An AI strategy for a professional services firm starts with what your people already do with AI, not with a tool list. Your staff are almost certainly using AI today, in ways management cannot see. Map that use first, pick two or three use cases with a clear review step, name who verifies the output, and measure results from the first month. A structured, anonymous staff survey that ends in a scored report and a 90-day plan, like the Calibre Forum readiness report, gives you that map before you spend on anything else.
This guide is for owners, managers, IT leads and planners at small to mid-sized Australian firms: accounting, legal, consulting, advisory, engineering and similar practices.
Calibreforumcalibreforum.auBuild Your AI Strategy on Clear InsightCalibre Forum’s readiness assessment maps how your team uses AI and provides a scored report with a clear 90-day action plan.VisitTL;DR
- Adoption is already widespread. Generative AI use in professional services organisations nearly doubled to 40% in 2026, from 22% a year earlier, so assume your staff are using it.
- Strategy begins with visibility: who uses which tool, for which task, with which client data. Ask every staff member, anonymously.
- Choose use cases with structured inputs, clear rules for good output and a named human reviewer. Meeting summaries, proposals and internal document summaries are the usual first wins.
- Verification is a professional duty that cannot be delegated. Name the responsible person, trace every citation to its source and keep a record.
- Set measures before rollout. Only 18% of professionals say their organisation tracks AI return on investment.
Table of Contents
- Why strategy comes after visibility
- What firms are doing and expecting
- Step 1: map current use across the whole firm
- Step 2: choose use cases that earn their place
- Step 3: put verification in the workflow
- Step 4: set rules, owners and a lightweight governance structure
- Step 5: measure what changes
- A 90-day plan in outline
- Our perspective
- Get your firm’s starting point
- FAQ
- Sources
Why strategy comes after visibility
Most AI strategy advice reads like a technology plan: choose a platform, run a pilot, scale. For a firm of 10 to 200 people, the problem sits earlier. Partners approve a few tools, while staff quietly use others to draft emails, summarise documents and tidy spreadsheets. Management sees neither the benefit nor the risk.
The National AI Centre, the federal body that publishes Australian guidance on AI adoption, says successful adoption is driven by people and team culture, and that meaningful consultation with workers is the best way to build trust and confidence. It also advises checking data accuracy and privacy, intellectual property and cybersecurity obligations, and whether your workflows can adapt to AI.
A strategy written without that picture is a guess. A strategy written with it can answer practical questions:
- Which tasks are staff already speeding up with AI, and which tools do they use?
- Where could client data be leaving the firm?
- Which teams are confident, and which are avoiding AI altogether?
- Which workflows are repetitive enough to redesign?
The answers are only honest if staff feel safe giving them. If people fear being reprimanded for using an unapproved tool, they will say they use nothing. That is why anonymity matters. Our guide on running an AI staff survey honestly and anonymously covers how to word the questions and how to explain to staff what happens with the results.
What firms are doing and expecting
The numbers give you context for the pace of your competitors, and for what clients will soon expect.
- Thomson Reuters’ 2026 AI in Professional Services Report, which surveyed more than 1,500 professionals, found generative AI use at 40% of organisations, with only 19% having no plans to adopt. More than 80% of current users use it weekly.
- NAB’s 2026 survey of around 100 Australian professional services firms found two-thirds expect AI to have a large or very large influence on operations, mainly as a productivity enhancer rather than a replacement for professional judgement.
- SPI Research reports that 45% of clients expect partners to use AI in delivery, and that firms with over 60% AI proficiency see EBITDA above 23%. That research draws on professional services firms more broadly, so treat it as a direction rather than a local benchmark.
The pattern is consistent: using AI is now normal, and using it well is the differentiator. Efficiency alone does not set a firm apart when everyone has it. What sets you apart is how AI fits your service model, your pricing and your client relationships.
Step 1: map current use across the whole firm
Before choosing anything, build a picture of reality. Three things need recording.
Tools. The National AI Centre advises keeping an up-to-date register of all AI systems used. Include the tools nobody approved. A register built only from what IT bought will be incomplete.
Workflows. The same guidance suggests starting process mapping with a single process that is high-volume, painful, data-rich and contained. For each step, record the action, the responsible person, the role or system, inputs, outputs and time. For a firm, think of client onboarding, monthly reporting, proposal drafting or a standard advice letter.
Data. Note which client information goes into which tool. This tells you where privacy, confidentiality and intellectual property exposure sits.

You can do this with interviews, but interviews of a 60-person firm take weeks and produce guarded answers. Our assessment surveys every staff member anonymously about their AI tool use, workflows and integration. It installs no software and collects no individual data, and it produces one scored, organisation-wide report. The report maps capabilities, identifies risks and sets out a tailored 90-day action plan with operational and security recommendations. The report page shows what it contains.
