claim that smaller firms see faster returns, since the solo and small firm report found those firms lagging larger peers on turning AI into revenue. Adoption is easy at small scale; conversion to billable gain is not automatic.

Which practice areas see the most benefit from AI?

The practice areas that benefit most from AI are the transactional ones covering contracts, mergers and acquisitions, and real estate, because drafting and review there are repetitive and high-volume. Litigation teams report more benefit from research and document analysis, as the Rupp Pfalzgraf account illustrates. In-house teams gain most on contract management and compliance work. Any practice with heavy document volume is a reasonable candidate, and the broader survey of where AI assists lawyers day to day maps the use cases by task.

How do law firms address confidentiality and privilege when using AI?

Firms address it through vendor selection and deployment terms rather than through tool settings alone. Enterprise legal AI platforms carry contractual confidentiality protections and defined data-retention terms, which general-purpose public AI services typically do not. The questions worth asking a vendor are whether and how data is retained, where it is hosted, whether the vendor or its model providers train on it, which subprocessors touch it, and what the data processing agreement commits to. Firms must confirm their own professional obligations in the relevant jurisdiction before routing any client material through any platform.

How can a firm convince partners to adopt AI?

Convincing partners to adopt AI tends to depend on evidence from peer firms rather than on capability demonstrations, which is part of why firms publish accounts like the ones above. Many firms start with a narrow pilot in one practice group to produce internal evidence before a wider rollout, which keeps the decision reversible and the scope small enough to measure. The commercial argument also carries weight, since clients increasingly expect the turnaround that peer firms have already demonstrated.

Firms comparing how practices at their own scale have approached contract work can start with Spellbook for law firms.