The structured deal context your AI needs to become a trusted colleague.
Matrix consolidates every moving part in a deal - every email, draft, side letter and election - into one real-time, single-source-of-truth "Omnigraph" that your existing AI calls into. Overnight, your existing AI stack gains the tirelessness of a junior, the nuance of an associate, and the institutional memory of a partner.
Your legal AI still lacks the matter context to output work above trainee level.
Ask it for a comment memo and you get output at a basic trainee level and requires extensive reworking. You still end up doing the nuanced 80% that matters to the matter.
It forgets deal context
No memory between sessions or between tools. It answers one query, then loses the context - blind to the thousands of clauses, elections and side letters that move together on any live deal.
Low-quality output
It changes the entity name and summarises the obvious. It tells you borrowing is restricted - not how that cascades into recycling, economics, the whole deal. The judgement is still yours.
It can't handle deal-scale data
Real matters are too large for a prompt: hundreds of drafts, email chains, side letters and elections. The AI samples a slice, misses the dependencies, and calls that the answer.
Messy provenance (if any at all)
...that you just have to trust, and hope for the best. Even 99% accuracy - which no one claims - is worthless if you don't know where the wrong 1% sits. So you end up having to check 100%.
Every source, live-updated into one single-source-of-truth Matter Omnigraph that persists between queries.
The live connective tissue of any matter. Matrix ingests and appends every email, draft and signed document (breaking down all clauses and nuances within those clauses) into one live graph, hosted in your region. It's append-only - so it remembers, and never re-reads what it already knows. And voilà, your AI can suddenly (actually) draft that 600-page SPA, properly.
Matrix gives your AI the full matter context - and keeps every answer traceable.
Today, your AI only knows what you upload in that moment - and even then, the answer is impossible to audit, and documents get stale very quickly in a live deal with moving parts. Matrix connects the live matter record: drafts, emails, side letters, elections and prior matters, so it can check context across documents and matters, not just one prompt. Ask again tomorrow and the answer comes from the same source trail, reducing hallucination drift from repeated queries.
We're just getting started: v1 → v2 → v3 all compound from the correct base architecture.
Answers, workflows and firm memory all call on the same immutable record.
Source trails are attached from ingestion, not reconstructed later.
Matter data stays sealed, region-hosted and scoped per user.
Retrieval becomes workflow execution, then institutional memory.
The same schema unlocks client and practice-area insight.
v1: same AI, drastically more usable outputs, with receipts - all from plugging it into the live context.
No new interface for fee-earners to learn. Matrix plugs the Matter Omnigraph into your existing AI stack, so AI output instantly upgrades from trainee work you have to rework, to work you can trust because each answer is tied to the record underneath.
1.4.3 The Fund may invest in any assets or securities as the General Partner sees fit, without limitation as to sector, geography, concentration or asset class and in the best interests of the Fund.
1.4.4 The Limited Partner may transfer all or any portion of its interest to any affiliate or qualified institutional investor with prior written notice to the General Partner.
1.4.3 The Fund may invest in any assets or securities as the General Partner sees fit, without limitation as to sector, geography, concentration or asset class, provided the General Partner considers such investment to be in the best interests of the Fund provided that (i) no single investment shall exceed 15% of Total Commitments, (ii) no investment shall be made in any entity that is publicly listed on a recognised stock exchange at the time of acquisition, (iii) no investment shall be made in any business whose primary activity involves the manufacture or sale of weapons, tobacco or fossil fuel extraction, and (iv) all investments shall be made in accordance with the Investment Policy set out in Schedule 2.
1.4.4 The Limited Partner may shall not transfer all or any portion of its interest in the Fund to any affiliate of the Limited Partner or to any third party that is a qualified institutional investor, without the prior written consent of the General Partner. The General Partner [and such consent] shall not be unreasonably withhold, delay or condition its consent to any other transfer. No transfer fee shall be charged in connection with any transfer withheld, provided that the General Partner may withhold consent in its sole discretion where the proposed transferee is a competitor of the General Partner, the Fund or any Portfolio Company, or where such transfer does not meet the requirements of clause 6.5 of the LPA.
Pulled concentration and restricted-sector guardrails from Ravenstone Growth Fund III LPA clause 5.2 and Helix Continuation Fund II LPA clause 6.1.
