The typical US in-house legal team in 2026 is small, about six people, and it is squeezed from both ends. Outside-counsel rates are climbing at roughly twice the rate of inflation, while generative AI adoption inside corporate law departments has doubled in a year to about half of teams.
The story that ties those two facts together is the outside-counsel paradox. In-house teams expect AI to cut their law-firm bill, the firms are the ones capturing the productivity gains so far, and the promised client savings have mostly not shown up. On top of that, the tools still get facts wrong often enough that trust, not adoption, is the gating problem.
This post pulls three detailed benchmarks into one picture: how big in-house teams are, what they spend, and how AI is changing the work. Every number here is third-party, named, and dated. Where a figure is a self-report or a projection rather than a measured result, we say so.
TL;DR
- The median in-house legal department is about 6 people, and the median company has about 7 lawyers (ACC, 2024 Law Department Management Benchmarking Report; median-6 from ACC/MLA 2022).
- Lawyer density falls fast as revenue grows: about 17 lawyers per $1B at companies under $1B, 4 at $1B to $5B, and 3 above $5B (ACC, 2021 Law Department Management Benchmarking Report).
- In-house GenAI use doubled, from 23% in 2024 to 52% in 2025 (ACC and Everlaw, Generative AI's Growing Strategic Value for Corporate Law Departments, October 2025).
- AI spend is small and poorly tracked: 26% of in-house teams spend under $100 a month, 9% spend over $2,000, and 44% do not know their spend (Counselwell and Spellbook, AI in Legal Departments: 2025 Benchmarking Report).
- Outside-counsel rates keep outrunning inflation: worked rates at large US firms rose 7.4% in 2025 against 2.8% inflation (Thomson Reuters Institute, Law Firm Rates Report 2026, October 2025).
- Accuracy is still the open problem: leading legal research AI tools hallucinated 17% to 33% of the time (Magesh et al., Journal of Empirical Legal Studies / Stanford RegLab, 2025).
What share of corporate law departments used generative AI in 2025?
How we built this
How big the in-house team actually is
Start with the headline most GCs want: what is normal. The median corporate legal department runs about 6 people counting all staff, and the median company has about 7 lawyers by a separate count that measures attorneys only rather than total headcount. The two figures come from different survey definitions, which is why they look close but not identical. Spread is enormous. A company under $1B in revenue tends to sit around 2 lawyers and 4 total staff, while a company over $20B carries a median of 92 lawyers and 158 total staff (ACC, 2024 Law Department Management Benchmarking Report; the median-6 figure comes from ACC/MLA 2022).
The cleaner way to think about staffing is lawyers per dollar of revenue, and here the pattern is steep. Overall it runs about 8 lawyers per $1 billion in revenue, but that average hides the shape. Small companies under $1B carry roughly 17 lawyers per $1B, medium companies from $1B to $5B carry about 4, and large companies over $5B carry about 3 (ACC, 2021 Law Department Management Benchmarking Report). The first lawyers a company hires are expensive on a per-dollar basis, and each additional billion in revenue needs proportionally fewer of them.
Industry matters too. Lawyer density ran from 10.2 per $1B in the information sector down to 7.7 in finance (ACC and MLA, 2025 Law Department Management Benchmarking Report). Measured against headcount instead of revenue, ratios move from about one lawyer per 200 employees at small companies to about one per 1,000 at large ones, with information-technology companies near one per 270 (ACC 2024). On composition, roughly 66% of a department is lawyers, 12% paralegals, and 5% legal operations (ACC 2022).
There is a practical rule buried in all of this. Practitioner guidance from Thomson Reuters (2024) says the moment to bring work in-house is when outside-counsel fees reach roughly twice the fully loaded cost of an in-house lawyer. That single ratio explains most first-lawyer hiring decisions.
For the full breakdown by revenue, industry, and headcount, see our deep-dive on in-house legal team size benchmarks. If you are earlier in the journey and still defining the role, start with what in-house counsel is.
What the team spends
Spending splits into two very different lines: the small, still-experimental AI budget, and the large, steadily rising outside-counsel bill.
The AI budget is tiny and mostly invisible
AI is not yet a meaningful line item for most departments. In a survey of 256 North American in-house professionals, 26% spend under $100 a month on AI, only 9% spend over $2,000 a month, and 44% do not know their AI spend at all (Counselwell and Spellbook, AI in Legal Departments: 2025 Benchmarking Report). Read that 44% twice. Nearly half of teams cannot name what they pay, which tells you AI is still bought at the seat level or buried inside software they already have: a Microsoft Copilot license, AI features bundled into a contract-management or e-billing renewal, a matter-management add-on. It is spent, but not tracked as a budget line.
