AI for Accountants: How Australian Firms Are Using AI in 2026

I'm Chris from Studio Slate. We built Reckon, an AI-powered accounting platform, and Sub Tracker, which uses Claude to parse bank statements and find recurring subscriptions. I wrote this guide because every accounting partner I speak with asks the same question: "Is AI real, or is it just marketing?" The answer is both, and the difference between the two is where most firms get stuck.

The adoption gap: global vs Australian

The global numbers paint a clear picture of acceleration. McKinsey's 2025 State of AI survey found that 88% of organisations now use AI in at least one business function, up from 78% a year earlier and 55% the year before that. Stanford's 2025 AI Index reported $252.3 billion in global corporate AI investment, with private investment jumping 44.5% year over year.

The accounting profession specifically has moved faster than most. Karbon's 2026 State of AI in Accounting report, surveying nearly 600 professionals across six continents, found that 98% of firms now use AI daily or multiple times a day. Average time savings reached 60 minutes per employee per day, a 7% increase from the prior year. That works out to roughly 250 hours per employee per year, or about six weeks of full-time work.

Australia is a different story.

The Deloitte/Amazon AI Edge for SMBs report, which surveyed more than 1,000 Australian SMBs in November 2025, found that while two-thirds of SMBs use some form of AI, only 5% are fully AI-enabled. A fully enabled business, by Deloitte's definition, has an AI strategy embedded in core processes, provides employee training, and maintains a centralised data system. The remaining 95% are experimenting, dabbling, or ignoring it entirely.

The CPA Australia Business Technology Report 2025 adds more context. While 89% of Australian businesses reported using AI (up from 69% in 2024), only 16% have AI fully embedded across operations. And Australian respondents were less likely to invest heavily in AI than their APAC peers: 15% identified AI as their top technology investment, compared with 32% across the wider region.

For accounting firm partners, the data says two things at once. First, your competitors globally are already using AI and saving meaningful time. Second, most Australian firms have not yet moved beyond basic experimentation. That gap is an opportunity, but not one that stays open indefinitely.

What AI can do in accounting today

The practical question is not whether AI works. It is what it works for. In 2026, AI handles certain accounting tasks reliably and others poorly. Here is what falls into each category.

Bank reconciliation

This is where AI has made the most visible progress. Matching bank transactions against accounting records is repetitive, rule-based, and high-volume, which makes it a natural fit.

Xero's JAX (Just Ask Xero) launched auto bank reconciliation in global beta in November 2025 for Australian users on Grow plans and above. JAX uses four layers to match transactions: existing bank rules, invoice and bill matching, patterns from reconciliation history, and machine learning predictions. Digit Business, a Xero-focused firm that has tested JAX across real client files, reports that Xero's target is to auto-reconcile more than 80% of bank statement lines.

MYOB's Smart Reconciliation takes a similar approach, matching bank feed transactions to categories automatically and auto-reconciling on a user-chosen schedule. It is currently in beta across Solo by MYOB, MYOB Business Lite and Pro, and AccountRight.

QuickBooks' Accounting AI groups high-confidence transactions into a "Ready to post" queue and auto-posts transactions from QuickBooks Payroll and Bill Pay with no manual intervention.

What works: Matching straightforward transactions where the AI has seen the pattern before. Direct debits, regular supplier payments, payroll transfers.

What does not work well: First-time transactions, ambiguous descriptions, split transactions, and anything that requires context about why a payment was made, not just who it was from.

Document processing and data entry

Receipt and invoice processing has been automated at scale longer than any other accounting function.

Dext (formerly Receipt Bank) has processed more than 1 billion receipts and invoices with claimed 99.5% accuracy. In January 2026 alone, the platform processed 31.4 million documents globally, and Dext estimates this saved more than two million hours of manual work based on a conservative three-to-four-minute estimate per document. Their new AI Assist agent, launched in March 2026, goes further by learning how individual users categorise transactions and suggesting automation rules based on those patterns.

Botkeeper offers a hybrid model where AI handles initial categorisation and a human team provides quality control. Pricing scales from $59 per client per month at volume (25+ clients) to $149 for smaller practices. Independent reviews put the overall cost at $500 to $2,000 per client per month when factoring in complexity, and report 97% to 99% accuracy on routine categorisation.

What works: Extracting data from structured documents (invoices, receipts, bank statements). Categorising transactions that follow patterns. Flagging duplicates.

