Ask why 87% of law firms use AI while only 14% trust it's working, and you don't get a tidy answer. You get a genuinely useful conversation.
That's what came out of our live panel on Wednesday, August 12 2026, with James Grandage (Hill Dickinson), Emily Florio (DLA Piper), and Marijah Sroczynski (Morrison Foerster) for Fragmented, Ungoverned, Unconfident: The State of Legal AI in 2026. Here's what they actually said.
"This isn't entirely new in law firms," said James Grandage, pointing back two decades to the last must-have legal tech wave. "Most firms were investing because every other firm was investing, even if they hadn't done a proper needs analysis." Emily Florio agreed the pattern is repeating: "For the first several years, it was keeping up with the Joneses. Now there's more scrutiny around what the benefit actually is," scrutiny arriving into budgets approved before anyone had a way to measure ROI.
Our report found 47% describe their firm's knowledge as fragmented, and 83.2% face friction just finding information. Emily's diagnosis: new AI tools tend to sit alongside existing systems, not replace them. "Unless all of those are truly connected, you're always going to have a fragmented knowledge system." James went further, landing the panel's sharpest line: "AI is just polishing what is, under the hood, data that's not very structured. It makes it more obvious how bad the data is." Rather than fixing fragmentation, AI can end up dressing it up in a confident, well-formatted answer.
Even in a perfectly connected world, would legal professionals trust the answers? "Do I know that this information is right? Is it up to date?" Emily asked. "I'm not sure confidence is going to follow just because that problem magically got fixed." James agreed, adding the panel's most human moment: his 17-year-old son's enthusiastic but unverified work-experience reports were his mental model for AI output, "really useful, gets us to a place very quickly," but "if I was to use that and send it on, I'd be absolutely insane." AI gets you to 80%, he said. "It's not all the way."
Our report found 43% of legal professionals aren't confident they know what AI tools their own lawyers use day to day. We polled the live audience: 66% said "roughly," 30% said "yes, precisely," 4% said no, close to our original finding. James wasn't convinced "roughly" was good enough: "If a regulator or a client asks what happened and why, 'roughly' isn't really an appropriate answer." Marijah's reaction to the confident 30% was the sharpest line of the session: "I'd love to dig into what makes you so confident, because with as much data as I have, I'm never confident they're not doing something I can't see."
41% of our respondents say their firm lacks clear AI governance. Emily argued a policy still earns its keep: "At least if you have a policy, there's something to tie behavior back to." Marijah offered a different take on whether governance can keep pace: "Governance should be bigger and broader than the specific tools... those are big-picture policies that can be enforced, and won't change just because the tools change." Client pressure is accelerating this, Emily added. Her firm's list of AI-related client requirements has grown from "limitations" to "limitations and obligations," with no sign of slowing.
This is the short version. The full 45 minutes includes the audience Q&A, including a live discussion on whether fixing your firm's data would make AI unnecessary in the first place.
Haven't read the underlying data yet? The full State of AI Readiness in Legal 2026 report is available now.
Media contact: Philippa McIntosh - philippa@vable.com - +44 (0)20 7422 0059