Key takeaways

  • Generative AI tools such as Harvey, Legora and CoCounsel are compressing tasks that once billed for hours into minutes, intensifying scrutiny of AI legal fees.
  • The billable hour rewards time spent, creating a structural tension with technology designed to save time.
  • Magic Circle firms including A&O Shearman, Linklaters and Slaughter and May have rolled out AI firmwide, yet top partner rates still reach roughly £1,000-£1,500 an hour.
  • Clients are pressing for alternative fee arrangements, fixed fees and value-based pricing as efficiency gains become visible.
  • The Solicitors Regulation Authority backs innovation but insists firms cannot delegate accountability for AI outputs.

For decades, the business model of the City law firm has rested on a deceptively simple proposition: the more time a matter takes, the more a client pays. Now that proposition is being tested as never before. Generative artificial intelligence can draft a first-cut contract, sift thousands of documents in a data room, or produce a research memo in minutes rather than the hours or days a trainee once spent. As those capabilities spread through Britain’s largest firms, an awkward question that clients have long muttered is being asked aloud: if the work is faster, why are the fees still so high?

The Financial Times has put the tension in stark terms. In its Lex commentary on legal services, the paper argued that deeper AI adoption should have senior lawyers “sharpening their hunting spears”, because the technology will only amplify client demands to cut fees. The image is pointed. It casts law-firm leaders not as passive beneficiaries of a productivity windfall but as combatants in a bargaining fight they may struggle to win. This article treats that FT reading as its news peg and builds out the evergreen picture around it: how City firms actually bill, what the new tools do, and why the collision between efficiency and the billable hour has become the defining commercial story of the profession.

The pressure on fees is no longer theoretical

For years, complaints about legal costs were met with a familiar defence: complex, high-stakes work demands scarce expertise, and expertise is expensive. That argument still holds for genuinely bespoke judgment. What has changed is that a growing share of the underlying labour, the reading, summarising, first-drafting and cross-checking, is now visibly automatable. When a general counsel can see a tool produce a serviceable first draft in seconds, the premium attached to a junior lawyer doing the same task overnight becomes harder to justify on the invoice.

The numbers involved are substantial. Analysis of top-end charge-out rates by legal-costs specialist Jim Diamond, reported by the Law Society Gazette, put Magic Circle partner rates at roughly £1,000 to £1,500 an hour, with five-year-qualified associates charged out at £650 to £850 and newly qualified lawyers at £450 to £600. US firms in London sat close behind. Diamond noted that top rates had roughly doubled over 15 years, far outpacing general inflation. Those figures pre-date the current wave of AI, but they frame the stakes: even modest efficiency gains applied to work billed at hundreds of pounds an hour translate into large sums, and clients know it.

The FT’s framing matters here because it reflects a shift in tone rather than a new fact. Corporate buyers of legal services have pushed on cost since the financial crisis. What AI adds is leverage and evidence. A client can now point to a specific tool, a specific task, and ask why it still appears on the bill at the old rate.

How the biggest firms bill, and why the billable hour is exposed

To understand the friction, it helps to understand the machine. The dominant pricing model at large UK firms remains the billable hour: lawyers record their time in six-minute units and clients are charged the sum of those units at each fee-earner’s rate. Layered on top is the concept of leverage, the ratio of junior lawyers to partners on a matter. A partner wins and supervises the work; associates and trainees do the bulk of the hours. Because juniors are charged out at a healthy multiple of their salary cost, high leverage is a central engine of firm profitability.

This is precisely the structure that generative AI unsettles. The billable hour rewards time; AI removes it. The leverage model depends on volumes of junior hours; AI is best at exactly the document-heavy, repeatable tasks juniors have traditionally absorbed. A firm that deploys AI effectively can complete a due-diligence exercise with fewer associate hours, which is good for the client’s bill and awkward for a revenue model built on those hours. The commentary from Lex captures this bind: the same tools that make a firm more efficient hand clients a stronger case for paying less.

None of this means the hourly model collapses overnight. Much premium legal work still turns on judgment, negotiation, relationships and risk appetite, which resist automation. But the mix is shifting, and the portion of a typical matter that AI can accelerate is the portion clients most resent paying full freight for.

What the new AI tools actually do

The tools driving this conversation are no longer speculative. Several are now embedded in the daily workflow of leading firms.

