KPMG Saved $59,000 on Its Audit Bill. Your Clients Just Learned the Same Trick.

In early 2026, KPMG International demanded a 14% discount from its own auditor, Grant Thornton UK. The reason was simple: KPMG had deployed AI across its operations, and that deployment had reduced the work Grant Thornton needed to perform. The 2025 audit fee came in at $357,000 — down from $416,000 the previous year. A $59,000 reduction, passed straight back to the client.

The explanation KPMG gave was direct: if AI lowers your costs, those savings should flow through to the client. This is precisely the argument KPMG has been making to its own clients for years. Now they have applied it to their own accountant. The irony is not lost on anyone who has sat in a Big Four fee negotiation.

Your clients read the same Financial Review article. They are making the same calculation. The question is no longer whether AI will reshape professional services billing — it is whether your firm will be prepared when the conversation starts.

What KPMG's Move Actually Signals

KPMG's action is not a one-off cost-cutting exercise. It establishes a market expectation: AI-driven efficiency gains are now considered pass-through savings. The precedent matters more than the $59,000.

This reframes the conversation every client will have with their accountant. The question will not be "are you using AI?" It will be: "You are using AI. Why hasn't our fee changed?"

Going Concern captured the dynamic precisely: KPMG has "inadvertently opened a Pandora's box filled with stingy clients." That box is now open. The firms that have an answer — a clear explanation of how AI benefits flow to clients, supported by measurable outcomes — will keep their relationships. The firms that do not will find themselves defending bills against competitors who can demonstrate tangible efficiency gains.

The signal is unambiguous. The largest professional services firm in the world has publicly acknowledged that AI reduces the cost of professional work, and that this reduction should be reflected in fees. Every Managing Partner in Australia now operates in that reality.

The Gap That Will Separate Winners from Losers

The data on AI adoption in accounting reveals a troubling pattern. According to Karbon's 2024 State of AI in Accounting Report, 71% of accountants believe AI will bring major change to the industry. Yet only 25% are actively training their teams on AI tools and workflows.

This is not a technology gap. It is a strategic execution gap. The majority of firms recognise the shift is coming. A minority are doing anything about it.

Thomson Reuters' 2025 Future of Professionals Report sharpens the picture: firms with a formal AI strategy are twice as likely to see revenue growth compared to firms without one. The correlation is not coincidence. Firms that treat AI as a strategic capability make different decisions about process design, talent development, and client value. Those decisions compound.

The gap between awareness and action is where competitive position is determined. The firms that close it over the next 12–18 months will establish operational advantages that are difficult to replicate. The firms that wait will find themselves explaining to clients why their fees have not moved, even as their own costs have.

Microsoft Copilot vs Building Your Own: What Does It Actually Cost?

For most mid-sized firms, the immediate question is practical: what should we actually deploy? The two primary options — Microsoft 365 Copilot and purpose-built AI systems — have fundamentally different cost structures and risk profiles.

	Microsoft 365 Copilot	Purpose-Built AI
Monthly cost per user	~AUD $46	~AUD $8–15 (API costs)
Your client data leaves your system	Yes	No
Configured for accounting workflows	No	Yes
Full audit log (your control)	Partial	Yes
Break-even (20-person firm)	Never (ongoing licence)	~18 months
Compliance defensibility	Limited	Full

Microsoft Copilot Enterprise is priced at $30 USD per user per month (confirmed 2026), which converts to approximately AUD $46 at current rates. For a 20-person firm, that is $11,040 per year in recurring licence fees — indefinitely.

A purpose-built system requires upfront investment — typically AUD $15,000–25,000 for initial development — followed by API usage costs of roughly AUD $200 per month for a firm of similar size. The break-even point arrives around 18 months. After that, the firm owns a system configured for its specific workflows, with full control over data handling and audit logging.

The calculation changes when client data is involved. Copilot indexes content across your entire Microsoft 365 tenant. It does not distinguish between internal administration and confidential client files. Purpose-built systems process only the data you explicitly provide, with no background indexing and no exposure to third-party model training.

For firms where client confidentiality is a professional obligation, this distinction is not technical preference. It is compliance architecture.

How Many Hours Can AI Actually Save an Accounting Firm?

The efficiency gains from AI in professional services are no longer theoretical. Thomson Reuters' 2024 Future of Professionals Report found that AI currently saves professionals 4 hours per week, with projections of 12 hours per week within five years.

Apply this to a 20-person accounting firm. Four hours per week, across 20 staff, over 48 working weeks: 3,840 hours per year. At a blended rate of AUD $80 per hour, that represents AUD $307,200 in annual capacity value.

Some firms are already realising these gains. CPA.com's 2025 AI in Accounting Report notes that some practices have automated over 80% of individual tax return preparation. The time previously spent on data entry and form population is redirected to client advisory and practice development.

The value proposition is straightforward: AI does not replace accountants. It removes the work that prevents accountants from doing what clients actually value. The firms that capture this shift will operate with fundamentally different economics than those that do not.

The Security Risk Most Firms Are Still Ignoring

The IBM Cost of Data Breach Report 2024 puts the global average cost of a data breach at USD $4.88 million — a 10% increase from the previous year and the largest annual rise since the pandemic. In Australia, the Office of the Australian Information Commissioner reported a 25% increase in data breach notifications in 2024, the highest level on record.

