What do law firms spend on AI per fee earner?

UK mid-market law firms spend between £1,200 and £4,500 per fee earner per year on AI tools, custom systems, and associated infrastructure. Top-50 firms spend £6,000 to £15,000. US firms spend roughly 40 to 60 percent more in absolute terms. The more important question is not how much you spend but where you spend it, because most firms waste their AI budget on low-impact tools while ignoring the workflows that would generate real returns.

Short answer: £1,200 to £4,500 per fee earner annually for UK mid-market firms. Top-50 firms spend £6,000 to £15,000. Allocation matters more than total spend.

Where these numbers come from

There is no single authoritative survey of AI spend per fee earner. The figures here are compiled from Law Society technology surveys, ILTA benchmarking reports, and our direct experience working with mid-market firms. We have also factored in publicly reported investments from larger firms and adjusted for the mid-market.

The numbers are imprecise because firms categorise AI spend differently. Some include ChatGPT Enterprise licences. Others only count purpose-built legal AI tools. Some amortise custom build costs over three years. Others expense them immediately. We have tried to normalise for these differences, but treat these as ranges rather than exact figures.

What is clear is that AI spend is growing faster than any other technology line item. Firms that were spending nothing on AI in 2024 are now allocating meaningful budget. The question has shifted from “should we spend on AI?” to “how much and where?”

Spend by firm size

Solo practitioners and micro firms (1 to 5 fee earners)

Typical AI spend: £300 to £1,000 per fee earner per year.

This is almost entirely SaaS subscriptions. ChatGPT Plus or Enterprise (£20 to £50 per month), a legal-specific AI tool like Harvey or CoCounsel (£100 to £300 per month), and occasionally a transcription or document tool. Custom builds are rare at this size because the volume does not justify the investment.

The opportunity: even at this level, the right £500 per year tool can save 5 to 10 hours per week for a solo practitioner. That is enormous ROI. The problem is tool selection. Most solo practitioners are not equipped to evaluate which AI tools will actually help their specific practice.

Small firms (6 to 25 fee earners)

Typical AI spend: £1,200 to £3,000 per fee earner per year.

Small firms start to see the benefit of at least one custom system, usually client intake automation. A £20,000 intake system spread across 15 fee earners costs £1,333 per fee earner in year one and under £400 per fee earner in subsequent years. Add SaaS tools and the total reaches the £1,200 to £3,000 range.

This is the segment where we see the highest ROI per pound spent. Small firms deploying one well-targeted custom system alongside 2 to 3 SaaS tools consistently outperform larger firms with sprawling, unfocused AI strategies.

Mid-market firms (26 to 200 fee earners)

Typical AI spend: £2,500 to £4,500 per fee earner per year.

Mid-market firms typically have 2 to 5 production AI systems, a handful of SaaS subscriptions, and are experimenting with more. The spend covers custom builds (often amortised), SaaS licences, API costs, hosting, and some internal resource for AI management.

This is where the allocation question becomes critical. A 50-person firm spending £200,000 per year on AI (£4,000 per fee earner) can either spread it thinly across ten tools or concentrate it on three high-impact systems. We consistently see better outcomes from concentration.

Top-50 and enterprise firms (200+ fee earners)

Typical AI spend: £6,000 to £15,000 per fee earner per year.

These firms have dedicated AI teams, enterprise licences, custom platforms, and ongoing consultancy relationships. The headline spend is higher but the ROI per pound is often lower because of bureaucratic overhead, slow deployment cycles, and a tendency to build platforms rather than ship solutions.

What the money goes on

SaaS subscriptions (30 to 50 percent of spend)

The bulk of most firms’ AI spend goes to subscription tools. ChatGPT Enterprise or similar general-purpose AI costs £20 to £50 per user per month. Legal-specific tools like Harvey, CoCounsel, or Luminance cost £100 to £500 per user per month. Document management AI add-ons cost £50 to £200 per user per month.

SaaS is easy to buy, easy to justify, and easy to cancel. It is also easy to waste money on. We regularly see firms paying for 6 to 8 AI subscriptions where staff actively use 2 to 3.

Custom systems (20 to 40 percent of spend)

Custom-built AI systems, either developed internally or by consultancies like us, typically cost £15,000 to £150,000 to build and £3,000 to £12,000 per year to maintain. Amortised over three years, a £45,000 system costs about £1,000 per fee earner per year for a 20-person firm.

Custom systems deliver the highest ROI when targeted at high-volume, firm-specific workflows. The Calder and Reid intake system we built cost £25,000 and saves £78,000 per year. That is a fundamentally different value proposition from a SaaS tool that saves each user a few minutes per day.

