We trained their advisor. Now she builds the firm's own AI tools.

Who are AustAsia Group?

A family-owned advisory firm in West Perth. They look after individuals, growing and established businesses, and retirees, across tax, accounting, finance, wealth and legal matters.

What was slowing the work down?

Manual steps around the AI they already used. The firm had Claude set up for its teams, but files had to be added by hand, and regular practice reports were pulled and processed manually.

Preparing a client's financials also passes through several hands, as it does in most firms: one person gathers the information, another drafts the working papers, and a manager reviews everything and checks that it balances. Each hand-off is a point where work waits.

What did we do?

We trained one of their business and tax advisors to build her own tools with Claude Code, the tool for building apps and automations with AI. We set it up on her computer, connected her everyday apps, and supported her as she built.

As the work moved towards real client data, we brought in the firm's IT team. Together we agreed where each tool should run and what rules it had to follow, and we stayed on hand with follow-up sessions while she kept building.

What did she build?

Two things: a client onboarding form, and a set of AI agents that share the work of preparing client financials.

• A web form where a new client enters their personal details, income and assets, and uploads their documents
• On submit, the form creates a PDF summary and files it with the client's details and documents
• One AI agent that gathers all of a client's information
• One that starts the draft
• One that reviews everything before a person signs off

Why does the form run on the firm's own server?

Because client documents are sensitive. We agreed with the firm's IT team that the form should run inside their own environment, under their control, rather than on an outside service. Their IT team deployed it at the firm's own web address.

Submissions follow the firm's existing rules for client documents: where they are stored, who can open them and how long they are kept. That was a requirement from the start, and it is the kind of decision that belongs with the firm. The form is in testing with dummy data while the firm completes its security checks before clients use it.

How do the agents share the work?

Through shared folders. Each agent drops its finished work into a folder, and a new file there starts the next one, so work moves from gathering to drafting to review without anyone passing it along.

We gave her a build guide so each agent starts the moment a file arrives, instead of checking on a timer. Nothing waits, and nothing runs when there is no new work.

What has it achieved?

The firm now builds its own AI tools in house, on its own terms. Client onboarding has a proper front door ready for clients once testing finishes, and the gather, draft and review steps of preparing financials are shared with AI, with a person still reviewing the result.

The next connections are under way: linking the agents end to end, having each new form submission create the client in the firm's CRM, and connecting practice reporting so it no longer needs pulling by hand.

AI for accounting firms in Australia

Accounting work suits AI well because so much of it is documents moving between people: client records, source documents, working papers, reviews and correspondence. AI is good at the gathering and first drafts, which frees qualified staff for judgement and client advice.

Common first wins

Client onboarding forms that collect details and documents in one go. Pulling a client's information together before a job starts. First drafts of working papers, summaries and client emails. Turning regular practice reports into something people can act on without re-keying.

Client data and privacy

Firms hold sensitive personal and financial information, so the Australian Privacy Principles apply in plain terms: know what you collect and why, where it is stored, who can see it and how long you keep it. Before connecting any AI tool, check where it keeps data, keep sensitive tools inside your own environment where you can, and involve your IT team early.

Keep a person in charge

AI can gather, draft and check, but a qualified person should review anything that goes to a client or a regulator. Start with one job, connect the software you already use rather than replacing it, and expand once the first tool has earned the team's trust.

If your team already uses Claude, our Claude training covers the step from projects to apps and automations. If you would rather we build it for you, see document processing.