Sponsored by
The key focus this week is "which parts of legal work will clients still pay a firm to do". PwC has published research that tries to answer it with numbers, and the answer is more uncomfortable than the headline suggests. The Online Procedure Rule Committee has set out what it wants a digital justice system to look like, and professional indemnity insurers have started treating AI governance as an underwriting question.
AI in Practice
PwC puts a figure on which legal work AI can actually take
PwC published research on 3 August, titled "The new rules of legal services: five moves to win as AI rewrites value", and the finding that has been discussed is the reassuring one: around 80% of the UK legal market retains what PwC calls "meaningful barriers" to full automation, including courtroom advocacy, complex negotiation, regulated judgement calls, and the matters where a client wants a trusted human in the room. Firms expect AI to unlock efficiency gains worth roughly 16% of chargeable hours this year, up from 11% last year, and PwC puts the slice of UK legal work that AI could deliver at about £6bn, or roughly a tenth of the market.
The part worth sitting with is the breakdown rather than the headline. PwC splits the market by two variables at once, namely how easily the work can be automated and how capable the client is of doing it themselves. On that basis, around 22% of fees come from work that is relatively easy to automate and where the client is sophisticated enough to bring it in-house with AI support, with transactional commercial contracting for large enterprises and patent drafting given as examples. A further 39% faces moderate barriers, including personal injury and clinical negligence work and property work for SMEs. The remaining 39% is both hard to automate and involves clients who cannot easily self-serve, which covers a good deal of consumer and SME employment, family and litigation work.
PwC's central argument is that the profession has been asking the wrong question. Whether AI can technically perform a lawyer's work is only half of it: the harder question is how AI changes what clients choose to buy, and which parts of the chain they are content to handle themselves. The threat to a commercial practice is perhaps less the models and more general counsels who decide they no longer need to send the work out. The report sets out four delivery models (the AI-augmented firm, AI-augmented outsourcing, AI-native firms, and self-serve, where the client goes to Harvey, Legora or Claude and skips the firm entirely). The report argues that most practices will have to choose deliberately where they compete. It also notes that 53% of firms are worried about clients using automation to reduce demand, 73% about keeping pace in what partners are calling a tech arms race, and that only a minority have turned AI activity into money so far.
This author would read the 80% figure with some caution, and not only because PwC sells the transformation consultancy that follows from its own diagnosis. The three numbers do not sit together as neatly as the report implies: 80% of work has meaningful barriers, the AI-deliverable slice is valued at around a tenth of the market, and yet 22% of fees are described as relatively easy to automate with a client able to take the work in-house. Those are different cuts of the same market, and the one that should concentrate a partner's attention is the 22%, because that is revenue which can walk without any AI tool ever being bought by the firm. The practical instruction PwC gives is the right one: take a portfolio view of your own work and decide, area by area, what you defend, what you industrialise, and what you let go. This week's practice prompt is built to help you do that on paper.
Takeaways
Act: Take your fee income by practice area and sort each area by two questions: how automatable is this work, and how capable is this client of doing it in-house. The areas that score high on both are where your revenue is exposed first.
Watch: Whether the self-serve model shows up in your own enquiry data, particularly a fall in the routine, high-volume instructions that used to arrive without being chased.
Risk: Treating the 80% as reassurance. The exposure sits in a specific fifth of the market, and a firm concentrated there has a strategy problem that no amount of AI adoption will fix.