What does not work well: Handwritten documents, poor-quality scans, multi-currency invoices from unfamiliar jurisdictions, and anything requiring judgment about whether an expense is personal or business.

Practice management and workflow

Karbon AI integrates directly into practice management workflows. Current features include client summaries (combining emails, notes, work items, and billing data into a single brief), email composition and refinement, smart suggestions for task assignment, and work conversation summaries. Karbon AI is now available to all customers with data sovereignty in the region where data is stored.

Karbon has also announced AI Agents for early 2026: AI-powered teammates modelled after roles in a firm (Bookkeeper, Tax Admin, Practice Manager, Fractional CFO). These agents will prepare workpapers, reconcile transactions, follow up with clients for missing documents, build forecasts, and draft email responses. They are in beta with select firms.

What works: Summarising email threads, drafting responses, suggesting task assignments, generating client briefs from existing data.

What does not work well: Understanding the politics of a client relationship, knowing when to escalate, managing the emotional labour of a difficult conversation.

BAS preparation

MYOB made the first move here. MYOB AI BAS, announced in March 2026, is described as Australia's first agentic BAS offering. It suggests BAS treatments for transactions, flags items that need review, links supporting documents, and produces a pre-populated report with lodgment totals for the accountant or bookkeeper to review. It is rolling out progressively throughout 2026, starting with sole traders.

What works: Automating the data preparation and categorisation that precedes BAS lodgment. Reducing the time between raw transactions and a draft BAS.

What does not work well: Judgment calls on mixed-use assets, determining the correct GST treatment for complex or unusual transactions, and knowing when a transaction pattern indicates a client's business structure has changed.

What AI cannot do (and why that matters)

The list of things AI cannot do in accounting is shorter than it was a year ago, but the items on it are the ones that define the profession.

Professional judgment on complex tax strategy. Should a client restructure as a trust? Is an aggressive R&D Tax Incentive claim worth the audit risk? Should a client change their reporting period to align with a new contract? These decisions require understanding not just the law but the client's risk tolerance, business plans, family situation, and relationship with the ATO. No AI model has this context, and the Tax Practitioners Board Code of Professional Conduct still requires a registered tax practitioner to review and approve anything that goes to a client.

Client relationships and advisory. The Hinge Marketing 2026 High Growth Study found that high-growth accounting firms rely on referrals 34% less than average firms because they have built stronger advisory relationships. Those relationships depend on partners who understand a client's business deeply and can offer strategic advice. An AI can summarise financial data. It cannot sit across a table from a business owner who is deciding whether to sell the company and read the room.

Interpreting ambiguous regulation. Tax law is written by legislators, interpreted by courts, and applied by practitioners. When the ATO issues a ruling, an experienced accountant understands not just what it says but what it means in the context of their clients' specific circumstances. AI language models can retrieve and summarise rulings, but they cannot reliably reason about edge cases where the answer is genuinely uncertain.

Ethical decisions. Is this deduction defensible? Is this disclosure sufficient? Should we report a client's error to the ATO, or advise them to self-correct? These are judgment calls that carry professional liability. No firm should delegate them to a machine, and no responsible AI vendor is suggesting they should.

The practical boundary in 2026 is clear: AI handles the data preparation layer (entry, categorisation, reconciliation, document processing, draft preparation), and humans handle the judgment layer (strategy, advice, relationships, ethics, sign-off). Firms that understand this boundary will extract the most value from AI. Firms that expect AI to replace professional judgment will be disappointed or worse.

The tool landscape: what is available today

Here is a summary of the major AI accounting tools available to Australian firms in early 2026, with what they actually do versus what their marketing implies.

A few things stand out.

First, every major platform is building AI into the product you already pay for. Xero, MYOB, and Karbon are not charging extra for their AI features during beta. That is deliberate: they want adoption, and they know firms will not pay a premium for features they have not tested. Whether these features stay free after beta is an open question. Xero has flagged that AI features may move to paid tiers.

Second, the standalone tools (Botkeeper, Vic.ai, Truewind) solve specific problems at scale. Botkeeper handles bookkeeping for firms that want to outsource it. Vic.ai is built for accounts payable in mid-market and enterprise finance teams. Truewind targets funded startups that need investor-ready books. If your firm does not fit those profiles, the platform-native AI (Xero, MYOB, Karbon) is where you will get the most immediate value.

Third, there is a meaningful gap between what these tools do and what a firm-specific AI build can do. Every tool on this list is built for the broad market. They handle the 80% of workflows that most firms share. The remaining 20%, the firm-specific intake quirks, the partner's tone in client emails, the legacy spreadsheet that a senior built in 2009 that still runs the BAS quarter, is where generic tools fall short and custom AI builds fill the gap.

