Most firms think they've "done" AI once the licence is signed. The really hard work starts after.

AI Adoption in Law Firms: A Tale of Misalignment

I've lost count of the number of lawyer conversations that start the same way: "we've rolled out [insert AI tool], but usage is limited - people don’t trust it."*

I’ve spoken with scores of lawyers in the UK, Australia and New Zealand and it tends to go like this:

• leadership signs the enterprise deal
• The vendor runs training sessions and sometimes they have internal AI demos / workshops
• Some people adopt the technology and do cool things with it
• Folks quietly assume the firm has modernised.

In my experience, choosing the tool is roughly 20% of the project. The other 80% is structural change management and cultural buy-in, and most law firm AI training doesn’t work for this.

1: Why Buying an AI Tool Isn't the Same as AI Adoption

One partner at a boutique firm I spoke to was - like most others - using Copilot. They found hallucinations slip through easily, and it's genuinely hard to know at what point in the workflow the AI was even used before the final answer lands on the partner's desk.

I keep thinking of gym memberships: buying the membership and turning up once is not the same as getting fit. The tool being present in the tech stack tells you almost nothing about whether it changed how anyone actually works.

2: Billable Hours and AI don’t mix

This is one that comes up constantly whether I ask or not. Most firms still measure and pay associates on hours billed. If AI turns a four-hour contract review into a one-hour job, the associate is cutting their main metric. Who would want to do that?

Lawyers will not adopt a tool that quietly shrinks their own targets, unless their KPIs get realigned - which may also mean revisiting their commercial model.

3: Workflow Friction: Why Lawyers Don’t Use The Tools They’re Given

Lawyers run on highly established, practice-specific routines, and “track changes” in Word is a default behaviour. "Why would we do this?" was the actual reaction of one in-house counsel to being asked to manually upload documents into a separate AI portal on top of an already-full SharePoint setup.

If a tool requires leaving tools you’re familiar with to go and use a bolted on alternative, usage drops off fast. Its not necessarily laziness, by the way - or even a fear of innovation. Its the fear of making a mistake - and in my opinion, that fear is entirely justified.

4: AI Hallucinations and Trust: The Real Barrier to Law Firm AI Adoption

Lawyers are trained, structurally, to avoid risk. Everything must have limits, caveats and justifications. When a tool's failure mode is "confidently wrong," you get exactly the reaction you'd predict. Roughly 77% of UK lawyers say they're concerned about inaccurate AI outputs.

The Pinsent Masons case is the one everyone in the industry now references:

• a junior used Harvey, ignored the tool's own verification warning (like many AI users)
• AI-driven factual errors reached a court
• the judge was “astonished” (in a bad way)
• The firm was embarrassed and self-referred to the SRA.

In fairness to Pinsent Masons, there are hundreds of other similar cases globally. Notwithstanding, those are the cases we know about.

Without mandatory, easy-to-use verification, fee earners simply revert to doing it manually to protect themselves. Which, from their side, is entirely rational.

What Does Good Look Like?

Firms that get real usage out of AI tend to address the following:

• Commercial realignment: targets and KPIs reviewed so that billable hours aren’t the only yardstick. The market is starting to do this anyway, btw.
• A little change management goes a long way: In an ideal world, you buy or build tools that fit your team's workflows perfectly. That's not realistic - sometimes the tool everyone loves isn't suitable (security, say), sometimes the tool just isn't there yet. The realistic middle ground is a bit of process redesign and change management, so people don't have to adjust too much. There’s strong data to back this, btw.
• AI Governance tailored for law firms: plain-English protocols mapping actual legal risk against perceived risk, so fee earners know when they're covered, not just told to "be careful."
• AI Verification workflow for lawyers: structured checkpoints on the highest-value outputs, not blanket manual review of everything, which just recreates the four-hour job you were trying to shorte

—

We work on all of the above and more at Monboard. Drop us a ping if you’d like to learn more.

*Yes, I’m aware some tools have better adoption rates than others

**Often staff have a different view on what works best. This may be valid, but it don’t rule out “I just like this tool more because I’m familiar with it.”