The full picture of AI pricing, from per-seat tools to enterprise contracts, is in our legal AI pricing benchmark.
The outside-counsel bill keeps rising faster than inflation
This is the line that actually hurts. Worked rates at large US firms rose 7.4% in 2025 against 2.8% inflation, the latest in a decade of increases running at roughly twice inflation. Sticker rates at the top of the market are up about 50% from 2020 to 2026, against roughly 22% CPI over the same stretch (Thomson Reuters Institute, Law Firm Rates Report 2026, October 2025).
What that buys in 2026: AmLaw 100 partners quote sticker rates of roughly $1,500 to $3,000 an hour, AmLaw 200 partners $1,100 to $1,800, mid-size regional firms $700 to $1,200, and solos $300 to $600. Effective rates after discounts land around 70% to 80% of sticker. The national all-practice average lawyer rate is $349 an hour, from a high of $492 in Washington DC to a low of $196 in West Virginia (Clio Legal Trends Report, data 2025, updated March 2026).
There is one early sign of pricing moving, and it reads as market signal more than a settled benchmark. In 2025, several AmLaw 100 firms reportedly began quoting first-pass NDA review at a fixed $250 to $500 per document, replacing a prior baseline of about 1.5 to 3 associate hours (Thomson Reuters Institute rates reporting, 2025). That is the kind of routine, high-volume task where fixed fees and AI both bite first.
For the full rate bands, discount data, and how to negotiate them, see our outside counsel rate benchmarks.
How AI is changing the work
Adoption is the most-surveyed number in legal AI, and it climbed across every source we checked. Inside corporate law departments, generative AI use rose from 23% in 2024 to 52% in 2025 (ACC and Everlaw, 2025). For comparison, all-lawyer AI use across firms was 30%, ranging from 18% at solo firms to 46% at firms of 100-plus attorneys (ABA, 2024 Legal Technology Survey Report, released March 2025).
In-house teams are ahead of firms on frequency too. Weekly generative AI use was 76% in corporate legal against 68% at law firms (Wolters Kluwer, Future Ready Lawyer 2024). Zoom out to the whole profession and the tool is moving to the center of the desk: generative AI as a primary or central tool nearly doubled among legal professionals, from 14% in 2024 to 26% in 2025 (Thomson Reuters, Future of Professionals, 2025). And the money is expected to follow, with 65% of legal professionals expecting their organization's AI investment to rise over the next three years (Wolters Kluwer, 2024).
The productivity case is real but mostly projected, not banked. AI could free up about 12 hours a week per professional by 2029, and 4 hours a week within a year, framed as roughly $100,000 a year in billable capacity for a US lawyer (Thomson Reuters, Future of Professionals, 2024). A separate modeled estimate put task exposure at about 44% of legal tasks that generative AI could automate (Goldman Sachs, 2023), though that is a model of what is exposed, not a measurement of what got done.
Three facts keep the hype honest. First, accuracy. Leading legal research AI tools, including Lexis+ AI, Westlaw AI-Assisted Research, and Ask Practical Law AI, hallucinated 17% to 33% of the time (Magesh et al., Journal of Empirical Legal Studies / Stanford RegLab, 2025). Second, measurement. Only 7% of in-house teams track AI with specific KPIs (ACC and Everlaw, 2025), so most departments cannot yet prove their own return because they are not measuring it. Third, depth. Wide adoption is shallow: only 23% of in-house lawyers use AI daily and 27% have not touched it in six months (Bloomberg Law, State of Practice 2026). Across all industries, 42% of enterprises abandoned most of their AI initiatives in 2025, up from 17% a year earlier (S&P Global, Voice of the Enterprise, 2025, an enterprise-wide figure, not legal-only). A tool that clears procurement and never enters the weekly workflow is the most common way legal AI fails.
The wider set of adoption, spend, and accuracy figures lives in our legal AI statistics reference.
The outside-counsel paradox in the state of in-house legal teams
Here is the sharpest tension in the data. In-house teams clearly expect AI to shrink their reliance on law firms: 64% of in-house respondents expect generative AI to reduce that reliance (ACC and Everlaw, 2025). The promised savings have not landed. In the same survey, nearly 60% reported no noticeable savings yet from firms' AI use, and 58% said their firms have not adjusted pricing at all.
So adoption is up, expectations are high, and the bill has not moved. Rates rose 7.4% in a single year while more than half of clients saw neither savings nor a price change. The firms are getting more efficient behind the scenes, and mostly keeping the gains.
The NDA fixed-fee experiment is the exception that proves the rule. Where firms did reprice, it was on the most routine, most AI-exposed work, first-pass NDA review at $250 to $500 a document, not on the complex matters where the leverage is. Expect the savings to arrive task by task, from the bottom of the complexity stack up, rather than as a broad rate cut.