The main platforms

Harvey, built on large language models and tailored for legal use, has become the most visible name in the sector. Legora, a Stockholm-founded rival, has won significant enterprise deals and, according to Lawyer Monthly, planned to double its headcount in 2026 as the market consolidated around a handful of players. CoCounsel, from Thomson Reuters, brings AI to legal research and drafting within a trusted content ecosystem, while Microsoft’s Copilot pushes generative assistance into the everyday Office tools lawyers already use. Reported pricing for the leading legal-specific platforms runs into four figures per user each year, a cost firms are absorbing in the expectation of efficiency and client-facing revenue.

From research to drafting to due diligence

In practice, these systems compress three traditionally labour-intensive workstreams. In legal research, they surface relevant authorities and summarise them, cutting hours of database work. In drafting, they generate first versions of contracts, clauses, memos and correspondence for a lawyer to refine. In due diligence and document review, they read and categorise vast document sets, flagging risks in transactions or disputes far faster than a team of juniors could. The SRA’s own risk assessment noted that AI can reduce document-discovery time “from weeks to seconds” while improving accuracy, a striking framing from the regulator itself.

Firmwide rollouts at the top

The adoption at elite firms is now concrete. A&O Shearman pioneered ContractMatrix, an AI contract tool built with Harvey and Microsoft, and in 2025 announced a revenue-sharing arrangement with Harvey to build and sell agentic tools, initially targeting antitrust analysis, cybersecurity, fund formation and loan review, on subscription or usage-based pricing. Linklaters rolled out Legora firmwide across its offices in September 2025, alongside proprietary tools including its chatbot Laila and due-diligence platform ReportIQ; managing partner Paul Lewis said the partnership “further strengthens our capabilities”. In April 2026, Slaughter and May adopted Harvey across all practice areas. Managing partner David Johnson stressed that “critical to our adoption is the investment we make in our people, as the vital human layer that supervises AI”, while partner Sally Wokes framed the firm’s edge as “the depth and quality of our experience”.

The efficiency-versus-revenue tension

The uncomfortable arithmetic is this: a firm that passes all its AI-driven time savings to clients erodes its own revenue; a firm that keeps billing as before invites the accusation that it is charging premium rates for machine-assisted work. Neither pole is stable.

If efficiency gains are fully shared, headline revenue on affected matters falls, and firms must either win more work, move up the value chain to higher-judgment tasks, or restructure how they charge. If gains are quietly retained, firms face reputational and competitive risk as rivals, and clients’ own in-house teams using the same tools, expose the gap. The FT’s “hunting spears” image sits squarely on this fault line: senior lawyers can defend margins for a while, but the direction of client pressure is one way.

Some argue the tension is overstated in the near term. AI still requires supervision, verification and accountability, and the time saved on drafting can be partly reabsorbed by the time spent checking outputs for errors. Bloomberg Law commentary has noted that productivity gains have so far boosted output without toppling the billable hour outright. But even sceptics accept that the pricing conversation has permanently changed.

Alternative fee models gain ground

The most likely response is not the death of the billable hour but its dilution. Alternative fee arrangements, long discussed and slow to spread, look better suited to an AI world where the labour input for a given output is falling and less predictable in hourly terms.

Fixed fees for defined pieces of work let clients buy an outcome rather than an input, and let firms keep the upside if AI makes delivery cheaper. Capped fees limit client exposure while preserving some hourly logic. Value-based or outcome-based pricing ties cost to the significance of the result rather than the hours logged. Subscription and productised models, of the kind A&O Shearman is pursuing with agentic tools, move firms toward selling software and repeatable expertise rather than bespoke time. Thomson Reuters and others have argued that AI makes some shift toward fixed and value-based billing close to inevitable for commoditisable work, even if bespoke, high-stakes matters remain hourly.

The strategic prize for firms is to decouple revenue from hours before clients force the issue. The risk is that they move too slowly and cede the initiative, and the pricing narrative, to buyers.

What it means for clients, junior lawyers and firms

The impact lands differently across the market.

For clients, especially well-resourced corporate legal departments, the near-term gains are real: faster turnaround, potentially lower bills on routine work, and stronger negotiating leverage. Many in-house teams are deploying the same tools themselves, further shrinking the tasks they are willing to outsource at premium rates.

