thirdbrain.

What's possible with AI in legal and compliance

Evidence current to mid-2026.

In legal and compliance, AI is ready today for the drafting and the volume work, like first-draft contracts from templates, policy documents, summarising large sets and extracting obligations from contracts, always with a lawyer checking and owning the output. The advice and the final call, unverified court submissions, privileged material in unapproved tools, and employment decisions stay with people. The tables below show exactly which is which.

Each task below sits in one of three bands: a strong fit today, worth a careful pilot, or keep with people for now.

A clear breakdown of where AI does and does not fit across the core tasks of a legal and compliance team. A practical starting point, not the last word.


Where AI is mature: strong fit today

Proven and available today. AI does the bulk of the work and a lawyer reviews it.

FunctionThe job todayWith AI
Contract drafting from templatesTurn a brief or template into a first-draft non-disclosure agreement (NDA), services agreement or employment contractAI drafts from the template and your instructions; a lawyer checks the terms and owns the final words
Policy and procedure draftingWrite or update internal policies, compliance frameworks and codes of conductAI drafts from the brief and your existing policies; a lawyer checks the legal accuracy and signs off
Document summarisingSummarise contracts, pleadings, legislation, regulatory guidance and board packsAI extracts the key terms and issues; a lawyer checks the summary is accurate and complete
Contract data extractionPull key dates, obligations, termination rights and liability caps from a large contract setAI extracts to a structured table; a lawyer spot-checks and owns the obligations register
External counsel briefingDraft instructions to external counsel, matter summaries and chronologiesAI drafts from notes and documents; a lawyer checks the accuracy and owns the brief

Where AI is emerging: consider piloting with a human gate

Promising but not yet proven at this scale. AI assists and a lawyer stays in the loop, so trial it on a contained scope first.

FunctionThe job todayWith AI
Legal research (with citation checks)Find relevant cases and legislation, summarise the position, check whether the law has movedAI finds and summarises; a lawyer independently checks every citation in a primary source before it goes anywhere
Document review and discovery triageClassify a large document set for relevance and privilegeAI triages; a lawyer reviews the privilege calls and owns the final set
Due diligence on standard documentsReview data rooms and extract issues from NDAs, supply agreements and standard contractsAI extracts issues on standard provisions; a lawyer checks novel or negotiated terms
Contract review and redlining (third-party paper)Review a counterparty’s contract against your standard positions, mark up the deviations, propose fallback wordingAI marks up against your playbook and drafts the redlines; a lawyer owns the materiality calls, the negotiation strategy and the unfair-contract-terms risk
Regulatory monitoringTrack changes to the laws and regulations that apply to the businessAI surfaces changes; a lawyer assesses the impact and directs the response
AML/CTF complianceTrack customer identity verification, risk assessments and AUSTRAC reporting obligations under the Anti-Money Laundering and Counter-Terrorism Financing Act 2006 (AML/CTF Act), which applies to legal practices providing designated services (including conveyancing, entity structuring and business financing) from 1 July 2026AI helps compile verification records, draft the compliance programme and surface monitoring alerts; the risk assessment, client due diligence judgement and reporting decisions stay with the practice
Compliance reporting and board packsDraft compliance status reports and board papers from data and notesAI drafts; a lawyer traces every claim, checks the accuracy and owns what goes to the board
Employment matter correspondenceDraft show-cause letters, investigation notices and procedural fairness correspondenceAI drafts from the brief; a lawyer checks the legal basis and the procedural fairness
Privacy impact assessmentsDraft privacy impact assessments (PIAs) and privacy governance documents from a brief of the processing activity, including assessment of exposure under the statutory tort for serious invasions of privacy (in force 10 June 2025)AI drafts from the template; a lawyer checks the privacy analysis, the tort exposure and the obligations

Where AI is not ready or suitable today: keep with people

These stay with people, either because the tooling is not reliable enough or because professional obligations rule it out.