Suggested (iv) streamlines investor excuse rights concessions. Schedule 2 has been amended to reflect top 5 excuse or restriction rights from 70% of investor base of prior fund.
Email from Amelie Laurent at 13:26 CET confirms Schedule 2 is the agreed source for investment restrictions and should be cross-referenced here.
2025 private funds benchmark set shows 82% of comparable LPAs preserve reasonable-consent language.
Any process-driven practice area is fair game for us.
Deals are fundamentally Omnigraph-shaped: thousands of documents with clauses each intricately relating to each other, hundreds of counterparties with MFN rights, with every change potentially cascading downstream. Every ontology within the deal economy is on our radar.
v2 is the Deal Engine: one UI end-to-end for the entire deal lifecycle, with Omnigraph-harnessed agents working in-workflow.
Your associates can't keep running billion-pound deals inside a chatbot. Matrix collapses the work into one matter workspace, then uses the graph to pre-empt what is relevant at each stage and surface it - right where the work happens.
Your Desk
Structured from ingestion - so cross-matter and cross-practice intelligence come for free.
A chatbot reads a document and forgets it. Matrix records every term as structured data the moment it lands - so a benchmark isn't a post-hoc reconstruction, it's a straightforward query from the ground up. See exactly where a client sits across fund formation, M&A, finance, project finance and beyond.
Firm-wide is yours from day one. Market intelligence is opt-in - de-identified, aggregated and lagged. The long-term prize is a live client memory across practice areas, because clients do not experience their transactions in neat departmental silos.
3 qualified lawyers and an ontology-obsessed engineer between 2 repeat co-founders: Matrix was always going to exist, it just needed the right team.
Two BAME, female, socially mobile lawyers. One bedroom startup founded 7 years ago that became the multi-award-winning social mobility charity, STRIVE Talent. Now building the intelligence layer that transactional lawyers have needed for decades. Between them: the transactional legal experience and the technical vision to build what nobody else has thought to build in quite this way.
Six years in practice at top fund practices in the City; formation of funds with AUM in the billions is her bread and butter. She has done, manually, everything Matrix is built to replace. Sana does not have a theory about how transactional lawyers work: she is one. She has sat in every closing, negotiated every side letter, and managed every MFN process that Matrix handles. When she says the current workflow is broken, she means she lived it — just last Thursday. She is the domain expert on call who shapes design and schema decisions, tightening the feedback loop from months to minutes.
Chancery-trained barrister. Silver-circle trained solicitor. 42-trained software engineer with systems/architecture depth and product/design instinct. She has acted for Deutsche Bank, Goldman Sachs and Barclays on transactions up to £350M, then as Chief of Staff at a fast-scaling professional services firm drove 40% revenue growth while digitising its entire client base. Her capability as a high-resolution translator between Sana's domain expertise and her varied skills stack means we can execute at breakneck velocity, with crystal clarity of vision, and stay leaner for longer.
Per matter, not per seat. Never more than a fraction of what the work is worth.
You pay for matters we actually run, never for seats you may or may not use. Price follows the depth of the work: lighter when Matrix is the graph layer behind your existing AI; fuller when Matrix becomes the workflow layer running the matter end to end.
No seat tax. A small team and a large team pay for the same thing: the work Matrix is entrusted to run.
The MCP layer is priced as infrastructure. The Deal Engine is priced as operating leverage on the matter itself.
The buyer should feel the economics are obvious: a small slice of the value created, with upside left on the table.
The pilot costs nothing. Full stop. Sign as a design partner and lock in founder economics before list pricing hardens.
Be a design partner - and feel the difference overnight.
A matter - ideally a few
One live or recently-closed matter is all we need to start - but bring two or three and we'll show you cross-matter analytics from Day 1 of deployment.
A non-binding LOI
A quick email to confirm that you'd like us to become a pilot customer - without committing to a full procurement process.
A domain champion
One person who knows the work who is open to questions on the real fine-grain as we build. They help shape the schema and the workflows, turning months of product iteration into weeks.
You share DMS access (or send us the matter bible)
We build an Omnigraph for each matter, hosted in your region. No engineering time from your side.
Half a day with your engineers
Once your Omnigraph is ready to deploy, your in-house team grants access for us to establish MCP connection between the Omnigraph and your existing tooling (Harvey or otherwise).