Largence Custo, built on Largence's knowledge graph

Most legal teams have tried a general-purpose assistant by now. It drafts quickly. It sounds confident. Then someone asks which section of CAMA 2020 the answer relied on, and the thread goes quiet.

That gap is not a prompt engineering problem. It is a context problem. A model that does not know your matter, your documents, or the statute your firm actually works under is guessing in polite prose. We built Largence Custo because practitioners deserve something closer to a junior who has already read the file, not a stranger at the other end of a chat window.

Matter-aware, not a chatbot

Custo lives inside the matter workspace. It inherits the same access controls as the rest of Largence, so it sees the chronology, the evidence index, the drafts in progress, and the correspondence your team has already logged. When you ask it to summarise a position, it is summarising your position, grounded in records your firm already trusts.

That distinction sounds subtle until you use it on a live transaction. The question is no longer "write me a clause about board composition." It becomes "given what we already agreed with the investor last Tuesday, what still needs to change before signing?" Custo can answer the second question because it is not starting from zero every time you open a new tab.

Built on Largence's knowledge graph

Underneath Custo is the same verified knowledge graph that powers Generate, Research, and Legal Agents. Largence maintains a living map of in-force Nigerian and UK statute. When Custo cites CAMA 2020 or the Companies Act 2006, the reference is checked against authoritative sources. If the system cannot verify a section, it says so plainly rather than inventing one that merely sounds right.

This is why we describe Custo as built on the knowledge graph, not merely connected to a model. The graph is what turns a fluent answer into a defensible one. Practitioners should not have to choose between speed and something they would put their name to.

A co-counsel that cannot show its working is not co-counsel. It is a liability dressed up as productivity.

Tools, with judgement left to you

Custo is not limited to conversation. It can call more than fifty platform tools: pull a document version, check a compliance flag, draft an update to a chronology entry, or prepare a research memo scoped to the matter in front of you. Tool activity is visible in the thread, and sensitive actions wait for human confirmation before they run.

We designed it that way on purpose. Legal work is full of steps that are mechanical once you know what needs doing, but consequential if done wrong. Custo is there to remove the friction, not the responsibility. You remain the counsel of record. It remains an assistant that knows when to stop and ask.

Three surfaces, one standard

Custo, Agents, and Generate are three distinct products. Generate produces structured documents with citations baked in. Agents run longer autonomous workflows on Vertex and Zenith, with deliverables routed to an approval inbox. Custo sits in the middle of daily practice: the colleague you message when you need an answer, a second read, or a nudge through the next task without leaving the matter.

They share one rule. Every surface validates citations against the same graph, respects the same matter boundaries, and writes to the same audit trail. A firm should not have to re-learn what "trustworthy" means depending on which button it clicked.

What we are watching for

Custo is live for firms in our soft launch cohort. We are paying close attention to where practitioners still have to explain context the system should already have, and where tool suggestions feel helpful versus noisy. That feedback is shaping what we ship next.

If you are already working in Largence, open a matter and start a Custo thread. Ask it something you would normally route to a busy associate at 9pm. See whether it answers like someone who has read the file.

If you are not in the cohort yet, view the newsroom for more on how we are rolling the platform out, or speak to us about access.

The Largence team