Skip to content
Sector programmes · Legal services

AI in legal services,
checked, cited, defensible.

Independent, standards-led support for law firms, chambers, and in-house teams putting AI to work: grounded in SRA guidance, judicial expectations, and the duties of confidentiality and privilege, free of vendor incentives.

— State of play —

The profession moved fast.
The duties did not move at all.

Generative AI use among UK lawyers has grown several-fold in two years, from research and first drafts to disclosure review and contract analysis. Clients are asking about it in panel reviews, and the larger firms are building dedicated capability. The pressure to adopt is now competitive, not just curious.

The duties, meanwhile, are unchanged and the courts have shown they will enforce them. The High Court's 2025 ruling in Ayinde v Haringey, where invented authorities reached filed submissions, made the position plain: the lawyer is responsible for every citation, however it was drafted. SRA guidance, judicial guidance on AI, and legal professional privilege set the rest of the frame. Adopting quickly and staying inside it is where most firms need help.

~2 in 5
UK lawyers report using generative AI for work
4x
growth in professional use in around two years
<1 in 5
firms with a firm-wide AI policy in place
— Use cases —

Where AI is already
earning its keep.

The strongest legal use cases put speed into the first draft and keep a qualified lawyer responsible for the final one.

Research and drafting

First drafts, case summaries, and research memos produced in minutes, with every authority verified by a lawyer before it goes anywhere near a client or a court.

Disclosure and review

Large document sets searched by meaning rather than keyword, surfacing the documents that matter earlier and at a fraction of the review cost.

Contract analysis

Review against playbooks, clause extraction, and risk flagging across whole portfolios, so negotiation time goes to the clauses that deserve it.

Client intake and triage

Enquiries assessed, summarised, and routed to the right fee earner, with conflicts and limitation issues flagged earlier than a Monday-morning inbox allows.

Knowledge management

Precedents, opinions, and firm know-how made genuinely findable, so the second lawyer to face a problem starts from the firm's best answer, not a blank page.

Risk and compliance

AML checks, engagement letters, and file supervision supported with audit trails that satisfy the COLP, the insurer, and the regulator.

— Sector considerations —

What we watch
most closely.

01
Accuracy

Invented authorities end careers

Generative tools fabricate convincing citations, and the courts have made clear that filing them is the lawyer's failure, not the tool's. Every authority is verified against the source before it leaves the building.

100%
of citations verified by a human
02
Confidentiality

Privilege is unforgiving

Client confidentiality and legal professional privilege do not bend for convenience. Where a tool runs, what it retains, and who can see it are settled contractually before any client matter touches it.

0
client data into unvetted public tools
03
Supervision

The signature still means something

SRA expectations on supervision apply regardless of who, or what, produced the first draft. Work is supervised, attributed, and owned by a qualified lawyer, and the file shows it.

1
qualified lawyer accountable for every output
04
Clients

Candour beats discovery

Clients are starting to ask how firms use AI, and finding out by accident is the worst answer. Engagement terms and panel responses should say what is used, for what, and under whose review.

100%
transparent with clients about AI use
05
Insurance

Your insurer is already asking

Professional indemnity insurers increasingly want to know how AI is governed before they price the risk. A documented policy, training records, and supervision evidence keep that conversation short.

1
question to expect at every renewal

Set your firm's
AI position.

A short call with the legal services team. Plain answers, a clear next step, and no software to sell.