For the tactics that actually move the number, see reduce outside counsel spend with AI.
What this means for your team
Three tensions define the in-house function in 2026, and a GC benchmarking a team should hold all three at once.
Adoption is up, and being in the majority is now the default. If your team is not using generative AI at all, you are behind the 52% of corporate law departments that are, and behind the 76% using it weekly. That is a real signal, not vendor noise. In-house teams move faster than firms here because they own the budget and feel the cost pressure directly.
Savings are lagging adoption. The value most of the survey population expects has not shown up on the invoice, and only 7% of teams are even measuring for it. So do not slow down, but set a baseline before you buy. In our own work with in-house teams, the thing almost nobody records first is the current cost and turnaround of the specific task they want to automate, usually contract review or first-pass research. Four numbers make a usable baseline: average turnaround time on that task, outside-counsel hours it currently consumes, internal review hours, and rework rate. Capture those once before rollout and you become one of the few teams that can prove its own return a year later.
Here is that baseline as a scorecard you can copy. Fill the "before" column the week before rollout, the "day 30" column a month in, and the gap between them is the return you can take to your CFO.
| Metric | Before AI | Day 30 | What good looks like |
|---|---|---|---|
| Weekly active users on the team | Climbing week over week | ||
| Turnaround on the target task | Down, at the same volume | ||
| Outside-counsel hours on that task | Down, or flat as volume grows | ||
| Internal review hours on that task | Down, net of verification time | ||
| Rework or error findings | Flat or lower, never higher | ||
| Monthly spend (seat cost) | Well under the hours saved | ||
| Continue or kill | Continue only if usage is rising |
The two rows that decide it are the first and the last. If weekly active use is not climbing by day 30, the tool has not entered the workflow and the rest of the table will not save it.
Here is the same scorecard filled in for an illustrative team, using round numbers you can sanity-check against your own. A $400M company runs a two-lawyer department, spends about $900,000 a year on outside counsel, and has a panel partner rate that rose from $850 to $1,020 an hour over two years. They pilot AI on first-pass NDA review for 30 days. Treat the per-task times as practitioner estimates, not sourced figures.
| Metric | Before AI | Day 30 | Read |
|---|---|---|---|
| Weekly active users (of 2) | 0 | 2 | Both lawyers in it, the pilot is real |
| NDA first-pass turnaround | ~40 min | ~18 min | Down more than half |
| Outside-counsel NDA hours / month | ~8 | ~2 | Routine work pulled back in-house |
| Internal review hours / month | ~10 | ~6 | Down, net of verification |
| Rework or error findings | baseline | flat | No new errors introduced |
| Monthly seat cost | 0 | a few hundred | Well under the hours reclaimed |
| Continue or kill | Continue | Usage rising, hours falling |
The CFO-ready conclusion from that table is not "AI is transformational." It is narrower and more defensible: on this one task the team pulled roughly six outside-counsel hours a month back in-house and cut its own review time, for a seat cost a fraction of the saving. That is the shape of a real 2026 legal AI return, one task at a time, proven on your own baseline, not a headline productivity number.
Accuracy gates trust. As long as leading research tools miss on a fifth to a third of citations, AI stays an assistant that speeds up a first draft, not an oracle that replaces review. Buy for the tasks where a human already checks the output anyway, which is where the fixed-fee NDA repricing is happening for a reason.
Put your own numbers next to the benchmarks. If you run 2 lawyers on $400M in revenue, you are near the small-company norm of roughly 17 per $1B. If your panel firm's rate jumped 20% in two years while the market moved 14%, you have a negotiation to reopen. Benchmarks are only useful when you compare your own line items against them.
The verdict
The 2026 in-house legal team is lean, roughly six people, and squeezed by an outside-counsel bill rising at about twice inflation. It has largely adopted generative AI, with corporate use doubling to 52%, but it has not yet captured the savings it expects, and it mostly is not measuring whether it will. The teams that come out ahead will be the ones that treat AI as a way to reset the cost of routine legal work and that instrument it well enough to prove the number, rather than the ones that adopt and hope.
FAQ
How big is the average in-house legal team? The median corporate legal department is about 6 people, and the median company has about 7 lawyers (ACC, 2024 Law Department Management Benchmarking Report; median-6 from ACC/MLA 2022). Size scales with revenue: companies under $1B tend to run about 2 lawyers and 4 total staff, while companies over $20B carry a median of 92 lawyers and 158 staff.
How much do in-house teams spend on AI? Not much, and most cannot say exactly. 26% of in-house teams spend under $100 a month on AI, 9% spend over $2,000, and 44% do not know their spend (Counselwell and Spellbook, AI in Legal Departments: 2025 Benchmarking Report). AI is still bought at the seat level for most teams rather than managed as a budget line.
