Why context beats the model in the AI race

A lawyer in a firm's Sydney office picks up a cross-border matter, this is a deal with moving parts in London and Singapore. She opens the same AI assistant a lawyer at almost any other firm in the country could open. The model is not the advantage. It never was. What she asks it, and what it's able to answer well, depends entirely on what sits behind & powers it, and that's where firms start to competitively differentiate from one another.

Every firm has access to the same AI

The frontier models moved past being anyone's private advantage a while ago. Copilot, Harvey, Legora, Claude, most firms now have a credible AI assistant available to their lawyers, often more than one. The genuine competitive question was never which model a firm licenses. It's what that model is allowed to know. An AI assistant given a stale draft or an unrepresentative document will answer just as confidently as one given the firm's actual best work. The output sounds identical either way. Only one version is right.

The model doesn't know your firm's work. Context does.

This shows up first as lost time. According to the iManage Knowledge Work Benchmark Report 2026, professionals spend an average of 37 minutes a day simply searching for documents — time spent trying to give an AI tool, or themselves, something worth working from. A model has no memory of which precedent a firm actually trusts, which client relationship a document belongs to, or how recently it was checked. That has to come from somewhere else. Experts decide; the system's job is to make sure they're deciding on something solid, quickly enough that the search barely registers.

That only works if a lawyer can see why an answer came back the way it did, which matter it drew on, whose work it was based on, how current it is. A more capable model doesn't solve this. Only context does.

The gap between firms is a knowledge gap, not a model gap

The knowledge maturity gap between firms is no longer a soft distinction — it shows up in performance. The same Benchmark Report found that organisations at the highest level of knowledge maturity are nearly twice as likely to report year-over-year revenue growth as those at the lowest, and that more than a third have already experienced a documented policy breach traced back to ungoverned AI use. Both firms may be running the same model. Only one of them has done the work to make it trustworthy.

This is the case for treating knowledge as governed infrastructure rather than a filing exercise. iManage Insight+ gives firms a way to organise & capture institutional knowledge — precedent, best practice, know-how, templates, matter history — with enough structure that an AI tool can differentiate between well-reasoned answer from a plausible-sounding one. Firms can define which sets of knowledge an AI assistant is allowed to draw from, so a query about an obligation or a deal structure is answered from what the firm actually knows, not from whatever a lawyer happened to have open. The difference between an AI tool that reflects the firm's best thinking and one that's working from the pile comes down to what it was given to draw from. None of this happens automatically. Every precedent worth trusting was carefully selected, tagged, and kept current by a knowledge team that has usually been doing this work long before anyone talked about AI models at all. That work has always mattered; it just wasn't always visible outside the firm. What's changed is that the same curation now underwrites something bigger than an easier search — it's the actual difference between two firms running the same technology.

Insight+ isn't another model. It's what makes the one you have worth using.

None of this asks a lawyer to give up the AI tool she already prefers. iManage Insight+ isn't built to compete with Copilot, Harvey, or whichever assistant a firm has adopted — it's built to make whichever one they're using worth trusting. The model does the drafting, the searching, the first pass. Insight+ makes sure what it's drawing on is actually the firm's own best work, not a plausible guess standing in for it.

Distance stops being a disadvantage

For firms with legal professionals spread across Sydney, Melbourne, and offices further afield, that matters more than it might elsewhere. Context that's properly governed is reachable by anyone with permission, regardless of which office they sit in, how many years they've been at the firm, or which model happens to be open on their screen. Our Sydney lawyer doesn't need a colleague in London to be awake, and she doesn't need seniority or tenure to get a trustworthy answer. She needs a system built to give one.

The firms pulling ahead on AI are rarely the ones with the most capable model. They're the ones whose knowledge was already well-governed enough to be trusted with a question — and confident enough in the answer to put their name to it. Allens, one of Australia's largest firms, has built exactly this kind of foundation, using curated knowledge collections to underpin its AI strategy across the firm — proof that the advantage was never the model.