Read: PwC, and the write-up at Legal Futures
On your radar
A voluntary AI code is coming to the digital justice system, and the core online rules bite on 7 September: The Online Procedure Rule Committee published a statement on 30 July setting out its priorities, and the immediate one is a voluntary Code of Practice covering inclusion standards, technology and data standards, and standards for the responsible and ethical use of AI, aimed at all providers in the digital justice system including those working before proceedings begin. The Code will be non-binding and will be published alongside the formal response to the Committee's 2025 consultation on the Pre-Action Model and Inclusion Framework. Separately, the Committee's core rules have been laid as The Online Procedure (Rules and Practice Directions) Rules 2026 and come into force on 7 September 2026, with the first procedural rules, for possession proceedings, to follow later this year. Why it matters for UK lawyers: a non-binding code sounds easy to ignore until a court or a client asks whether you follow it, and the possession rules will land on a high-volume, deadline-driven practice area. Diarise 7 September, and if you do possession work, put the forthcoming practice direction on your team's agenda now. (Online Procedure Rule Committee, legislation.gov.uk)
Three quarters of the top 100 UK firms now have someone whose job is AI, and most of them are not lawyers: Research by Search Acumen, reported on 30 July, found that 76 of the top 100 UK firms have appointed a chief AI officer or equivalent in a standalone, often C-suite role, against 96 of the top 100 in the United States. Only 35% of the UK appointees are qualified lawyers, 81% are men, and all but two of the top 50 firms have made such a hire compared with 56% of firms ranked 50 to 100. Why it matters for UK lawyers: where AI is governed by someone with no practising certificate, the professional obligations that attach to the output still sit with the fee earner and the COLP, so the reporting line between the two matters more than the job title. Find out who owns AI governance in your firm, and whether a lawyer sits anywhere in that chain before something goes out the door. (Legal Futures)
You can now practise your submissions against an AI judge: BenchSim, built by a US litigation partner, runs video simulations of oral argument: you upload your written case, choose a judicial temperament (hot, quiet or neutral, drawn from trial and appellate styles) and argue against it. Artificial Lawyer covered the move from text and audio into video on 5 August. Why it matters for UK lawyers: advocacy practice is one of the few genuinely scarce things in a junior's training, and unlimited repetitions against a difficult bench has obvious value, though it is built for US courts and the questioning will not track an English judge's habits. If you test it, use anonymised or non-confidential papers, because uploading a live skeleton to a third-party tool raises exactly the confidentiality question in this week's For review section. (BenchSim, Artificial Lawyer)
The UAE has put AI into the judicial process itself: The UAE announced on 28 July what it describes as the world's first fully integrated AI-powered judicial platform, intended to speed up litigation and support legal research and decision-making, with the stated position that final judgments remain under human oversight. Services are due to start rolling out from September. Why it matters for UK lawyers: this is the comparator our own courts will be measured against, and it goes considerably further than the Digital Justice System's current ambitions, so it is worth understanding what "AI-assisted with human oversight" turns out to mean in practice. If you advise clients with Gulf exposure, ask how the platform affects timetables and case handling before September. (Gulf News)
Following on from last week's pricing story, unpredictable AI cost is being pitched as the push towards value pricing: A piece for Legal Futures on 4 August, by Rachel Coleman of iManage, argues that once AI cost varies matter by matter, the billable hour stops working as a pricing mechanism and firms are pushed towards pricing on value instead. It is a vendor byline and the argument is not new, but it follows directly from the consumption-based pricing this newsletter covered a fortnight ago and is worth reading if you are the person who will have to explain the numbers. Why it matters for UK lawyers: a variable input cost you cannot forecast is difficult to recover under an hourly rate, so the pricing conversation is arriving whether or not your firm wanted it. Work out how your engagement letters treat AI cost, as disbursement, overhead, or something priced into scope, before a client asks the question for you. (Legal Futures)
Ad Break
In order to help cover the running costs of this newsletter, please check out the advert below. In line with my promises from the start, adverts will always be declared.
The best voice models now listen, adapt, and resolve too.
Most CX platforms don't own the voice. They orchestrate a workflow, then call a third party for speech and transcription. Every hop adds latency, and latency is what turns a frustrated customer into a churned one.
ElevenAgents is the opposite. Built on the voice models the market already builds on, it runs voice, transcription, chat, and reasoning in one vertically integrated pipeline. Responses come back in under 400 milliseconds and sound human, not synthetic. When a caller gets frustrated, the agent detects it and shifts tone in real time: calm, reassuring, patient.
You keep full control. Plug in any LLM, connect tools, webhooks, and MCP servers, and ground every answer in your knowledge base. Launch in minutes, A/B test with Experiments, enforce Guardrails, and version every change.
More resolved conversations, less infrastructure stitching. Pricing is transparent and flat at $0.08 per minute.
For Review
The new rules of legal services: five moves to win as AI rewrites value (PwC)
The full report behind this week's lead, with the market segmentation, the four delivery models, and the five strategic moves set out in enough detail to test against your own fee income. Read it if you have any role in practice-area strategy or pricing, and read the segmentation table before the executive summary.
