The received wisdom on legal AI treats it as a model selection problem. Which tool to buy. How to price it. Whether the partners will adopt it. These are real questions, but they’re downstream of a more fundamental one that most firms aren’t asking.
What does the information layer actually look like?
AI doesn’t create knowledge. It works with the information it has access to. Feed it curated, accurate, current sources and it produces outputs lawyers can rely on. Feed it whatever your lawyers happen to find themselves - sources of variable quality, updated on no particular schedule, accessed differently depending on individual search ability - and it produces outputs that look confident and aren’t.
The model doesn’t create this gap. It exposes it. A fragmented, uneven information foundation was always going to produce fragmented, uneven results. AI just makes the consequences visible faster, and more publicly.
AI is only as good as the knowledge behind it. The information question has to come first.
Every law firm has access to broadly the same AI models. The frontier shifts every few months, but the gap between the best model on the market and the one your firm can buy at enterprise pricing is rarely wide enough to matter at deployment.
What does differ, sharply, is what each firm can put in front of those models.
A firm with centralized, curated sources, automated delivery to the right people, and visibility into what lawyers actually engage with produces AI outputs lawyers can use. A firm with the same model and a fragmented, manually managed information layer produces AI outputs lawyers don’t trust. And the trust gap is very difficult to close once it opens.
Every firm will have the same AI. Not every firm will have the information to make it useful.
The information lawyers work with - and that AI works with - usually has no single owner. It comes from multiple disconnected places: vendor databases, manually subscribed alerts, practice group emails, individual search habits built up over years.
That fragmentation creates three compounding problems. Coverage is inconsistent - what one lawyer monitors, another misses entirely. Delivery is unreliable - updates that matter don’t always reach the people who need them, or reach them too late. And visibility is absent - no one knows what’s actually being read, what’s being ignored, or whether the effort going into distribution is producing any value.
This has always been inefficient. The reason it’s become urgent is that AI amplifies it. AI works with what it has access to. A fragmented, manually managed information layer - with uneven coverage and no engagement signal - produces fragmented outputs. Unlike a missed email, the gap in an AI-generated summary isn’t always obvious to the lawyer reading it.
The information infrastructure is the variable AI has to work with. The quality of what it produces follows directly from the quality of what goes in.
For years, KM has been positioned as a support function: a team that curates content, manages databases, runs the alerts. AI changes that positioning - but not in the direction most people expect.
The KM function isn’t being replaced by AI. It’s becoming the layer that determines whether the firm’s AI strategy works.
When KM professionals curate sources properly - verifying accuracy, ensuring currency, maintaining licensed access - they build the information foundation that AI needs to produce reliable outputs. What AI adds is the ability to make that KM-quality intelligence accessible to every lawyer in the firm, regardless of their search ability. Not just the ones who know how to use a boolean filter.
That’s the combination that changes the capability equation: curated, current intelligence, delivered to everyone. The KM function sets the standard. AI makes it universally accessible. But the AI can only deliver on that promise if the information layer underneath it is good enough to build on.
We made a related argument earlier this year in The AI Reality Check: What You Need to Ask First, looking at the same problem from the question-set side: strategy, skills, governance, technology. The most productive AI conversations we see in firms aren’t about the tools; they’re about the foundations underneath them. Worth reading alongside this piece.
Three things, applied to the information layer before any AI sits on top:
Centralized source monitoring: one system aggregating the sources that matter - curated, licensed, continuously updated - rather than individual lawyers managing their own fragmented alerting in parallel.
Automated, targeted delivery: relevant updates reaching the right lawyers automatically - by practice area, by client, by individual role - without requiring the lawyer to go looking or the team to distribute manually.
Engagement visibility: data on what’s actually being read, by whom, so the team can improve what it sends and the firm can see the value of what its information function produces.
Get those three right and the AI conversation changes. The question stops being “why aren’t the outputs consistent?” and becomes “what can we now do with the infrastructure we have?”.
If you’re evaluating an AI tool right now, or using one and quietly wondering why the outputs vary more than expected, the information question is worth asking first.
Take the AI Information Check. Four questions, instant result. Find out whether your firm’s information layer gives AI something reliable to work with - or where the gaps in currency, access, and reach are already affecting what it produces.
Or book a 15-minute call with our team. We’ll walk you through what a strong information foundation looks like in practice - and where Vable fits in.
Further reading: The AI Reality Check: What You Need to Ask First