  • Legal advice and opinions. AI can draft a working document; the advice and the final call are a professional judgement that must come from a qualified lawyer. AI is not a lawyer.
  • Court submissions with unverified authorities. A lawyer must independently verify every citation in a primary source before filing. More than 60 published Australian decisions involving AI-generated citations in court filings have been documented through mid-2026, with consequences including formal referral to professional disciplinary bodies and restriction to supervised practice. The rule from every Australian court that has addressed this: independent verification, not AI verification.
  • Privileged material in public AI tools. Public AI tools are incompatible with legal professional privilege. The Federal Court (GPN-AI, April 2026), the Federal Circuit and Family Court of Australia (FCFCOA) (PD-AI, May 2026) and every major Australian professional body have said so. Use only enterprise tools with contractual data isolation.
  • Employment decisions. Hiring, termination, disciplinary findings and workplace investigation outcomes stay with the decision-maker. AI may help draft correspondence; the decision and the procedural fairness steps are human.
  • Signing, filing and lodging. AI does not execute documents or lodge filings. The relevant director, company secretary or authorised signatory signs and lodges.
  • Novel or unsettled law. AI answers confidently from training data that may be outdated or jurisdiction-poor. Independent research found Sydney legal queries hallucinate at 61% in general-purpose AI, with some specialist Australian legislation at 100%.
  • Risk and commercial judgement calls. Whether to settle, what position to take, how hard to push: that is strategy and judgement. AI can draft options; the lawyer or business owner decides.

Worth weighing: for any of these, the upfront setup (building template libraries, defining the clause checklist, connecting the approved enterprise tools with the right contractual controls) is a one-off, separate from the ongoing effort, which on the strong-fit items is mostly verification and review. The time reclaimed from first-pass drafting and document reading goes back to the advice, the strategy and the client relationships that require a qualified lawyer’s judgement: exactly where the value lies and where AI is not a substitute.

This guide sits on top of the things that stay the same whatever your function. See the ground rules

Questions

The questions leaders ask.

Will AI replace my legal team?

No, it changes what they spend time on. The drafting, the document reading and the obligations tracking shrink, and their judgement, their advice and their accountability matter more. The compliance-critical work, the advice itself, employment decisions, court submissions and anything that requires a professional to sign off, stays firmly with a qualified lawyer or responsible officer. The goal is a legal function that does more with the same people, not fewer.

Is it safe to put confidential client information into AI?

Not into a public AI tool, and every major Australian professional body has said so clearly. Public AI chatbots are incompatible with legal professional privilege: uploading privileged matter material to a tool that may retain or train on inputs risks inadvertent waiver. The safe path is an enterprise-licensed tool with contractual data isolation, Australian data residency where available, and a ban on model training on your inputs. The December 2024 joint statement from the NSW Law Society, the Victorian Legal Services Board and the Legal Practice Board WA is explicit on this point.

Where should we start?

Pick something high-volume, easy to check and verifiable against the original. Document summarising is a common first step: AI reads the document and extracts the key terms, a lawyer checks the output against the source. Contract data extraction is the other strong one, pulling obligations, dates and key clauses from a large contract set into a register. Both remove real busywork and both have a natural verification step built in. Prove one before adding the next.

Can we use AI for legal research and court submissions?

With strict rules, and with much more caution than most people assume. AI legal research tools hallucinate at meaningful rates on Australian law. An independent peer-reviewed study found Sydney legal queries hallucinate at 61% in general-purpose AI, with some specialist legislation at 100%. The dedicated legal research platforms perform better, but still need independent verification. More than 60 published Australian decisions involving AI-generated citations submitted to courts have been documented through mid-2026, with consequences ranging from formal referral to a professional disciplinary body through to restriction to supervised practice. Every citation from any AI tool must be independently verified in a primary source before it goes into any document filed with a court.

What can AI do in legal now that it could not six months ago?

Quite a bit on both sides of the ledger, which is why this is dated. Through early 2026, Australian courts got serious: the Federal Court issued binding guidance on AI use in proceedings (GPN-AI, April 2026) and the Federal Circuit and Family Court followed (PD-AI, May 2026), covering virtually all major federal litigation. Harvey opened a Sydney office and Australian law firms began deploying legal-specific AI for contract review and research at scale. At the same time Australia accumulated its first disciplinary findings against practitioners for AI-generated fake citations, making the risks concrete rather than theoretical. The capabilities and the obligations both grew. A year ago this was distant guidance; today it is current professional obligation.

How current is this, and what if the rules have moved on?

The evidence is current to mid-2026 and we refresh it as the tools and the obligations change, which in legal and compliance they do often. Where a finding comes from an independent source we lead with it; where it comes from a vendor we treat it as marketing. If something has shifted since you read this, the fastest way to get the current picture for your own function is a quick chat.

Book a discovery call

A first conversation

See how this applies in your world.

Book a confidential chat. Tell us where you see the opportunity, in legal and compliance or anywhere else, and we will agree a sensible next step. No deck, no pitch.