Firms sell judgement. Most of the week goes elsewhere.
until the audit finds work worth doing. Confidentiality obligations mean we scope where client material may go before we scope what an agent does with it.
Assembling documents, chasing clients for information, keying data between systems, and writing up what was already decided. That is where the billable week goes, and it is the work partners cannot delegate to anyone junior enough to make the numbers work.
Accounting · Financial and wealth · Insurance · Mortgage broking · Property and real estate
What we hear from firms.
The professional services firm in our case studies said all four in the first meeting.
Client chasing
Half the delay in any job is waiting for a document the client has and has not sent. Somebody senior ends up doing the chasing.
Keying between systems
Practice management, the ledger, the CRM, the lender portal, the document store. None of them agree on what a client is.
Writing up the decided
Advice letters, file notes, statements of advice and workpapers that restate a decision already made in the meeting.
Shadow AI
Staff pasting client material into public chatbots because nobody gave them a sanctioned tool or a policy.
First use cases, with the number we scope to.
Each is scoped in the free audit with a target, a named owner and a human review point. The targets are what we set; the audit puts your own number on them.
Client information collection
Requests sent, followed up and reconciled against the checklist automatically, across email and SMS, with the file flagged complete or the gap named.
Workpapers and advice drafts
File notes, workpapers and advice letters drafted from the meeting and the file, with every figure traced to its source, for the adviser to review and sign.
Intake and onboarding
New client details captured once and written to practice management, the ledger and the CRM, with identity and AML checks queued for a person.
The rules that apply.
Client confidentiality is the constraint everything else sits inside. This is what we check each use case against. It is not legal advice; it is the list your professional body and your auditor will ask about.
- 01Privacy Act 1988 and the Australian Privacy Principles. Client personal and financial information, including where it may be sent and which models may process it.
- 02Tax Practitioners Board Code of Professional Conduct. Confidentiality and competence obligations for registered tax and BAS agents apply to any tool that touches a client file.
- 03APES 110 Code of Ethics. Confidentiality, professional competence and due care for CPA Australia and CA ANZ members, whatever drafted the workpaper.
- 04Corporations Act, AFS and credit licensing. Best-interests duties and record-keeping for advisers and brokers, and ASIC RG 271 on complaints, shape what an agent may say to a client and what must be kept.
- 05APRA CPS 230 and CPS 234. For APRA-regulated entities, operational risk and information security standards apply to any AI service provider in the chain.
- 06AML/CTF Act. Customer due diligence and reporting obligations, extended to accountants, real estate and other services from 1 July 2026. Identity checks stay with a person.
Guardrails that matter most here
- 03Data governed: client material never goes to a model you have not approved, and the policy says so in writing.
- 09On the record: every drafted document traces to its sources and its reviewer.
- 05Human in control: advice, identity checks and anything with a licence obligation behind it is reviewed by the licensed person.
How it ran at a 150-person firm.
A partnership under board pressure to have an answer on AI. Partners disagreed on direction, staff were pasting client material into public chatbots, and a full-time Chief AI Officer was neither findable nor affordable.
Fractional AI leadership, two days a month. The roadmap came from the audit findings; a firm-wide AI use policy grounded in the firm's confidentiality obligations went out in month one; build-versus-buy calls were made on three capabilities; the board got a quarterly report.
Board-endorsed roadmap in month one, unsanctioned chatbot use ended, two systems in production within six months including a research assistant on the firm's own knowledge base. Read the full case.
Questions.
Can client material go into an AI model at all?
Yes, under a policy that names which models, where they run, what may be sent and what may not. The audit maps that for each use case. Public chatbots with no such policy are how firms get into trouble; a governed tool is how they get out of it.
Who is responsible for a drafted advice letter?
The licensed adviser who signs it, exactly as now. The agent drafts from the file and shows its sources; the review and the decision stay with the person whose name is on the licence.
Will this work with our practice management system?
Practice management, ledgers, CRMs and document stores connect through the Digital Brain via their APIs. If your data lives in it, it can be wired once and used by every workflow.
Where do we start?
With the free audit: a readiness scorecard, your three highest-return use cases with a number on each, and a 90-day roadmap. Three hours of your time.
Can we use AI without breaching client confidentiality?
Yes, if the system is built for it. Client data stays inside your controlled environment, agents respect matter-level permissions, and nothing trains a public model. That is a very different proposition from staff pasting into free chatbots, which is usually what is already happening.
Where do firms see payback first?
Document drafting and client chasing. Both are high-volume, low-judgement, and measured easily in hours returned. Proven value there funds the more ambitious work.