Contract Clause Review
First-pass triage of an inbound contract against your firm's playbook — every clause classified, the risky ones flagged with a suggested redline, before a lawyer opens the document.
It reads each clause, matches it to your firm's preferred position, and flags the gaps with a draft redline for a legal practitioner to accept, edit or reject.
Where this works well
The slow, invisible problem this makes visible is the first pass — the hour or two a lawyer spends just locating the clauses that matter in an inbound contract before any real thinking starts. On a 47-clause professional services agreement, most clauses are boilerplate that match your firm's standard accepted position; the value is buried in the four or five that don't. This pattern earns its keep by doing that locating mechanically: it reads every clause, matches it to the playbook, and surfaces the indemnity that says "without limitation", the reverse IP assignment, the 30-day termination notice below your 60-day minimum — so the practitioner opens the document already pointed at the risk.
It helps most in a firm or in-house team with a high volume of broadly similar contracts — services agreements, supply agreements, NDAs — and a settled house view on the common clauses. The reviewer who today triages a stack of inbound contracts by hand gets that triage as a starting draft, and redirects the recaptured hours to negotiation and the genuinely novel clauses.
Where it works badly
It works badly where your positions aren't actually settled. If three partners would mark up the same indemnity three different ways, the tool has no single "preferred position" to measure against, and its flags become noise. It also works badly on bespoke, heavily negotiated, or unusual agreements — a one-off joint-venture deed, a contract drafted in an idiom it hasn't seen — where a clause can be mis-classified, or a real risk scored green because it doesn't pattern-match to anything in the playbook.
The dangerous failure mode is a false green: a clause that matches the playbook on its face but is dangerous in *this* deal because of the counterparty, the dollar value, or how it interacts with another clause two pages away. The tool rates clauses one at a time against a standard; it does not weigh the contract as a whole.
The honest test: write down how your firm would mark up an uncapped indemnity, a reverse IP assignment, and a short termination notice. If three reviewers produce three different answers, fix that first — the tool will only amplify the inconsistency.
What it doesn't do — and shouldn't
It does not give legal advice and it does not approve anything. It surfaces where a clause departs from your playbook and drafts a suggested redline; a person decides whether that redline is right for the deal, edits it, or throws it out. The classification ("red — uncapped indemnity") is a prompt for the reviewer, never a clearance for the clause.
That boundary is deliberate and it is not optional. Reviewing a contract and advising on it is reserved legal work in Australia — only an admitted practitioner holding a current practising certificate may do it, and clients who rely on unqualified work lose the protection of professional indemnity cover. So the tool stops at "here is what looks off and a draft of how we'd fix it"; the practitioner owns the advice, the negotiation, and the signature. This is where "we accelerate the thinking, we do not replace the thinker" is literal: it accelerates the locating, never the deciding.
What your data has to look like for this to work
The thing most firms underestimate is the playbook. For the tool to flag a clause, your preferred position on that clause type has to be written down in a form it can read: not "we usually push back on broad indemnities", but "indemnities capped at 2x annual fees, consequential loss excluded, mutual where possible" — a position, a fallback, and the rationale. The same for liability caps, IP ownership, termination notice periods, data-retention and deletion obligations, and governing law. The demo flags a 30-day termination notice precisely because the playbook records a 60-day minimum; with no recorded minimum, there is nothing to flag against.
The contracts themselves need to arrive as machine-readable text — a clean Word document or a well-structured PDF, not a scanned image or a fax — and clauses need to be numbered and titled consistently enough to be located. Most firms have the contracts in reasonable shape but the playbook only half-written, sitting across partners' memories, old marked-up precedents, and a few checklist documents. Pulling that into one explicit, owned, current set of positions is usually the real first job — a matter of capturing what your senior people already know, not buying a new tool — and it is the part we help with. It is almost always bigger and more valuable than the AI layer that sits on top of it.
Could it tell us a dangerous clause is fine, or push us to accept something we shouldn't?
It can, which is exactly why it never has the last word. The tool classifies each clause and rates it green, amber or red against your playbook — but a green rating only means "matches our standard accepted position", not "safe in this deal". Bespoke or unusual drafting it hasn't seen before can be mis-classified or scored too low. Every flag and every non-flag is a prompt for the reviewing practitioner, not a clearance. The safeguard is that the redline only leaves the firm after a person with a practising certificate has signed off.
Our playbook lives in three partners' heads and a folder of old marked-up Word docs. Will this work?
Not until the playbook is written down. The tool can only flag a 30-day termination clause as short if it knows your minimum is 60 days, and can only call an uncapped indemnity risky if your preferred position (say, capped at 2x annual fees, consequential loss excluded) is recorded somewhere it can read. Turning tacit partner judgement and scattered precedents into an explicit, current set of preferred positions and fallbacks is the real first job — and usually the more valuable one. It is the work we help with before any AI layer goes near a contract.
Does this replace our lawyers or our contract reviewers?
No. Giving legal advice and reviewing contracts is reserved legal work in Australia — only an admitted practitioner holding a current practising certificate can do it, and unqualified practice carries real penalties. The tool does the mechanical first pass — locating every clause, matching it to the playbook, drafting the redline — so the practitioner spends their time on judgement, negotiation and the clauses that genuinely need a human. It recaptures reviewing capacity; it does not remove the reviewer.
How current does our playbook have to be for the flags to be trustworthy?
As current as your actual negotiating position. If the firm quietly moved its liability cap or started accepting a clause type it used to reject, but the playbook still says the old thing, the tool will confidently flag deals against a stale standard — or wave through something you'd now push back on. The playbook needs an owner and a review cadence, the same as any precedent bank. Stale positions are the most common cause of wrong flags.
Where does our contract data go, and is it exposed to a third party?
Contracts are commercially sensitive and often privileged, so this matters. The clause text and playbook are sent to the language model that does the classification; which model, where it runs, and what its data-retention and training terms are, are decisions we scope with you up front, including options that keep data within agreed boundaries. We design the data path so privilege and confidentiality obligations are respected — that is part of the build, not an afterthought.
It read a 60-page contract in seconds — how do I know it actually found everything?
You don't take that on faith. The tool reports how many clauses it parsed and which it flagged, and the reviewer works from the full document, not just the summary — the flagged-clause list is a starting point for the human, not a substitute for reading. If the contract is a scanned PDF, badly OCR'd, or structured in a way the parser misreads, clauses can be missed entirely, which is why the practitioner stays responsible for the document as a whole.
Estimated build: 3–5 weeks. Most of it is template work we've already done.
Fixed scope, fixed price, fixed dates.
The cost band reflects the engagement shape, not a per-feature line item. We work on fixed scope, fixed price, fixed dates — see the services catalogue for what falls inside each band.
Considering this for your org?
The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where it hurts; we’ll play it back, scope it, and show you what’s possible.