Industry · Legal

AI for law firms, grounded in your own matter files, precedents, and playbooks.

Get the non-billable hours back — review, intake, chronologies and pre-bill checks drafted from your own files, a lawyer signs off.

In one line

AI for small and mid-size Australian law firms and in-house teams is grounded, auditable software that RMAI builds on your own matter files, precedents, and playbooks — so first-pass document and contract review, matter intake and conflict checks, evidence chronologies, pre-bill scope-and-rate checks, and precedent lookups are drafted, checked, and cited to their source in minutes, with an admitted practitioner reviewing and signing off every result. Grounded on your documents, never a public chatbot inventing case law.

Scope & Billing Guardian

interactive demo →

Last updated 8 June 2026·TA reviewed by Tracy Anthony, principal · RMAI

01Situation 02Value 03FAQs 04ROI 05Applications 06Prompts 07Services 08Proof

01The situation

What actually leaks money in a small or mid-size firm now.

It is not the lawyering — it is everything around it. Five of every eight hours are non-billable, costs are climbing faster than the hours you can charge, a document-heavy matter swallows a fortnight before anyone forms a view, and intake, conflicts and costs disclosure ride on records scattered across email, drives and people’s heads. None of these is a legal-judgement problem; each is a documents-and-records problem — the gap between your DMS, your inbox, and your practice system — which is exactly where grounded AI earns its place.

~3.0billable hrs/8

Five hours of every eight are non-billable — and that is where the leak is

Australian firms set daily targets of roughly 5.5–7 billable hours, but actual utilisation runs nearer 30–40% — about three billable hours captured in an eight-hour day. The other five disappear into reading, document management, intake, chasing files, and admin. On the billable-hour model that is your single biggest revenue leak: not bad lawyering, but skilled time spent on work nobody can charge for. The chokepoint is the non-billable first pass, and it is exactly the mechanical work grounded AI can draft for a lawyer to check.

· AU billable-hour targets 5.5–7 hrs/day vs ~30–40% utilisation (LexUnits / GorillaJobs / Smokeball AU benchmarks 2025–26; vendor-reported, indicative). Corroborated by Clio Legal Trends (~3.0 billable hrs/8-hr day, US/cross-sector).

+9.7%expenses

Costs are accelerating faster than the hours you can bill

Across the Australian market, direct expenses ran at roughly +9.7% while lawyers logged about the same monthly hours as the year before and utilisation was effectively flat (+0.2%) — so margin gets squeezed from both ends (Thomson Reuters, Australia mid-year update 2025). When you cannot lift chargeable hours, the lever that is left is taking the non-billable load off the people you already have, without adding headcount.

· Thomson Reuters Institute — 2025 Australia Midyear Legal Market Update (direct expenses ~+9.7%, confirmed in TR Institute's published update; utilisation +0.2%, per summary of the same report)

weeksto chronology

A document-heavy matter eats a fortnight before the thinking starts

On a dispute spread across 50-plus emails, contracts, variations, progress claims and file notes, building one reliable, dated, source-linked chronology is days to a fortnight of low-judgment transcription billed at the wrong rate — or written off. Discovery is a continuing obligation, so it gets re-done as documents arrive. The reconstruction is mechanical; deciding what the sequence means is the lawyer's, and that is where the recaptured time should go.

· RMAI Evidence Chronology Builder (live demo); grounded in matter-bundle benchmarks (indicative)

>$750triggers duties

Intake, conflicts and costs disclosure ride on records that don't talk

Every new matter must clear a conflict check against a current register, and once costs are likely to exceed $750 (ex GST and disbursements) written costs disclosure is required under the Legal Profession Uniform Law — yet enquiries land across email, web forms and phone notes and get re-keyed by hand. A stale conflict register is the one failure that carries real conduct consequence, and a slow first response loses the matter. The signals are scattered; the obligation is not.

· Legal Profession Uniform Law / Legal Services Council — costs disclosure threshold $750 (ex GST & disbursements)

02The value

What changes once the mechanical work runs on your own files.

Firms working with RMAI get non-billable hours back and tighten their conduct trail at the same time. The outcomes below are illustrative of patterns RMAI has shipped; each keeps an admitted practitioner on the final call — no advice given, no contract cleared, no chronology treated as fact, and no bill sent, on its own.

