AI Discussion Takeaways for 2027

At the start of 2026, most conversations about legal AI were about getting started. A year on, the question has changed. Firms have bought the tools and run the pilots - so is it actually working?

For many firms, the honest answer is "we're not sure." Our State of Legal AI in 2026 report found that 52% of legal professionals aren't confident AI is delivering value, and only 14% are very confident that it is.

We took those findings to two webinar panels of knowledge, research and information leaders this year: Fragmented, Ungoverned, Unconfident in August and Value Before Cuts in October. The panelists and questions were different, but the same points kept coming up.

Are We Mistaking Adoption for Progress?

Both panels agreed that firms have moved quickly on buying tools and much more slowly on the work that helps those tools pay off. As James Grandage of Hill Dickinson put it, "don't confuse adoption with progress."

Experimentation still has its place, especially when people are trying to fix a real workflow problem. But as Courtney Toiaivao of Grey Letter Advisory pointed out, "experimentation is not a strategy." Without a clear idea of what good looks like, it's very hard to know whether you're getting anywhere.

AI Is Shining a Light on Old Problems

Many people hoped AI would fix fragmented knowledge. So far, it has mostly made the problem more visible. In our report, 47% of respondents described their knowledge systems as fragmented, and 83.2% of lawyers said they face friction finding information.

Our panelists pointed to a few reasons:

• New AI tools are being added alongside existing research subscriptions, not replacing them
• Almost every platform now has its own built-in AI, so it's harder to know which tool to use for what
• AI can make poorly organized data look polished, which makes problems harder to spot

Connected knowledge has been the goal of knowledge management for a long time. AI has just made it much harder to ignore.

Governance Needs to Keep Up With Behavior

Our report found that 41% of firms don't have clear AI governance, and around 42% don't know which AI tools their lawyers are using. When we polled our August audience, 66% said they could tell us "roughly" which tools were in use.

"Roughly" might be fine day to day, but it's not an answer you'd want to give a client or a regulator. And clients are asking. Emily Florio of DLA Piper described how her firm's list of client AI "limitations" has grown into "limitations and obligations."

The panel's advice was to build governance around principles rather than individual tools, much like most firms already do for internet use. "Check your citations" and "don't put client data into unapproved tools" stay relevant whatever launches next.

Where Do People Fit In?

This came up in both sessions. AI can get you most of the way there, but a person still needs to check the output and apply their judgment - especially when a hallucination looks convincing.

That makes training and talent planning critical:

• Make training regular. Occasional blocks of training leave long gaps where questions pile up. A short monthly session can work better.
• Protect time to learn. It's hard to get up to speed on AI on top of a full workload.
• Think about the pipeline. Only 9.4% of respondents listed losing junior talent as a top AI concern. Cutting graduate and intern roles now could leave firms short of experienced people in five years' time.

As Courtney put it, "AI replaces work. Only the firm decides whether it replaces people."

Who Should Own It?

65.6% of respondents told us their firm isn't ready, or they aren't sure it's ready, to scale AI effectively and safely. Part of the problem is ownership. When a committee owns AI adoption, nobody really does.

Our panelists recommended one clear owner, supported by a cross-functional group from IT, knowledge, research and the wider business. Knowledge and research teams are well placed to help lead this, as skills like information architecture and knowledge management are exactly what good AI use depends on.

Questions to Ask Going Into 2027

• What problem is each AI tool meant to solve, and how will we know it's working?
• Is our data organized and connected enough for AI to build on?
• Does our AI policy reflect how people actually work?
• Are we training people regularly, and protecting the talent pipeline?
• Who owns AI adoption at our firm, and is our knowledge team involved?

As we head into 2027, the firms that see real value from AI will likely be the ones that take the time to get the foundations right - their data, their governance and, most of all, their people.

Watch the webinar recordings here and you can read the full report here.

Alternatively, book a meeting with one of our team to discuss how we can support you to answer these questions.