Read or listen: PwC
Artificial intelligence and legal professional privilege (Miller, via Legal Futures)
A broker's piece, which is exactly why it is useful: it looks at AI and privilege from the professional indemnity side and reports that insurers are beginning to treat AI governance as an underwriting consideration, with Euclid's Chris Savvas quoted as saying what matters is whether a firm can demonstrate enforced governance through policies, training, supervision and an audit trail. It works through why privilege cannot be reasserted once confidentiality has gone, why closed enterprise systems reduce rather than remove the risk, and it draws on the Upper Tribunal's warning in Munir about uploading confidential material to open tools. There is a self-audit at the end.
Read or listen: Legal Futures, and the Upper Tribunal decision at The National Archives, Munir [2026] UKUT 81 (IAC)
The Online Procedure Rules 2026: a summary, and the rules themselves (Civil Litigation Brief)
Gordon Exall's walk through the rules which come into force on 7 September, with the text alongside the summary. If you conduct civil litigation, this is the practical preparation for the first radar item, and it is a quicker route to the substance than the statutory instrument on its own.
Read or listen: Civil Litigation Brief
Practice Prompt
Try the below prompt to run PwC's portfolio question over your own practice, so you can see which of your work is exposed to clients taking it in-house rather than to a tool replacing you. Ensure you fill in context and constraints and other aspects marked with {}. Remember to adhere to the Golden Rules and do not upload confidential or privileged information to public tools.
You are assisting the management of a law firm in England and Wales with a strategic review.
Your task is to produce an automation and disintermediation exposure assessment for the
firm's practice areas, as a planning aid only.
Context to apply:
- The firm: {size, offices, and total fee income if you are willing to state it}
- Practice areas and rough share of fee income: {e.g., "commercial property 30%, corporate
20%, employment 15%, private client 15%, commercial litigation 20%"}
- Client base for each area: {e.g., "corporate clients with in-house legal teams" /
"owner-managed businesses with no in-house counsel" / "individual consumers"}
- Current AI use: {e.g., "a firm-wide assistant used for summarising and first drafts" /
"nothing formal"}
- Pricing model: {e.g., "hourly rates, some fixed fees on property"}
Produce the assessment under these headings:
1. Exposure grid
For each practice area, score two things separately on a scale of low, medium or high:
(a) how far the work can be automated with currently available tools, and (b) how capable
that area's clients are of doing the work themselves with AI support, taking account of
whether they have in-house legal resource. Set the results out as a table and explain each
score in one line.
2. Highest-exposure work
Identify the areas scoring high on both measures, since this is the revenue most likely to
disappear through clients in-sourcing rather than through the firm failing to adopt a tool.
Say what specifically within each area is exposed, at task level rather than area level.
3. Defensible work
Identify the work where either the task resists automation or the client cannot realistically
self-serve, and say what makes it defensible: regulated judgement, advocacy, negotiation,
relationship, or the client's lack of internal capability.
4. Portfolio options
For each area, set out the realistic options (defend, industrialise so it can be delivered
profitably at lower cost, reprice, or exit) with the trade-offs of each. Do not recommend a
single course: set out what each option would require.
5. Evidence to gather
List the internal data the firm should pull before acting on any of this, for example
realisation rates by area, enquiry volumes over time, matter mix by client type, and which
clients have added in-house legal capacity.
6. Questions the assessment cannot answer
Flag where the analysis depends on assumptions you have had to make, and what the firm would
need to know to replace each assumption with a fact.
Constraints:
- {Add firm-specific constraints, for example a merger under discussion, a lock-up problem, or
a practice area the partnership will not exit.}
- Apply the market and regulatory context of England and Wales, and the SRA Standards and
Regulations where relevant.
- Do not invent market data, competitor behaviour, or figures for the firm. Where you do not
have a number, say so and mark it as an assumption.
- Do not treat AI capability as fixed: note where a score would change if tools improve.
- This is a strategic planning aid, not legal, financial or investment advice, and the
partnership remains responsible for any decision taken on the back of it.How did we do?
Hit reply and tell me what you would like covered in future issues or any feedback. We read every email!
Thanks for reading,
Serhan, UK Legal AI Brief
Disclaimer
Guidance and news only. Not legal advice. Always use AI tools safely.
Recommended Newsletters
Below are a few newsletters that I recommend, for various reasons. Check them out!