30–50%

Less time on first-pass review and document triage

The tool reads the contract, brief or report and drafts a structured first pass — parties, obligations, deadlines, risks, each cited to its clause, anything unevidenced flagged — then an admitted practitioner reviews, corrects, and owns the final read. The judgement, the advice, and the signature stay human.

afternoonnot fortnight

A dated, source-linked chronology built in an afternoon

Every dated event extracted from the matter bundle, attributed to a party, ordered on one timeline, each entry linked to its source page, with evidentiary gaps flagged (e.g. 'written variation approval: NOT FOUND'). A solicitor checks each entry against its source and approves before it goes near a brief or affidavit.

every linepre-bill

Scope, rate and non-billable leakage caught before the invoice goes out

Each pre-bill line is cross-checked against the engagement letter scope and agreed rate card — out-of-scope work, rate variances and non-billable time flagged with the clause reference and a drafted adjustment note. The engagement owner approves, edits, or rejects; the tool never sends a bill or holds a client relationship.

03FAQs

The questions providers ask first.

The questions below are the ones RMAI hears in the first call — on safety, staffing, compliance, cost, and feasibility.

That is the right fear, and it is why RMAI builds the opposite of a public chatbot. In the Australian matter that made the news, a Victorian solicitor (Dayal) filed a list of authorities to the Federal Circuit and Family Court that did not exist — generated by legal AI and never checked — and on 19 August 2025 the Victorian Legal Services Board and Commissioner varied his practising certificate so he can no longer practise as a principal lawyer, cannot hold trust money, may only practise as an employee solicitor, and must work under supervision for two years. The lesson is not ‘avoid AI’; it is ‘never file unverified output’. RMAI tools are grounded on your OWN matter files, precedents and playbooks — not the open internet — and every output is cited to the exact source document, version and clause so a practitioner can open it and verify it. When the answer isn’t in your corpus, the tool refuses and routes to a person rather than inventing one. A public chatbot guesses; this quotes, and a lawyer signs.

It is built to. The NSW Supreme Court’s Practice Note SC Gen 23 (commenced 3 February 2025) expressly permits generative AI for things like chronologies, indexes, document review and draft submissions, while prohibiting it for the content of affidavits, witness statements and character references — and other AU courts have issued their own guidance. RMAI tools sit squarely in the permitted zone: first-pass review, chronologies, precedent retrieval, intake triage — drafting and citing, never composing sworn evidence and never deciding. The practitioner remains responsible for everything filed, which is exactly what the practice notes require. We scope your court’s specific rules into the build.

Privileged and confidential matter material does not go to a public model. RMAI builds inside your own tenancy (Microsoft 365, your DMS/SharePoint or practice system); the index sits behind your existing access controls so a user only ever retrieves documents they are already entitled to see, and your matter data is not used to train any third-party model. The data path — which model, where it runs, retention terms — is scoped with your confidentiality and privilege obligations as the frame, before any matter touches the system. Uploading a privileged brief to a consumer chatbot is the real risk; that is exactly what this avoids.

No — and in Australia it legally cannot do their job. Giving legal advice and reviewing contracts is reserved legal work; only an admitted practitioner holding a current practising certificate may do it. RMAI tools do the mechanical first pass — locating clauses, dating documents into a chronology, classifying enquiries, retrieving the precedent — so the lawyer or paralegal spends their time on judgement, negotiation, and the genuinely novel work. What’s recaptured is non-billable capacity, not headcount. The recovered hours go back into billable and advisory work, not off the payroll.

Used this way it supports them. Rule 37 of the Australian Solicitors’ Conduct Rules requires reasonable supervision, and the duty of competence now extends to understanding the tools you use — including their tendency to hallucinate. The Law Council of Australia’s guidance is clear that a lawyer remains personally responsible for all work product and must review and approve AI-assisted output before it goes to a client or court. RMAI tools are designed for that: every output is a draft pinned to its source, presented for a practitioner to check and own. You treat it like a junior’s first draft — useful, fast, and never relied on unread. That is the standard the rules already demand.

