Back to Industry Insights

AI in Law: Hype vs reality

A clear-eyed view of adoption, performance, opportunity, and risk

Executive Summary

AI is now a practical tool inside UK solicitor firms rather than an experimental curiosity. Adoption has accelerated sharply over the last 18 months, driven by pressure on margins, client expectations around speed and cost, and the growing maturity of legal-specific AI tools. However, most firms remain early in organisational maturity. Individual solicitors and teams are using AI day to day, but relatively few firms have embedded AI into repeatable, governed workflows.

A growing and under-discussed dynamic is the externalisation of AI use by clients. Solicitors increasingly receive AI-generated analyses from clients that are lengthy, poorly reasoned, or legally incoherent. These submissions often arrive with the expectation of rapid validation and low cost. In practice, firms report that this increases workload rather than reducing it, as solicitors must now deconstruct and correct flawed reasoning before providing advice. Several solicitors noted that, in many cases, it would be faster and safer to start from a blank page.

Large firms are moving faster and more visibly. They are piloting AI at scale, formalising use cases, and measuring outcomes such as turnaround time, capacity, and write-offs. Smaller firms are adopting more cautiously, focusing on immediate productivity gains in drafting, research, and administration, often without the same level of formal governance.

Key findings:

  • Where AI is working well, firms report faster execution, reclaimed time for higher-value work, and improved consistency.
  • Where it fails, the causes are predictable: unclear rules, unmanaged shadow AI, over-reliance on unverified outputs, and poor alignment between AI outputs and client-specific nuance.
  • In practice, most current AI tools operate at roughly a junior solicitor level: useful for gathering context and structuring first drafts, but incapable of applying judgment, commercial awareness, or client-specific "spin" without senior oversight.
  • The next phase of impact will come from moving beyond single-prompt copilots to structured, multi-step workflows and early agentic systems. This creates real opportunity, but also expands risk. Firms that treat AI as an operating model change rather than a collection of tools will be best placed to benefit.

1. How UK Solicitor Firms Are Testing AI

Common Early Testing Patterns

Across firms of all sizes, AI adoption typically begins in areas where outputs are easy to review and the consequences of error are limited. These "safe-start" use cases include:

  • First drafts of emails, letters, clauses, and client updates
  • Summarising long documents, correspondence, or bundles
  • Accelerating legal research using tools grounded in trusted legal sources
  • Clause identification, document classification, and anomaly spotting in reviews
  • Internal productivity tasks such as meeting summaries and knowledge search

These uses allow solicitors to build confidence while maintaining human control over final outputs. Importantly, firms reporting positive outcomes emphasise that AI outputs are treated as inputs to thinking, not answers.

Tooling Signals from the Market

A frequently cited platform is Harvey, which has been designed and developed in collaboration with A&O Shearman. Feedback is mixed. Some solicitors value its research acceleration and structured outputs, while others remain unconvinced about its ability to handle nuance or complex client context. This divergence reflects a broader truth: perceived value is highly dependent on the type of work, quality of internal precedents, and clarity of review processes.

2. How AI Is Performing in Practice

What Firms Report Is Working

Solicitors consistently identify AI's strongest capabilities as:

  • Advanced search and summarisation: Across large volumes of case law, correspondence, and internal documents
  • First-pass drafting: Of emails and client communications that are then refined by the solicitor
  • Context gathering: Ahead of deeper legal analysis

AI's ability to filter and structure large bodies of information, which can then be peer-reviewed, is widely seen as a genuine time saver. In this role, AI acts as a force multiplier rather than a decision-maker.

Where Performance Disappoints

In addition to previously identified issues, firms report a growing challenge created by client use of generative AI:

  • Clients submit long, AI-generated analyses containing legal errors or false assumptions
  • Clients expect rapid validation rather than substantive advice
  • Cost expectations are misaligned with the effort required to correct flawed reasoning

This shifts work from advisory to remedial review, increasing cognitive load and professional risk. Firms that fail to address this explicitly risk margin erosion and client dissatisfaction.

Internally, disappointment also arises where AI outputs are mistaken for "near-final" work rather than junior-level drafts, solicitors lack clear guidance on how much verification is required, and junior solicitors are unclear how AI fits into their development path.

3. Implications for Legal Talent and Training

Several senior solicitors raised concerns about the impact of AI on junior development. If AI routinely performs tasks historically used for training, firms risk weakening the pipeline of future senior solicitors.

Leading firms are beginning to respond by:

  • Treating AI explicitly as a junior solicitor, not a replacement
  • Redesigning training to focus juniors on review, critique, and judgment rather than raw drafting
  • Making AI use visible and discussable rather than implicit

Handled well, AI can accelerate learning. Handled poorly, it can hollow out core skills.

4. Opportunities in the Practice of Law

The core opportunities remain, but with clearer boundaries.

Legal Research and Advisory Work

AI is most effective when used to:

  • Produce structured first-pass research notes
  • Surface relevant authorities from large corpora
  • Create issue-spotting frameworks

Its value lies in coverage and speed, not judgment.

Drafting and Client Communication

AI performs well at:

  • Starting emails and letters
  • Drafting standard clauses from known precedents
  • Producing plain-English summaries

Final positioning, tone, and risk assessment remain human responsibilities.

5. Back-Office and Knowledge Infrastructure

Most AI discussion remains focused on fee-earning work, but many of the lowest-risk, highest-return opportunities sit in business services.

Knowledge Management and Retrieval

There is strong potential in:

  • Knowledge graphs linking matters, precedents, clauses, and outcomes
  • Retrieval-augmented generation systems grounded in internal and historic case data
  • Agent-based systems that search, summarise, and route information for review

These approaches reward firms that invest in structured data and disciplined document management.

On-Prem and Private AI Considerations

Some firms operate primarily with on-premises storage and express concern about AI accessibility. This is not a blocker. With modern open-source models and private deployment options, firms can host AI tooling on-prem or in isolated environments. The constraint is cost and access to skills rather than feasibility, driven largely by model size, bringing in AI expertise and usage patterns.

6. Risks and Controls

False Authority and Client Expectations

Firms increasingly need to manage not just AI risk, but AI-inflated client confidence. Practical responses include:

  • Setting expectations early about how AI-generated client material will be treated
  • Reframing advice as outcome-based rather than validation-based
  • Documenting review and correction effort explicitly

Conclusion

AI is already changing how UK solicitor firms operate, but its most profound effects are organisational rather than technical. In practice, today's tools function like junior solicitors: helpful, fast, and occasionally wrong. The firms seeing sustained benefit are those that recognise this and design workflows, training, and client communication accordingly.

Over the next two years, the gap will widen between firms that embed AI into disciplined operating models and those that allow it to emerge ad hoc, whether internally or via clients. The difference will not be model quality, but professional maturity.

Ready to Transform Your Legal Practice with AI?

Our experts can help your firm develop a practical AI strategy that delivers measurable results while managing risk effectively.