The SaaS vs custom spectrum

Not every firm needs a custom AI build. Most firms should start with the AI built into the tools they already use. Here is how to think about the spectrum.

Start with platform AI (cost: $0 incremental). If your firm uses Xero, MYOB, or Karbon, you already have access to AI features. Turn on Xero JAX's auto bank reconciliation. Enable Karbon AI's email summaries and client briefs. Try MYOB's Smart Reconciliation. These cost nothing above your existing subscription and provide immediate time savings on routine tasks. Karbon's research shows that firms with formal AI training see the biggest productivity gains, so invest an hour training your team on the features before expecting results.

Add specialist SaaS where the gap is clear (cost: $7 to $149/client/month). If your firm processes high volumes of receipts and invoices, Dext's AI Assist will save more time than manual entry or basic Xero functionality. If you want to outsource bookkeeping entirely, Botkeeper provides a hybrid AI-plus-human service. These are point solutions that solve specific problems, and they integrate with the platforms you already use.

Consider a custom build when generic tools hit their ceiling (cost: $5,000 to $35,000 upfront). This makes sense when your firm has specific workflows that no SaaS product handles, when you need AI trained on your actual client base and templates, when you want integrations that the platforms do not offer, or when per-seat pricing becomes a constraint at scale. A custom build means you own the code, the prompts, and the deployment. There are no per-seat fees and no dependency on a vendor's product roadmap.

The wrong approach is skipping straight to a custom build without first extracting value from the tools you already have. The also-wrong approach is waiting for Xero and MYOB to solve every problem, when the firm-specific 20% is where the most time is actually lost.

The talent shortage context

AI in accounting is not just a technology question. It is a staffing question.

In October 2025, the Australian government added taxation accountants to the Occupation Shortage List. A CA ANZ survey of 395 members who advertised vacancies in 2024 found that taxation accountant was the role at highest risk of national shortage, accounting for 29% of advertised roles. Vacancy fill rates for management accountants sat at 45%, and for taxation accountants at 59%.

The main reason firms could not fill roles was not salary. It was a lack of suitable applicants with the required experience and technical skills.

This is where AI stops being optional and starts being operational. If you cannot hire another accountant, the next-best option is making the accountants you have more productive. Saving 60 minutes per employee per day (the Karbon figure) across a 10-person firm is equivalent to hiring 1.25 additional full-time staff, at a fraction of the cost and without the recruitment timeline.

MYOB's blog on their Discover 2026 event made the point directly: "Talent shortages are expected to worsen over the next decade." Their AI strategy, building AI features into the software firms already use, is designed specifically for firms that cannot add headcount but need to do more work.

How to evaluate AI ROI for your firm

Most AI vendors talk about time saved. That is useful but incomplete. Here is a framework for evaluating whether an AI tool or investment is worth it for your firm.

Step 1: Identify the hours. Pick the three most time-consuming repetitive tasks in your firm. For most firms, these are bank reconciliation, receipt and invoice processing, and email management. Estimate hours per week spent on each, per employee.

Step 2: Attach a cost. Multiply the hours by the loaded cost of the employee doing the work. If a bookkeeper costs $80,000 per year fully loaded (salary plus super plus overheads), that is roughly $41 per hour. Ten hours per week of reconciliation across three staff is $63,960 per year.

Step 3: Estimate the reduction. Be conservative. AI will not eliminate these tasks. It will reduce them. Xero's target is 80%+ auto-reconciliation. Dext claims 90%+ reduction in document processing time. Assume 50% reduction to be safe. That turns $63,960 into $31,980 in recovered time.

Step 4: Compare the cost. Platform AI (Xero JAX, MYOB AI, Karbon AI) costs nothing incremental. Dext costs $7 to $18 per user per month. Botkeeper costs $59 to $149 per client per month. A custom build costs $5,000 to $35,000 upfront. Compare the annual cost of the tool to the annual value of recovered time.

Step 5: Factor in what the recovered time enables. Recovered bookkeeping hours have different value depending on what they are redirected to. If staff go from reconciliation to advisory work billed at $300 per hour, the ROI multiplies. If the time is simply absorbed by other compliance work, the value is the cost saving alone.