API and infrastructure costs (10 to 20 percent of spend)

This is the hidden cost that catches firms off guard. GPT-4 API calls, embedding storage, vector databases, and cloud hosting add up. A moderately busy custom system can cost £200 to £800 per month in API calls alone. Firms running multiple systems need to budget £5,000 to £20,000 per year for infrastructure.

Training and change management (10 to 15 percent of spend)

Often overlooked in budgeting, training costs are real. Staff need time to learn new tools. Workflows need to be redesigned. Champions need to be identified and supported. Budget 10 to 15 percent of your total AI spend for this, or risk building systems that nobody uses.

Where firms under-invest and over-invest

Common over-investments

Too many general-purpose AI subscriptions. Firms buy ChatGPT Enterprise, Microsoft Copilot, and a legal AI tool that all do roughly the same thing for document drafting. Pick one and use it properly.

Enterprise platforms before proving use cases. Some firms spend £50,000 to £200,000 on AI platforms before they have a single production use case. Build the use case first, then decide if you need a platform.

Proof of concepts that never ship. Money spent on POCs that sit in a sandbox and never reach production is wasted. If a POC cannot ship within 8 weeks, the concept was wrong.

Common under-investments

Client intake automation. This is the single highest-ROI use case for most firms, yet many allocate zero budget to it. A firm spending £4,000 per fee earner on AI but nothing on intake automation is almost certainly misallocating.

Data preparation and integration. AI systems are only as good as the data they work with. Firms under-invest in data cleaning, system integration, and workflow documentation.

Ongoing maintenance and improvement. AI systems are not set-and-forget. They need monitoring, updating, and refinement. Budget for year-two improvements, not just the initial build.

How to benchmark your own firm

Calculate your current AI spend per fee earner by adding:

All AI-related SaaS subscription costs (annual)
Custom system build costs amortised over three years
Custom system annual maintenance and hosting costs
API and infrastructure costs (annual)
Internal staff time allocated to AI management (valued at cost)

Divide by your number of fee earners. If you are below the range for your firm size, you are likely under-investing. If you are above it, check that your spend is generating proportional returns.

The more important benchmark is ROI per pound spent. A firm spending £1,500 per fee earner with a 500 percent return is in a better position than one spending £8,000 with a 50 percent return.

Start with a focused AI audit that identifies your two or three highest-ROI opportunities. Allocate 60 to 70 percent of your AI budget to those opportunities and keep 30 to 40 percent for SaaS tools and experimentation. Plan to spend £2,000 to £4,000 per fee earner in year one and adjust based on measured returns.

The firms getting the best results are not the ones spending the most. They are the ones spending deliberately, measuring rigorously, and expanding based on evidence. A £250 workflow audit can tell you where your money will work hardest before you commit to anything.

FAQ — RELATED QUESTIONS

What is the average AI spend per fee earner in a UK mid-market law firm? +

Between £1,200 and £4,500 per year. This includes SaaS subscriptions, custom system costs amortised over their useful life, and API costs. Firms at the higher end typically have 2 to 4 production AI systems running.

How does AI spend per fee earner differ between UK and US firms? +

US firms spend roughly 40 to 60 percent more in absolute terms, driven by higher billing rates and larger technology budgets. As a percentage of revenue per fee earner, spend is comparable at 0.5 to 1.5 percent.

Should AI spend be allocated per fee earner or per matter? +

Per fee earner is the standard benchmark but per matter is more useful for budgeting specific systems. An intake system cost is better understood per enquiry handled. A drafting system is better understood per document produced.

What percentage of a law firm's technology budget goes to AI? +

In 2026, AI represents 8 to 18 percent of the total technology budget for firms actively deploying it. This is up from under 3 percent in 2024. The share is growing as firms shift spend from legacy case management upgrades to AI systems.

Is it better to spend on AI SaaS tools or custom builds? +

SaaS tools are cheaper per seat but deliver less competitive advantage. Custom builds cost more upfront but generate higher ROI for high-volume workflows. Most firms should use SaaS for generic tasks and custom systems for their core money-making processes.

How do I justify increasing AI spend to the partnership? +

Frame it as cost per hour saved, not cost per seat. If a £3,000 per fee earner investment saves 200 hours of solicitor time per year at a blended cost of £70 per hour, the return is £14,000 per fee earner. Present the ratio, not just the cost.

What are the hidden costs of AI beyond the headline per-fee-earner figure? +

Training time, change management, data preparation, and integration with existing systems. These typically add 20 to 40 percent to the headline cost in year one, then drop to under 10 percent in subsequent years.

Do smaller firms spend less per fee earner on AI than larger firms? +

Yes, but the gap is narrowing. Smaller firms spend less in absolute terms but often get better ROI because they deploy AI on a smaller number of high-impact workflows. A 15-person firm spending £2,000 per fee earner on one well-chosen system can outperform a 200-person firm spending £8,000 across ten mediocre ones.