Step 2: choose use cases that earn their place
The search results for this topic are full of long use-case lists. A firm needs a short list, ranked. Use these filters:
- Structured inputs. The task starts from documents, transcripts or forms, not from a blank page.
- Clear rules for good output. A reviewer can tell quickly whether the result is right.
- Low tolerance for error, with human review. The higher the stakes, the more review the task gets, not less AI-free time by default.
- A measurable time or quality gain. You can state the before and after.
SPI Research advises starting with use cases that work, not those that sound innovative: proposal development, meeting summaries, project management office activities and risk analysis are the quickest wins that produce measurable returns and build team confidence. It suggests judging each initiative against four goals: sustainable growth, talent strength, client experience and organisational profit.
Here is how that looks in practice across typical firm types.
| Use case | Typical firm | Reviewer checks | Measure |
|---|---|---|---|
| Meeting transcription and summary | All | Actions and decisions match what was said | Minutes of admin time saved per meeting |
| Proposal and engagement letter drafting | Consulting, accounting, legal | Scope, fees and terms against firm templates | Hours from brief to sent proposal |
| Internal document summarisation | Legal, advisory, engineering | Summary against source sections | Reading time per document |
| Client onboarding data capture | Accounting, financial advice | Extracted fields against originals | Onboarding days, error rate |
| Research synthesis | Consulting, advisory | Every claim traced to a source | Research hours per deliverable |
Note what is missing: anything where AI output goes straight to a client unreviewed. Keep it that way at the start.
Your staff survey data is useful here. If twelve people independently use AI to summarise meetings, you have found a use case that already has demand, and a pilot will meet little resistance. The 90-day plan in our report turns that kind of finding into a sequence of actions for your firm.
Step 3: put verification in the workflow
AI tools produce confident text that can be wrong: invented citations, wrong figures, misread documents. In professional services the error lands on a client, so verification belongs in the process, not in good intentions.
Thomson Reuters’ guide for law firm partners sets four steps that apply well beyond law:
- Use only approved tools.
- Name the person responsible for verification. For lawyers, the responsible lawyer’s duty cannot be delegated.
- Trace every citation to its primary source, such as the judgment or legislation, and read it to confirm it supports the proposition. Do not use another AI tool for this check.
- Keep a record of the tools, outputs and checks.
The Legal Practitioners’ Liability Committee adds that principals remain responsible for supervising and reviewing all client work, including AI-assisted documents. Checking citations alone is not enough: the authority must support the intended proposition, still be current and fit the client’s circumstances. It also points to firm policies, regular training and monitoring for unauthorised tool use.
Accountants, consultants and engineers can adapt the same logic. Every number in a deliverable traces to a source document. Every regulatory reference is checked against the current official text. Every AI-assisted piece of work has a named reviewer who signs off.
Two practical habits help. First, label AI-assisted drafts internally so reviewers know to look harder. Second, record what went wrong. A short log of caught errors teaches the team where each tool fails, and it feeds directly into your AI risk register. Our guide to generative AI governance covers the policy side.
Step 4: set rules, owners and a lightweight governance structure
A small firm does not need a bureaucracy, but it needs four things written down:
- An AI policy stating approved tools, what client data may be entered and what must be reviewed. See what to include in an AI policy in Australia.
- An owner. One person, often an operations manager or IT lead, keeps the tool register current. A small group, even three people, can act as an AI governance committee.
- Data rules. Which client information never goes into a public tool. Our AI data governance guide lays out the basics.
- A security baseline. Account controls, vendor checks and access. The AI security checklist is a starting point.
On regulation, Australia currently has no AI-specific statute for businesses. Existing laws such as the Privacy Act and the Australian Consumer Law still apply to how you use AI, and some Privacy Act changes affect disclosure about automated decisions. Our overview of AI regulations in Australia explains what applies to your business. The Australian Institute of Company Directors’ guide stresses accountability, planning, risk measurement and management, transparency, testing and human control, which is a sound checklist for a partner group.
Step 5: measure what changes
The weakest part of most firms’ AI efforts is measurement. Thomson Reuters found only 18% of professionals say their organisation tracks AI return on investment, and another 40% do not know whether anyone does. Firms are spending without knowing the result.
The AICD advises boards that ROI assessments should focus on specific, measurable use cases, and notes that more value is realised where AI transforms processes, systems or business models, yet relatively few organisations have moved beyond the pilot phase.
For each use case, set four measures before you start:
- Time: hours per task before and after.
- Quality: review corrections per document, client rework requests.
- Client value: turnaround time, client feedback, capacity to take new work.
- Risk: errors caught at review, data-handling incidents, unapproved tool use.