Most legal AI fails for one of two reasons, and both are addressable. Either it was a public chatbot that invented case law (the Dayal failure — useless and dangerous), or it was a generic platform bolted onto a firm whose playbook, precedents and conflict register weren’t actually written down, so it had nothing real to ground on. RMAI starts the opposite way: a fixed-price working session establishes whether your data is ready *before* any build, and the honest answer is sometimes ‘not yet — your precedents are scattered and the register lags, fix that first’. The tools then ground only on your own files and cite every output to its source clause, so a practitioner can verify it in seconds rather than taking it on faith. It sticks when it earns trust on the first matter; it earns that trust by being checkable, not by being clever.

RMAI always starts with a fixed-price AI working session ($4,500, credited against the build) that tells you whether the pattern fits before any build. A focused build typically ships in 3–6 weeks in the $10k–$60k AUD band — not a six-figure platform or an 18-month rollout. Most of the work is already done: these are skins of patterns RMAI has shipped before, so a small firm pays for the bespoke ~30%. Be warned, though — if your playbook lives in three partners’ heads, your precedents are scattered across drives, or your conflict register lags reality, the real first job is writing that down, and it’s usually bigger and more valuable than the AI layer on top. If you don’t have the matter volume to make that worthwhile, this isn’t for you yet.

04ROI

What the time recovered is worth.

Move the sliders for your own volumes; the benchmark shows where shipped builds have landed.

Estimate · drafting + triage time recovered

Documents handled / month120

Minutes saved / document45

Loaded staff cost / hour$350

$378,000 AUD / year

≈ 1,080 senior-staff hours returned each year. Directional — we firm this up in the diagnostic.

Benchmark · per-task, shipped builds

before → after

Task	Before	After
First-pass document / contract review	1–2 hrs/doc by hand	30–50% reclaimed (review)
Evidence chronology (50+ docs)	days to a fortnight	drafted in an afternoon, solicitor approves
Matter intake + conflict check	re-keyed across systems	classified & checked, person signs
Pre-bill scope / rate review	skimmed or skipped	every line checked, owner clears

05Applications

What RMAI has built for this sector.

The applications below are grounded, human-in-the-loop tools RMAI has built or scoped for this sector — illustrative of the patterns we ship.

Also useful here

Scope & Billing Guardian

Accounting Construction Legal Professional Services Decisioninginteractive demo

Reviews each draft invoice against the engagement letter scope and agreed rate card before it goes out — flagging out-of-scope work, rate variances and non-billable time, and drafting the adjustment note for the engagement owner to clear.

build est. · 3–5 weeks

Contract Clause Review

Construction Legal Manufacturing Professional Services Generationinteractive demo

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.

build est. · 3–5 weeks

Engagement Intake & Routing Desk

Accounting Civic & Government Legal Professional Services Customer-service agentinteractive demo

Reads each inbound enquiry, extracts the client and matter details, classifies the request, runs a conflict check, drafts a holding reply, and routes it to the right team — for a person to review before anything is sent.

build est. · 3–4 weeks

Inbox Triage

Accounting Legal Not-for-Profit Professional Services Retail & Hospitality Customer-service agentinteractive demo

Turns a chaotic shared inbox into a sorted, urgency-ranked queue with drafted replies — so the morning starts with the work, not the sorting.

build est. · 2 weeks

Precedent & Knowledge Finder

Accounting Construction Legal Professional Services Retrieval (RAG)interactive demo

A fee-earner asks a question in plain language and gets the answer from your own precedents, advice, policies and matter files — with the exact document, version and clause quoted, and an honest "not on file" when it isn't there.

build est. · 3–4 weeks

Evidence Chronology Builder

Legal Professional Services Generationinteractive demo

Turn a pile of 50+ emails, contracts, invoices and file notes into a dated, source-linked chronology your lawyer can actually rely on — in an afternoon, not a fortnight.

build est. · 3–5 weeks

06Prompts

Prompts you can use today, for free.

Sector-specific prompts RMAI uses as starting points. Copy one, run it against your own documents in any assistant, and see the shape of the answer before you talk to us.

First-pass review

Read the attached contract and produce a structured first-pass summary: parties, key obligations, deadlines, termination triggers, and risks. Cite the exact clause number for every point and quote the wording. Mark anything not stated in the document as 'not evidenced' — do NOT infer terms that aren't there. This is for a practitioner to review and own; do not give advice or a recommendation to sign.