For junior lawyers, the picture is more anxious. The document review, first drafting and research that AI does best have historically been how trainees and newly qualified solicitors learn the craft and how firms generate leverage-driven profit. If those hours shrink, firms must rethink both training, how juniors develop judgment without the grind, and headcount. Firm leaders including Slaughter and May’s have publicly insisted that the “human layer” supervising AI remains essential, and that investment in people is central to adoption. Whether trainee cohorts and career structures survive intact is one of the profession’s most consequential open questions.

For firms themselves, AI is simultaneously a threat to a lucrative model and an opportunity to build higher-margin, technology-led services. Those that productise expertise, win share, and reprice intelligently could thrive. Those that treat AI as a cost centre bolted onto an unchanged billing model may find margins squeezed from both ends.

Opportunities, challenges and material risks

The opportunities are significant. Efficiency can widen access to legal services and improve margins on the right work. New revenue streams from selling AI-enabled products, as A&O Shearman is testing, could diversify firms beyond time-based billing. And better tools can free senior lawyers for the judgment-intensive work that justifies premium fees.

The challenges are equally real. The billable hour is deeply embedded in firm economics, partner compensation and client expectations, and unwinding it is hard. Training pipelines and junior recruitment may need wholesale redesign. Competition from legal-tech vendors and from clients’ own in-house adoption threatens to disintermediate parts of the value chain.

The material risks are where the regulator focuses. Generative AI can “hallucinate”, producing plausible but false content; the SRA has warned that lawyers have already cited non-existent cases generated by AI. Confidentiality and privilege are exposed when sensitive data is fed into cloud systems. Errors can propagate at scale, affecting many clients at once and complicating supervision. The SRA’s stance is pro-innovation but firm on accountability: it does not dictate which technologies firms adopt, yet insists they cannot delegate responsibility for AI outputs to IT teams or vendors, must test tools before deployment, and should train staff and, where relevant, inform clients. In short, faster does not mean unaccountable.

Counterarguments

Not everyone accepts that AI spells a fee reckoning. Sceptics make several points worth weighing.

First, premium legal work is bought for judgment, relationships and risk-bearing, not typing speed, and clients pay for outcomes and reassurance that resist commoditisation. Second, verification overhead is real: a lawyer must still stand behind every AI-assisted output, and that supervision consumes time and carries liability, which supports continued fees. Third, induced demand may offset efficiency, as cheaper, faster analysis encourages clients to seek more legal work, not less. Fourth, the billable hour has been declared dead many times, most recently with the rise of legal-process outsourcing and earlier document-review technology, and has proved remarkably durable. On this reading, AI reshapes the model at the margins rather than dismantling it.

These arguments have force, and they explain why the transition is likely to be evolutionary. But they address the pace of change, not its direction. Even firms confident in their premium positioning concede that clients will ask harder questions about AI legal fees, and that answering them credibly will require more transparency about how work is done and priced.

What happens next

Expect the pricing conversation to intensify rather than resolve. As AI adoption becomes standard across the Magic Circle and the wider City, differentiation will shift from whether a firm uses these tools to how well it deploys them and how honestly it prices the result. Alternative fee arrangements, fixed fees and productised services should gain further ground on commoditisable work, while bespoke, high-judgment matters stay closer to the hourly model. Firms will experiment with sharing AI-driven revenue, as A&O Shearman is doing, and clients will keep building in-house capability that reshapes what they buy externally.

The training question will move up the agenda. Firms that solve how to develop junior lawyers when routine hours shrink, and that retain talent through the transition, will hold an advantage. Regulators are likely to keep refining guidance, with the SRA and Law Society under continued pressure to clarify competence and disclosure duties. And the FT’s central warning, that AI hands clients a stronger hand in the fee negotiation, will be tested matter by matter, invoice by invoice.

Conclusion

The arrival of capable generative AI has not, on its own, made legal work cheap. But it has made the question of price impossible to dodge. When a client can watch a task that once justified hours of billing complete in minutes, the burden of explanation shifts onto the firm. Britain’s biggest law firms are responding, adopting the tools firmwide, protecting the human judgment that remains their real product, and beginning, cautiously, to rethink how they charge. The billable hour is unlikely to vanish. But the era in which it went unquestioned is ending, and the firms that confront the fee question openly are better placed than those hoping it goes away.