Is in-house or law firm AI adoption higher? In-house is higher on both counts. Corporate law department generative AI use hit 52% in 2025 (ACC and Everlaw, 2025), versus 30% of all lawyers across firms (ABA, 2024). On frequency, 76% of corporate legal used it weekly against 68% at firms (Wolters Kluwer, 2024). In-house teams own their budget and feel cost pressure directly, which speeds adoption.
How fast are law firm rates rising? Worked rates at large US firms rose 7.4% in 2025 against 2.8% inflation, part of a decade of increases at roughly twice inflation, with top-of-market sticker rates up about 50% from 2020 to 2026 (Thomson Reuters Institute, Law Firm Rates Report 2026, October 2025). The national average lawyer rate is $349 an hour (Clio, data 2025).
How accurate is legal AI? Less accurate than the marketing suggests. Leading legal research AI tools, including Lexis+ AI, Westlaw AI-Assisted Research, and Ask Practical Law AI, hallucinated 17% to 33% of the time in a Stanford study (Magesh et al., Journal of Empirical Legal Studies / Stanford RegLab, 2025). That error rate is why human review of AI output stays mandatory in practice.
When should a company hire its first in-house lawyer? Practitioner guidance from Thomson Reuters (2024) points to the moment outside-counsel fees reach roughly twice the fully loaded cost of an in-house lawyer. Small companies under $1B in revenue tend to run around 2 lawyers, or roughly 17 lawyers per $1B of revenue (ACC, 2021 and 2024), so the first hires are the most expensive per dollar.
Are in-house teams cutting outside counsel with AI? They expect to, but it has not landed yet. 64% of in-house respondents expect generative AI to reduce reliance on outside counsel, but nearly 60% reported no noticeable savings yet and 58% said firms have not adjusted pricing (ACC and Everlaw, 2025). The one exception is routine work, where some AmLaw 100 firms now quote first-pass NDA review at a fixed $250 to $500 per document (Thomson Reuters Institute, 2025).
Sources
- ACC and Everlaw, Generative AI's Growing Strategic Value for Corporate Law Departments (October 2025), survey of corporate law departments: ACC newsroom.
- ABA, 2024 Legal Technology Survey Report (released March 2025). Coverage: ABA Journal.
- Wolters Kluwer, Future Ready Lawyer 2024 (named without a link; source blocks automated requests).
- Thomson Reuters, Future of Professionals (2024 and 2025). 2024 coverage: 12 hours per week by 2029. The 2025 primary report blocks automated requests, so it is named without a link.
- Magesh et al., Journal of Empirical Legal Studies / Stanford RegLab (2025): Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools.
- Goldman Sachs (2023), task-exposure estimate (named without a link; source blocks automated requests).
- Counselwell and Spellbook, AI in Legal Departments: 2025 Benchmarking Report (survey of 256 North American in-house professionals).
- Bloomberg Law, State of Practice 2026 (US lawyers; daily-use and six-month non-use figures; named without a link, the report page blocks automated requests).
- S&P Global, Voice of the Enterprise: AI and Machine Learning, Use Cases 2025 (survey of about 1,000 IT and business leaders; AI-abandonment figure is enterprise-wide, not legal-specific; named without a link).
- ACC, 2024 Law Department Management Benchmarking Report (survey of corporate legal departments), team-size and lawyers-per-employee figures. The median-6 figure comes from the ACC/MLA 2022 benchmarking report (both named without a link; the primary reports sit behind ACC membership).
- ACC and MLA, 2025 Law Department Management Benchmarking Report: MLA Global.
- ACC, 2021 Law Department Management Benchmarking Report (lawyers per $1B by company size).
- Thomson Reuters Institute, Law Firm Rates Report 2026 (published October 2025): Thomson Reuters.
- Clio Legal Trends Report (data 2025, updated March 2026), national and state lawyer rates.
Last updated: July 2026.
New legal AI guides, weekly.
Further Reading
Legal AI Statistics 2026: Adoption, Spend, and Accuracy
Read postWhat Lawyers Want From Legal AI in 2026: A 534-Lawyer Study
Read postAI Medical Record Review for Lawyers: How It Works, What It Costs, and Where It Breaks
Read postIs Legal AI a Bubble? What the 2026 Market Data Says
Read postWhat Lawyers Really Think of Legal AI in 2026 (Reddit + Reviews)
Read postHow to Verify a Legal AI Accuracy Claim (Before You Trust the Number)
Read post
Co-Founder & CEO · Attorney
Arshita leads product and strategy at Vaquill, building the legal AI suite that solo, small-firm, and in-house US lawyers use to run a matter end to end.