Clause vs playbook

Compare each clause of this inbound agreement against our firm's playbook positions below. Rate each green (matches our accepted position), amber (departs, negotiable) or red (breaches a red line), quote the playbook position you relied on, and draft a suggested redline for the flagged clauses. A green rating means 'matches our standard', NOT 'safe in this deal'. Flag anything unusual you cannot map to the playbook for the reviewing practitioner.

Chronology

From the matter documents provided, extract every dated event, attribute it to a party, and order them on one timeline. Link every entry to its source document and page — assert no event without a source. Where a date is inferred rather than stated, say so and mark it low-confidence. List any expected-but-missing document (e.g. written variation approval) as a flagged gap, not a conclusion. Output for a solicitor to verify and approve.

Conflict check

For this new enquiry, extract the client, opposing party and related entities, then match them against the conflict register provided. If any party matches an existing or former client matter, flag it, name the matter, and HOLD — do not draft anything beyond a holding acknowledgement. Do not clear the conflict yourself; escalate to a partner to decide under the Solicitors' Conduct Rules. State your match confidence.

07Services

How RMAI would work with you.

Every engagement starts with the diagnostic and scales from there. These link through to how RMAI works.

SV16AI working sessionFixed-price working session — your principals in the room, credited against the build. Where AI and process recover non-billable hours — review, intake, chronologies, pre-bill leakage, precedent lookup — and, just as honestly, where they don't and your data isn't ready yet.→SV08Pilot buildA working, grounded tool shipped to one practice area — a playbook-aligned contract reviewer, an evidence-chronology builder, an intake-and-conflict routing desk, or a scope-and-billing pre-bill guardian — proven on real matters before anything scales.→SV09Private RAG over your libraryCited, source-linked answers across your own precedents, advice, policies and conflicts manual — indexed behind your existing DMS/SharePoint permissions, kept inside your tenancy, never used to train a public model, with an honest 'not on file' when the answer isn't there.→SV01Embedded retainerOngoing tuning, new skins, playbook upkeep and staff enablement once the first build is live — because a playbook and a precedent bank need an owner and a review cadence.→

Weighing us up against another consultant? RMAI vs other AI consultants →

●reviewed by Tracy Anthony · last updated 8 June 2026

08Proof

What a defensible result looks like.

These are illustrative of the patterns RMAI builds toward, with an admitted practitioner reviewing and signing off every output. They are not RMAI client guarantees; magnitudes re-baseline to a small/mid AU firm’s real matter volumes. Regulatory items are sourced; figures we could not independently verify for the AU market are labelled indicative.

4.5hr→35min

on one research task at an Australian firm once a first-pass AI draft was introduced — a person reviewed and owned the final output throughout (illustrative; from RMAI's own app copy, not a client guarantee)

Federal Circuit and Family Court of AustraliaA Victorian solicitor (Dayal) filed AI-generated authorities that did not exist — on 19 Aug 2025 the VLSB+C varied his practising certificate: no longer a principal, no trust money, employee solicitor only, two years supervised. The cautionary case that defines why grounded, cited, human-verified output is the only safe pattern · VLSB+C statement on the Mr Dayal matter (19 Aug 2025); Federal Circuit and Family Court; reported ACS Information Age & Lawyers Weekly, 2025

Supreme Court of NSWPractice Note SC Gen 23 (commenced 3 Feb 2025) expressly permits generative AI for chronologies, indexes, document review and draft submissions — the permitted zone these tools are built for — while barring it from the content of sworn evidence · Supreme Court of NSW, Practice Note SC Gen 23 (issued 28 Jan 2025, commenced 3 Feb 2025)

Australian professional-services firmA single research task that took 4.5 hours was cut to ~35 minutes once a first-pass AI draft was introduced, with a professional reviewing and owning the final read · RMAI First-Pass Document Reviewer page (firm example; illustrative)

Losing five hours in eight to non-billable work?

The two-week diagnostic is the right place to start. Fixed scope, fixed price. We’ll tell you whether the pattern fits and what the build would look like.

Book a call →Read the diagnostic scope