If you bill by the hour, faster work can reduce revenue unless pricing changes. That is a business-model decision, and it belongs in the strategy: fixed fees, value-based pricing or redeploying freed hours to advisory work. A practical guide to measuring AI automation ROI can help you structure the business case.
Repeat the measurement. Because staff behaviour changes as tools improve, a one-off snapshot ages quickly. Our annual membership runs the assessment quarterly, so you can see whether adoption, confidence and risk are moving the way your plan intended.
A 90-day plan in outline
A structured start keeps the project from drifting. This is a sensible shape, similar to what a tailored plan contains. For another view of a first quarter, see this 90 day plan to embed AI in professional services.
Days 1 to 30: see and decide.
- Survey all staff anonymously on tool use, workflows and confidence.
- Build the tool register and note where client data is involved.
- Choose two or three use cases using the filters above.
- Write a one-page AI policy and name an owner.
Days 31 to 60: pilot with review.
- Run each use case with a small team, a named reviewer and baseline measures.
- Train the pilot team on verification, including how to trace sources.
- Log errors and near misses.
Days 61 to 90: decide and extend.
- Compare results to baselines and keep, adjust or drop each use case.
- Update the policy and risk register from what you learned.
- Brief all staff on what is approved and why.
- Plan the next quarter, including any agentic tools.
On that last point, Thomson Reuters reports only 15% currently use agentic AI, though 53% are planning or considering it. Agentic AI means tools that carry out multi-step tasks with some autonomy. These raise the stakes for review and access control, so they come after the basics are working.
Our perspective
We built Calibre Forum because the first question most firms cannot answer is the simplest: what are our people already doing with AI?
Partners tend to assume one of two things. Either “we don’t really use it yet” or “we have it under control”. Surveys of the whole staff usually show a third picture, with a handful of enthusiastic users, a few hidden risks and a large group unsure what is allowed.
That is why our assessment asks every staff member, anonymously, and installs nothing. People answer honestly when they know nobody is being singled out, and honest answers are the only ones that produce a useful strategy. The result is a scored report that shows where the firm is strong and where it is exposed, and a 90-day plan that gives management concrete next steps rather than a generic framework.
Our view is that a firm should know its starting point before it chooses tools, writes policy or sets targets. Strategy built on that map is shorter, cheaper and far more likely to be followed.
Anthony
Get your firm’s starting point
If you want the map before the strategy, our assessment is built for this. Staff complete a short, anonymous survey on how they use AI tools and where they work. You receive a combined, scored report that outlines your existing capabilities, risks and opportunities, together with a tailored 90-day action plan. The one-off report costs A$1,990. If you want to track progress, membership at A$2,388 per year provides recurring quarterly surveys and reports.

Firms with a specialised profile can also see how we apply the assessment to similar organisations, such as engineering consultancies.
FAQ
What should be the first step in an AI strategy for a professional services firm?
Find out what your staff already use AI for, and with which client data. An anonymous, firm-wide survey is the quickest way to get honest answers. From that you can pick use cases, write a policy and set measures.
Which AI use cases should a firm start with?
Start where inputs are structured, good output is easy to recognise and a human reviews the result. Meeting summaries, proposal drafting, internal document summaries and onboarding data capture are common first choices. Leave unreviewed client-facing output until you have a track record.
How do we stop AI hallucinations reaching clients?
Make verification a named responsibility, not a habit. Use approved tools, trace every citation or figure to its primary source, and keep a record of the tools, outputs and checks. Regulators and professional bodies are clear that the responsible professional remains accountable for the work.
How do we know whether AI is paying off?
Set baselines before the pilot: time per task, review corrections, turnaround and errors caught. Compare after 60 to 90 days and keep only the use cases that show a gain. Repeating the measurement every quarter shows whether the gains hold.
Sources
- 2026 AI in Professional Services Report, Thomson Reuters
- Professional Services Performance Era 2026, NAB
- Impact of AI on Professional Services, SPI Research
- Get ready for AI, National AI Centre
- Map your processes, National AI Centre
- How to verify legal AI: guide for law firm partners, Thomson Reuters
- Supervising AI, Legal Practitioners’ Liability Committee
- A Director’s Guide to AI Governance, AICD
- Measuring AI automation ROI
- 90 day plan to embed AI in professional services
Recommended
- AI Adoption Playbook for Australian Businesses: A 90-Day Plan
- AI Policy for Law Firms Australia and NZ
- AI Readiness Assessment for Accounting Firms
- AI Readiness Assessment for Marketing Agencies, AU and NZ
General information only. Articles on this blog are written with the help of AI writing tools and published by Calibre Forum. They are not legal, insurance, tax or financial advice, so check anything important with a qualified adviser.