If you work in a law firm, you've probably had some version of this conversation recently. Someone senior wants to know what the firm is doing with AI. Someone else is quietly using a tool nobody signed off. And somewhere in the middle is you, trying to work out whether any of it is actually safe.
You're not imagining the tension. When we surveyed 160 legal professionals for our State of AI Readiness in Legal 2026 report this summer, one theme kept coming through. People aren't worried that AI won't pay off. They're worried it will go wrong, in public, with their name on it.
We've pulled the sharpest of those findings into a one-page handout, the Legal AI Fear Index. Here's what's in it, and why it matters.
87% of the people we surveyed are using or experimenting with AI in some form. Yet 52% aren't confident it's delivering value.
That's an odd place to be. The tools are in, the budget has been spent, and more than half of people still can't say whether it's working. Interestingly, cost barely registered as a worry, with only 8% picking it. The question has shifted from "is this worth it?" to "what happens when it goes wrong?"
When we asked what concerns people most, the answers were about exposure and personal mistakes:
Far fewer were worried about the things you might expect, like a big investment not paying off (18%) or losing junior talent who expect modern tools (9%). The fear isn't strategic, it's human.
This is the part that should give everyone pause, because the data backs those worries up.
Plus, 66% say their firm isn't ready to scale AI safely, or they aren't sure if they're ready.
You might expect the most AI-active firms, the ones led by knowledge and innovation teams with formal governance in place, to be well ahead. They're four points ahead: 37% feel ready, compared with 33% across the wider market. More tools and more policy haven't produced much more confidence.
Every number here points to the same problem. Firms are deploying AI on top of fragmented, ungoverned knowledge, and they know it. Nearly half (47%) describe their knowledge as fragmented across systems, and only 16% say it's fully centralised and structured.
The usual fix is to add another tool. But AI is only as good as the knowledge behind it. If the information feeding it is scattered, out of date or unverified, a new tool just repeats the same problem faster.
Readiness isn't built on tools or policy. It's built on trusted information underneath. That's the layer most firms haven't solved yet. The good news is that it's solvable.
The Fear Index puts all of this on one page, so it's easy to share with a partner, your IT team or whoever owns AI at your firm (assuming someone does).
If you want the full picture, including governance, regulator readiness and the real barriers to scaling, you can download the complete State of AI Readiness in Legal 2026 report.
AI moves fast. Our survey closed in June, and a lot has happened since. So we're running a short survey to find out where the profession thinks AI is heading in 2027, and whether those fears have eased or grown.
We'd love to hear what you think.
It's one minute of your time, and it'll give all of us a better sense of what next year really looks like.