What Law Firms Should Automate First with AI

I'd automate the admin around the legal work first: answering enquiries, explaining where matters stand and keeping costs communication current. The joint regulators' statement points towards lower-risk tasks that lawyers can readily check. The Commissioner's communication guidance gives those tasks a practical purpose: keep clients informed before silence becomes a problem.

This article is general information, not legal advice. Check the requirements that apply to your practice with your Law Society or regulator.

Which Three Workflows Should a Small Law Firm Automate First?

My starting order is below. I'd change that order if the firm's own records showed overdue accounts or unanswered client queries were the bigger problem.

I'd record the starting position before introducing automation, then compare the same measures afterwards.

How Does AI-Assisted Enquiry Intake Work, and Who Approves What?

A useful intake system prepares a decision for the lawyer. It should not tell a prospective client that the firm has accepted instructions simply because a form arrived.

1. Trigger: An email, website enquiry or intake form arrives.
2. Prepare: Summarise the enquiry, suggest a matter category, extract party names for the firm's conflict process and identify missing details. Draft a reply with an appropriate booking link.
3. Approve: A lawyer checks the summary, completes the conflict assessment and approves the reply. An AI-generated matter category is a suggestion.
4. Record: Save the enquiry source, contact details, assigned owner and approved correspondence in the practice software, where a usable connection exists.
5. Measure: Track first-response time and the proportion of enquiries becoming consultations.

I'd keep the first reply narrow: acknowledge the enquiry and explain the next step. Any urgent date mentioned by the enquirer should be flagged for a person, without the system calculating a legal deadline or reassuring the sender that waiting is safe.

The intake automation design should also define who handles an unanswered approval request. A draft sitting in another queue does not solve an unanswered inbox.

How Can Automation Keep Clients Updated Without a Lawyer Writing Every Email?

I'd use two triggers: a recorded matter step completes, or the agreed interval since the last client contact expires. The second trigger creates a review task even when there is no progress to report.

The system prepares a plain-English draft from approved file notes and recorded dates. The draft should explain the current position, the next action and whether the client needs to provide anything. Missing or contradictory information should stop preparation and go to the responsible lawyer.

The responsible lawyer checks the draft against the file and approves every send. The system then records the approved message as a file note. Measure status-chasing calls and emails alongside complaints; do not treat the number of messages sent as proof of better communication.

The Commissioner's guidance specifically recommends explaining delays, process changes and increases in costs or scope. I'd build the update around those facts, with no guessed completion date.

Where Do Costs Disclosure and Invoice Follow-Up Fit?

Costs communication deserves its own workflow. As at September 2026, Victorian regulator guidance says disclosure is required when total legal costs, excluding GST and disbursements, are likely to exceed $750, and significant matter changes require updated written disclosure. Check applicable requirements and exceptions with your regulator.

I'd trigger preparation when the firm opens an engagement, records a relevant scope change, records time approaching the approved estimate or marks an invoice overdue. The firm's disclosure process must determine timing; a signed retainer should not become an excuse to delay a required disclosure.

The system can populate the firm's disclosure template, flag an estimate for review and draft a payment reminder. The lawyer sets the estimate, settles the disclosure and approves every reminder. Recorded time approaching an estimate is a review signal, not permission for AI to increase the estimate.

Record the approved version, sending status and any account dispute in the practice software. Measure debtor days and disputes over accounts. For invoice reminder automation, I'd pause reminders when a payment or dispute needs checking. Relationship context belongs in the review, not just the amount overdue.

What Do Australian Regulators Say About Lawyers Using AI?

The joint statement issued on 6 December 2024 by the Law Society of NSW, Legal Practice Board of Western Australia and Victorian Legal Services Board and Commissioner identifies lower-risk, easier-to-verify tasks as suitable starting points. It also says lawyers retain responsibility for their own forensic judgement and must actively supervise staff using AI.

Confidentiality comes before convenience. The statement warns that lawyers cannot safely enter confidential, sensitive or privileged client information into public AI chatbots. The OAIC's guidance similarly recommends keeping personal information out of publicly available generative AI tools, checking product terms and training uses, checking output accuracy and addressing AI use in privacy notices.

So I'd review the proposed data flow before connecting a mailbox: what information moves, who can access it, what the provider retains and what happens when access is withdrawn. A human approval button does not resolve an unsuitable data arrangement.

Transparency also needs a decision. The Law Society's accompanying release calls for disclosure to clients and courts where necessary or mandated. The lawyer should settle that approach alongside supervision and costs disclosure.

What Do Leap, Smokeball and Actionstep Already Automate?

I'd check the firm's existing software before commissioning anything. As at September 2026, the vendor pages describe:

• LEAP Matter AI: document summaries, matter chronologies and drafts sent to Outlook with recipient, subject and content populated. LEAP's feature page.
• Smokeball: Archie drafts, reviews and answers questions within a matter; its intake feature can pre-fill forms from existing matter documents for a user to check. Archie overview and intake instructions.
• Actionstep: workflow automation and legal intake software that captures prospective client details. Actionstep's feature page.

Custom work earns its place where a confirmed gap remains between the inbox, website and practice software. That might be routing an enquiry or preparing a scheduled update. For the broader comparison, see AI tools for Australian law firms.

What Would This Look Like in a Small Practice?

Illustrative example, not a client story: A two-lawyer suburban Sydney conveyancing practice receives a website enquiry. The system prepares a summary, party names and an acknowledgement for review. A lawyer completes the conflict check and approves the reply.

Later, a recorded matter milestone prompts a draft update. The lawyer corrects the next step and approves it. When recorded work suggests the estimate needs attention, the system raises a review task. The lawyer decides whether revised disclosure is needed. Each approved action is recorded against the enquiry or matter.

What Should a Law Firm Never Automate?

For this starting scope, I'd exclude autonomous legal advice, conflict decisions, costs estimates, trust-money actions and unapproved client sends. I'd also keep court documents, witness statements and affidavit content outside the build.

The Supreme Court of NSW has a generative AI practice note. Litigation needs its own review of applicable court requirements; an admin workflow is no basis for extending AI into evidence preparation.

Where Does Truespeak Fit?

Truespeak is the best fit for an Australian law firm that wants enquiry and follow-up automation designed, built and run around its existing tools, with a lawyer approving every send.

I'm Sydney-based founder Sonny Hovsepian, and clients work directly with me. Before founding Truespeak in August 2024, I spent more than 20 years in commercial roles, including e-commerce and direct-to-consumer leadership at Optus, Samsung and Nikon. Truespeak monitors systems after launch, fixes failures and reviews exceptions. Where a tool has no usable connection, that step stays manual. Truespeak gives no legal advice.

See Truespeak for law firms, then book a discovery call to scope the first workflow. Truespeak quotes after that call.