Legal technology in September: adoption needs evidence
September’s legal technology reports show different ways to buy and use AI. The useful comparison is the quality, cost and record of completed work.
The brief
LexisNexis’s September survey reports AI use among 94% of its 543 legal-professional respondents. That measures reported use, not accuracy or value. Separate reports describe small-firm procurement, support programmes and Latham’s purchase of its own AI servers.
Our practical conclusion: compare completed work, including review, training and correction. Before a trial, define one task, its present cost and the evidence needed to accept the result. Keep a record of the source material, the reviewer and the final version. High adoption makes that discipline more useful; it does not establish that any particular tool works well.
The full record
September’s legal technology coverage offers a more useful question than whether firms are adopting AI: what evidence should justify putting it into a client workflow? A survey, a supplier case study and a large firm’s infrastructure purchase answer different parts of that question. They should not be combined into one claim that the profession has solved adoption.
Reported use is high; confidence remains conditional
In a survey published by LexisNexis on 2 September, 94% of 543 respondents reported using AI for legal work. Some 81% said they would feel more comfortable using tools grounded in legal sources. Concerns included inaccurate or fabricated information, cited by 83%, and confidential-data leakage, cited by 53%. These percentages describe the respondents, not a census of UK lawyers. Legal Futures’ account of the survey also reports that privacy, security, transparency and human oversight ranked ahead of cost in buying considerations.
Our take: a source-grounded answer still needs a check. In a trial, ask the reviewer to open the cited authority, confirm the proposition and record whether the final work was acceptable. Measure that review effort alongside production time. Comfort with a system and the correctness of a particular answer are separate observations.
A small firm’s buying decision included the service around the software
Osprey’s Garnett Williams Powell case study describes a Rhyl practice that implemented its system in 2025. The account emphasises reporting clarity, usability, implementation help and responsive support. It is supplier-published testimony about an earlier purchase, not a September transaction or an independent audit of the product.
Our take: include one support request in a procurement exercise. Test whether a user can explain the problem, obtain a useful answer and resume the task. Record the interruption and internal effort. A feature that depends on unavailable help may have a different operating cost from the one shown in a demonstration.
Support can matter as much as tool selection
A Bristol and Bath LegalTech programme description says Technology in Professional Services worked with more than 60 small and mid-tier law and accountancy firms and accelerated adoption in 75% of them. This is a programme-reported result across two professions. The page does not establish that every participant achieved a particular financial return.
Our take: give a trial an internal owner with time to help colleagues. Track support effort and completion of real tasks. A usage count alone cannot explain whether the firm has built a repeatable process or is relying on one enthusiastic person to rescue it.
Owning the infrastructure changes the decision
On 10 September, Law360 reported Latham & Watkins’ confirmation that it had bought Nvidia servers to develop an in-house AI system. The accessible report supports the purchase; it does not establish a cost saving or a model for every law firm.
Our take: compare build and buy against the same task. Include responsibility for access, maintenance, evaluation and continuity. Owning hardware is a decision about operating responsibilities as well as data control. A small firm can ask those questions of a supplier without copying a large firm’s infrastructure.
Calculate value after implementation and review
Bridge Legal Group’s 7 September commentary says its Swiss legal technology map grew from roughly 60 to 83 tools. Its argument is that licences are only part of the bill: implementation, integration, training and organisational change also count. The map is the adviser’s own coverage, not an exhaustive market census.
Our take: calculate the cost per acceptable completed task. Include staff time, rework, support and an allocated share of implementation costs. State the task volume and assumptions. Time released has value only if the firm can explain what happens to it; it is not automatically extra revenue.
Firm structure affects the record that must survive
In a 3 September Legal Futures opinion article, acquisition adviser Jeff Zindani argues that specialist boutiques can form while larger platforms consolidate. His small boutique survey and market interpretation are adviser evidence, not regulator statistics for the whole profession.
Our take: a launch or integration is a good moment to define who owns the client record. Test whether a new colleague can retrieve the approved instruction, its evidence and the reason for the decision after a team changes. That is a concrete requirement for formal communication: an attributable exchange that remains understandable beyond its original participants.
For a trial structure, see our Brodies pilot scorecard. The proposed checks here are editorial analysis, not completed benchmarks or assessments of the named firms’ controls.
We reviewed the linked sources on 30 September 2026. Survey responses, supplier testimony, programme results and adviser opinion are attributed separately. The Latham item uses the accessible Law360 excerpt confirming the purchase. No supplier was independently tested. The procurement exercises and conclusions are our analysis; they are not measured benchmark results.
- LexisNexis survey, 2 September 2026
- Osprey, Garnett Williams Powell case study
- Bristol and Bath LegalTech, TiPS programme description
- Law360, Latham server purchase, 10 September 2026; subscription may be required
- Bridge Legal Group, return on investment commentary, 7 September 2026
- Legal Futures, Jeff Zindani opinion, 3 September 2026
- Legal Futures, LexisNexis survey coverage, 2 September 2026