Professional services firms face a specific exposure. Copilot's architecture grants it broad access across Microsoft 365 environments. Without careful configuration, it can query engagement files, financial records, and client correspondence. Recent incidents demonstrate the risk is not hypothetical:

February 2026: Microsoft confirmed a Copilot bug (reference CW1226324) that allowed unauthorised access to confidential emails in draft and sent folders.
Late 2025: Security researchers demonstrated that Copilot Studio could access files classified as "High Restricted" on SharePoint without generating audit log entries.

Purpose-built AI systems address this through architectural design rather than configuration. Client data is processed in isolated environments with no cross-client access. Every interaction is logged to the firm's own audit trail. There is no background indexing, no implicit permission inheritance, and no exposure to third-party model training pipelines.

For a firm holding client records under professional obligation, the difference is material. Regulators and professional bodies do not distinguish between accidental and intentional disclosure. The question is whether your firm can demonstrate control over client data — or whether you are relying on a vendor's default settings.

What Does a Purpose-Built Accounting AI Actually Look Like?

A purpose-built system for an accounting firm is not a general-purpose chatbot with a different interface. It is a workflow-specific tool designed around the actual work of practice management.

The architecture is straightforward: client files upload to a secure, isolated environment where they are automatically classified and indexed. Staff query the system in natural language — "what was the GST treatment for the Chen property acquisition?" — and receive answers grounded in the specific client files they have permission to access. The system assists with BAS preparation, flags compliance issues, and drafts email correspondence based on engagement history.

Every interaction is logged. Client data never leaves the firm's controlled environment. The system is configured for accounting workflows rather than adapted from general productivity software.

We have built a working demonstration of exactly this architecture. See it in action →

Three Steps Your Firm Can Take This Quarter

The path from awareness to action does not require a six-month programme. For most mid-sized firms, three focused steps establish the foundation:

1. Audit what AI you're already running

Most firms have Copilot features active in their Microsoft 365 tenant without realising it. Determine what is enabled, what it can access, and whether your current permissions align with client confidentiality requirements. This is a days-long exercise, not a months-long project.

2. Quantify the gap

Calculate the automation potential in your current workflows. How many hours per week does your team spend on tasks that AI could handle — data entry, document review, correspondence drafting? The Thomson Reuters estimate of 4 hours per week is a useful benchmark. Apply your firm's rates to that number. The result is your annual efficiency opportunity.

3. Start small: one workflow, one client segment

Do not attempt firm-wide transformation. Select one high-volume, low-complexity workflow — individual tax return preparation is a common starting point — and implement AI assistance for a single client segment. Measure the results, refine the process, and expand from a proven foundation.

Frequently Asked Questions

How much does it cost to build AI for an accounting firm?

Purpose-built AI systems for mid-sized accounting firms typically require an initial investment of AUD $15,000–25,000 for development and deployment. Ongoing costs consist primarily of API usage, generally AUD $200–400 per month depending on firm size and transaction volume. Compared to Microsoft 365 Copilot at ~AUD $46 per user per month, purpose-built systems reach break-even around 18 months for a 20-person firm.

Should accounting firms use Microsoft Copilot?

Copilot can be appropriate for general productivity tasks — email drafting, meeting summaries, internal research — provided it is properly configured with sensitivity labels and restricted from accessing client files. However, Copilot is not designed for accounting workflows and presents data governance challenges when processing confidential client information. Most firms will find a combination works best: Copilot for internal administration, purpose-built tools for client-facing work.

How is KPMG using AI in auditing?

KPMG has deployed AI across its audit practice to automate document review, risk assessment, and anomaly detection. In early 2026, KPMG demanded a 14% fee reduction from its own auditor (Grant Thornton), citing AI-driven efficiency gains as justification for lower audit costs. This move established a precedent: AI savings are now expected to flow through to clients.

How long does it take to implement AI in an accounting firm?

A pilot implementation targeting a single workflow — such as individual tax return preparation or BAS processing — can be operational within 4–6 weeks. Firm-wide rollout depends on scope but typically extends to 3–6 months. The critical factor is not technology deployment but process definition: knowing exactly which tasks AI will handle, how staff will interact with it, and how quality will be verified.

Is client data safe with accounting AI tools?

Safety depends entirely on architecture. General-purpose tools like Microsoft Copilot process data through Microsoft's infrastructure with limited visibility into how that data is used or retained. Purpose-built systems can be configured to keep client data entirely within the firm's environment, with full audit logging and no third-party access. For firms under professional obligation to protect client confidentiality, the latter architecture is typically required for defensible compliance.

What Happens Next

KPMG's $59,000 discount is not the story. The story is that the largest professional services firm in the world has publicly acknowledged that AI reduces the cost of professional work — and that clients are entitled to those savings.

Every Managing Partner in Australia now operates in that reality. The firms that have measured their AI opportunity, secured their data architecture, and can articulate the value flow to clients will capture market position. The firms that have not will find themselves in difficult conversations they are not prepared to win.

The gap between awareness and action is where this gets decided. Most firms know AI is coming. A minority are doing something about it. Which side your firm ends up on is determined by what you do this quarter.

If you're wondering where your firm sits — or what a purpose-built AI system could look like for your practice — get in touch →

Sources: Australian Financial Review (February 2026); Going Concern; Karbon State of AI in Accounting Report 2024; Thomson Reuters Future of Professionals Reports 2024–2025; CPA.com 2025 AI in Accounting Report; IBM Cost of Data Breach Report 2024; Office of the Australian Information Commissioner Notifiable Data Breaches Report 2024.