Why Copilot isn't working in your law firm (and what would)

Wednesday , September 23, 2026

Author

Nexian
featured-image

Author

Nexian

A pattern most firms will recognise


The licences were bought. A pilot ran. People went to the introductory session and the early enthusiasm was real. Several months on, a couple of people use Copilot well every day, and everyone else tried it twice and went back to doing the work by hand.

If that describes your firm, it is not an outlier.

Commentary across the UK legal sector this year has described exactly this trajectory: licences purchased, pilots run, then patchy usage, low confidence and a plateau [1].

Survey data tells the same story from both ends. LexisNexis found 61% of UK lawyers were using generative AI in their work by September 2025, up from 46% at the start of that year [2], and Wolters Kluwer's 2026 Future Ready Lawyer survey found over 90% of legal professionals now use at least one AI tool daily [3].

Yet independent survey work by Recon Analytics, covering more than 150,000 respondents, puts the proportion of workplace Copilot seat holders who actually use the tool at roughly 36% [4].

In other words: the profession has adopted AI, but a large share of paid Copilot seats sit idle. The licences are not the problem.

Nexian_CTA_AI_ROI_Calculator

The standard diagnosis, and why it falls short

Most advice on stalled Copilot adoption reaches the same conclusion: it is a training problem, a culture problem, a change management problem. There is something in each of those. A one-off demo session does not change how anyone works, and firms that treat rollout as a licence exercise get licence-exercise results.

But training only sticks when the tool can do something a fee earner actually needs done. And this is where the standard diagnosis runs out, because it never asks what Copilot was being asked to do when people gave up on it.

The real gap: Copilot can't see your matters

Out of the box, Copilot drafts an email, summarises a document and writes up a meeting. Genuinely useful. What it cannot do is answer the questions that actually run a law firm. It cannot tell you which matters are approaching a key date with nobody assigned. It cannot tell you where the unbilled time is sitting, or what stage a matter reached, or whether a new enquiry conflicts with an existing client.

That information lives in your practice management system, and Copilot has no route into it unless someone builds one. So a fee earner asks a matter question, gets a generic answer, concludes the tool does not understand their work, and stops. They are right, in a narrow sense: it is not connected to the systems that hold their work. The disappointment is a configuration gap, not a verdict on the technology.

This is the part that a training day cannot fix. You can teach people better prompts all year; a disconnected Copilot still cannot see the matters.

The second gap: nobody can evidence anything

There is a quieter version of the same problem at management level. Microsoft 365 includes native usage reporting for Copilot, but in most firms it has never been switched on, let alone segmented by department. So the managing partner cannot tell whether the licences are earning their keep, and the firm's compliance officer has no record of how AI is being used or supervised.

That second point matters more than it first appears. How AI use is supervised, and how that supervision is evidenced, is a question your COLP will want a good answer to, and it is worth putting to them before an audit or a client asks first. A firm that can show who is using AI, on what, under which policy, is in a very different position from a firm that can only say the licences exist.

What fixing it actually involves

Three things, and none of them is buying more licences.

Connect AI to matter data.

When Copilot is connected to the practice management system, fee earners can ask about matters, work in progress, clients and key dates inside the Microsoft 365 tools they already use. This is the single change that turns Copilot from a drafting toy into something a legal team relies on.

Build habits, not awareness.

One guided task a day for a month changes behaviour in a way a single training session never does, because people adopt tools that visibly help with their own caseload. Adoption programmes work when they are built around the firm's real work rather than generic exercises.

Switch the evidence on.

Turn on usage reporting, segment it by department, and put an acceptable use policy behind it. This is what lets the managing partner see whether the investment is working and lets the compliance officer show how use is supervised.

Where to start

Not with a platform decision, and not with more licences. Start by finding out which AI use cases are actually worth something in your firm, because the answer depends on your practice areas, your charge-out rates and the system your matters live in.

We have mapped 46 use cases for Copilot and Claude across UK legal practice, and built a free five-minute calculator that filters them to your practice areas and your practice management system, with an illustrative figure against each. No hype and no hard sell: some firms use it to sanity-check what a vendor is claiming, and that is fine too.

Nexian_CTA_AI_ROI_Calculator

Sources

[1] Legal Futures, 'Why so many firms lose momentum with Copilot' (March 2026), describing the stalled adoption pattern across UK firms. https://www.legalfutures.co.uk/associate-news/why-so-many-firms-lose-momentum-with-copilot-and-what-leaders-can-do-to-get-things-moving-again

[2] LexisNexis survey figures (61% of UK lawyers using generative AI, September 2025, up from 46%), as reported in SpotDev's AI for Law Firms UK guide. Secondary source: verify against the LexisNexis original before publication if preferred. 

[3] Wolters Kluwer 2026 Future Ready Lawyer Survey (over 90% of legal professionals use at least one AI tool daily), as cited by the American Bar Association. Secondary source: verify against the Wolters Kluwer original before publication if preferred. 

[4] Recon Analytics survey of 150,000+ respondents (~36% of workplace Copilot seat holders actively use it), as cited by Citizen Development Academy. Secondary source: verify against the Recon Analytics original before publication if preferred. 

Download the 
Whitepaper

nexian.co.uk/legal-modern-workplace-whitepaper

Or get in touch with the team if you would like to talk through where your firm is on its technology journey.