A semantic layer, in plain language.
Max is a legal semantic layer. It sits above AI models, inside the tools lawyers already use, and turns scattered AI interactions into continuous, governed legal workflows.
It is not an assistant, not a chatbot, not a platform to open. It is the layer that gives legal AI a memory, a methodology, and a chain of accountability.
The semantic layer is the architectural idea that lets a complex system speak the language of those who use it.
Thirty years ago, Business Objects introduced a semantic layer above enterprise databases so business users could query their own data without going through IT. The category that emerged was business intelligence, and the idea has been quietly load-bearing in enterprise software ever since.
Max brings the same idea to legal work. It sits above AI models and translates their raw power into the language of the firm: matters, positions, validation chains, methodology.
No.
Assistants live in the same layer as the models. They answer one question at a time, they restart from scratch at every session, and they compete with each other for the lawyer's attention.
Max lives one layer above. It orchestrates the assistants, holds the memory they cannot hold, governs the work they cannot govern alone.
An assistant gives an answer. Max carries a matter.
Max is not designed to replace the AI tools a firm already uses. It is designed to make them work together.
A firm with a drafting assistant, a research tool, and a review platform has three tools that do not talk to each other. Max sits above them, gives them shared memory, applies the firm's methodology across them, and tracks who did what. The investment already made in those tools is preserved. What changes is the layer that holds them together.
Max does not use a model. It orchestrates several.
One model drafts. A second, which did not draft, contests. A third stage confronts the result with the official sources of the applicable law. This is the four-eyes principle: a controller who shares the perspective of the controlled does not see what the latter has missed.
When a model surpasses the others, it enters the orchestration.
The value of Max depends on no model. It depends on the layer that governs them.
Every new model strengthens Max rather than threatening it.
Because Max is not tied to any single model, a better model simply means a better engine to orchestrate. The semantic layer above is unchanged. The memory it holds, the methodology it has learned, the governance it enforces, all of it remains exactly where it was and benefits from the improved engine underneath.
This independence can be checked: on September 3, 2026, during the simultaneous unavailability of three consumer assistants, Max kept working by drawing on three distinct providers.
A new interface is a new place lawyers have to go, a new tool they have to learn, a new tab competing for their attention.
Max was deliberately built without one. It operates from inside the tools lawyers already use, doing its work without asking to be noticed.
The user does not visit Max. Max arrives where the user already is.
Max supports the entire legal work of the firm, across drafting, review, research, analysis, synthesis, and multi-document reasoning.
But the question of which tasks misses the architectural point. Max is not a tool for tasks, it is a layer for matters. Its value lies less in any single task than in the coherence it maintains across all of them.
A matter is not a session. It is an institution in miniature.
Max holds the matter as the primary unit. It remembers the trade-offs made three weeks ago, the positions taken by a partner now on holiday, the version of the clause that was settled and should not be reopened.
When a colleague picks up the matter, the matter is still itself.
Yes. Every reasoning step is grounded in underlying sources, prior work, or documented methodology, and the same reasoning can be surfaced as a comment, a summary, a table, or a full document.
This is not added as a compliance layer. It is a structural property of how Max is built. A semantic layer that cannot explain itself is not a layer, it is a black box.
Max integrates into Outlook, Word, SharePoint, and the rest of the stack a firm already operates. There is no migration, no parallel platform, no adoption curve weighing on the team.
The deployment is technical, not behavioural. Lawyers keep working in the environment they already know, and Max arrives in the background.
As Max operates inside a firm, it structures the methodology, the trade-offs, the positions, and the standards that emerge from the firm's actual work. This knowledge remains strictly private and remains the firm's property.
Over time, it becomes an accumulated asset that no other vendor and no new model can rebuild. It is the firm's institutional memory, encoded into its own semantic layer.
The longer Max operates inside a firm, the more specific it becomes to that firm. The more specific it becomes, the harder it is to replace, not because of any contractual lock-in, but because the accumulation itself is the value.
No.
Max makes legal judgment more explicit, more reusable, and more traceable. But the judgment itself remains where it has always been, with the lawyer. Final positions, final decisions, and final responsibility never leave the practitioner.
Max carries the architecture. The lawyer carries the matter.
Security is not a layer added to Max. It is a property of how Max was built.
Because Max has no separate platform, the firm's data does not have to leave the firm's environment to be useful. Because Max does not train the underlying models on client data, there is no propagation of confidential information across users.
A defence around a risk is always weaker than the absence of the risk.
Data remains within environments controlled by the firm, hosted within the European Union, and the entire processing operates under the GDPR and the emerging requirements of the EU AI Act.
Depending on deployment choices, Max can run on-premise, in a dedicated isolated environment, or in another secure configuration agreed with the firm. Sovereignty is preserved by design, not by promise.
No. There is no cross-firm learning, no shared knowledge pool, no propagation of client information between users. Each firm's data, documents, and accumulated experience remain strictly isolated within its own environment.
Confidentiality, in Max, is not a contractual commitment that could be modified. It is an architectural property that cannot be undone without changing the product itself.
Write to us: we answer every one, and those that come up most often end up joining this page.
Contact us