The Deloitte/Amazon AI Edge report found that Australian SMBs moving from basic to intermediate AI adoption can expect a 45% increase in profitability. Moving from intermediate to fully enabled use produces a 111% increase. The compounding effect of AI on firm economics is real, but it requires moving beyond experimentation into structured adoption.

Practical first steps for firm partners

If you are an accounting firm partner reading this and have not yet done anything structured with AI, here is a starting point ordered by effort and impact.

Week 1: Turn on what you already have. Enable Xero JAX auto bank reconciliation. Turn on Karbon AI email summaries. Review MYOB's AI BAS beta eligibility. These features are free, already in your tools, and require no procurement.

Week 2: Measure one workflow. Pick bank reconciliation for one client. Time how long it takes manually. Run it with AI for a week. Compare. This gives you a real number, not a vendor's number.

Week 3: Write a one-page AI policy. Karbon's research found that only 21% of firms have a documented AI strategy, yet firms with formal policies see significantly stronger outcomes. The policy does not need to be complex. Cover which tools are approved, what data can and cannot be used, and the human review requirement for anything client-facing.

Week 4: Train the team. Spend one hour showing staff how the AI features work in your existing tools. Karbon's data shows that firms investing in AI training save meaningfully more time than those that simply turn features on. The tool is only as useful as the person using it.

Month 2 onwards: Evaluate the gaps. After a month of using platform AI, you will have a clear picture of what it handles well and where it falls short. Those gaps inform whether you need a specialist SaaS tool, a custom build, or nothing more than what you have.

Accounting Firm Marketing: Getting Clients Beyond Referrals

How accounting firms actually get clients in 2026. Referral-to-Google pipeline, website requirements, Google Business Profile, LinkedIn, and budget benchmarks.

Read more

What comes next

The direction of travel is clear. Every major accounting platform is moving from AI-assisted features (suggestions, summaries, auto-categorisation) to AI agents (autonomous teammates that complete multi-step workflows with human oversight). Karbon is launching Bookkeeper and Fractional CFO agents. Xero describes JAX as an "AI financial superagent." MYOB's AI BAS is an agentic offering that handles the full BAS preparation pipeline.

This shift matters because agents do not just answer questions or make suggestions. They take actions. They follow up with clients for missing documents. They prepare workpapers. They draft BAS returns. They route approvals.

The firms that will benefit most from this next wave are the ones that have already built the foundation: AI policies, trained staff, measured workflows, and a clear understanding of where human judgment is required. The firms that have not started will face a steeper learning curve when the tools move from suggestion to action.

The economic incentive is real. Deloitte estimates that if just 10% of Australian SMBs advanced one rung on the AI adoption ladder, $44 billion could be added to GDP annually. At the firm level, the maths is simpler: the firms that use AI to recover time and redirect it to advisory work will be more profitable than those that do not. The gap will compound.

Sources

• McKinsey: The State of AI (2025) - 88% of organisations use AI, up from 78% a year ago, 62% experimenting with AI agents
• Stanford HAI: AI Index 2025 - $252.3B global corporate AI investment, 78% of organisations used AI in 2024
• Deloitte/Amazon: The AI Edge for SMBs (Nov 2025) - Only 5% of AU SMBs fully AI-enabled, $44B GDP opportunity, 45-111% profitability increase
• CPA Australia: Business Technology Report 2025 - 89% AU adoption but only 16% fully embedded, 15% vs 32% APAC investment
• Karbon: State of AI in Accounting 2026 - 98% of firms use AI daily, 60 min/day savings, only 21% have AI strategy
• CA ANZ: Occupation Shortage List Submission (2025) - Taxation accountant at highest risk of national shortage, 29% of advertised roles
• Digit Business: Xero JAX Explained (2026) - Auto reconciliation target 80%+, four-layer matching, included in subscription
• Xero Blog: Accountable Intelligence (2026) - JAX as "AI financial superagent", vision to automate 90%+ of routine tasks
• MYOB: AI Products Launch (2026) - AI BAS (first agentic BAS in AU), Smart Reconciliation, AI Business Insights
• Dext: AI Assist Launch (2026) - 31.4M documents processed in Jan 2026, 2M+ hours saved, 99.5% accuracy
• Karbon: AI Agents (2026) - Bookkeeper, Tax Admin, Practice Manager, Fractional CFO agents launching 2026
• Hinge Marketing: 2026 High Growth Study - High-growth firms rely on referrals 34% less, invest 9% of revenue in marketing
• Accountants Daily: Taxation Accountant Shortage (2025) - 395 members surveyed, vacancy fill rates 45-59% for accountant roles