Ledgerbots: Assessing AI assistants from Xero, MYOB, Intuit & Sage

How are Xero, MYOB, Intuit & Sage using AI? We review the reality of AI in accounting software – and measure it against the hype of a fully automated "ledgerbot".

by | Sep 11, 2025

In Apple’s 2025 show Murderbot, a robot gains the ability to act unrestrained by its human creators and their rules. Image: Apple


At a glance

  • Xero’s AI chatbot, JAX, has disappointed users with its slow speed and unreliability.
  • MYOB is cautiously upgrading existing workflows like receipt capture and transaction categorisation with AI.
  • Intuit is aggressively marketing an “army of AI agents” to automate business functions.
  • Sage aims to eliminate the month-end close, but its Copilot chatbot faced reliability issues.

Accounting has long been seen as one of the professions most susceptible to automation. The logic is straightforward: it’s built on rules and repeatable processes, making it fertile ground for software.

And now the AI revolution has given us powerful new chatbots that some people suspect – even fear – will indeed automate accounting.

In Apple’s 2025 show Murderbot, a robot gains the ability to act unrestrained by its human creators and their rules. Perhaps it’s only a matter of time before an AI-powered “ledgerbot” gains a similar capability to reconcile bank accounts, process expenses and run payroll without human intervention?

Vendors are heading towards this ledgerbot future from different directions. Some see AI as a way to speed up routine work and support decision-making. Others are pitching a digital finance team: multiple AI-driven agents working in tandem to handle core financial functions, an idea that comes close to fulfilling the ledgerbot vision. The question is: How close are they to delivering on that promise?

Xero: High hopes, mixed reality

Xero was a pioneer in the use of AI in accounting software. Its introduction of machine learning to categorise transactions was celebrated as a genuine breakthrough that sped up bookkeeping and improved accuracy.

That early success meant expectations were high when Xero announced its next leap into generative AI. Many were keen to see how far the company could go with an embedded version of ChatGPT for accounting.

The result was JAX (Just Ask Xero), a chatbot designed to work not only inside Xero but also through SMS, email and WhatsApp.

A small business owner can send a couple of text messages and JAX will create an invoice, prefill the details with inventory item codes, and let the user approve and send it. (Creating an invoice through a text conversation with your accounting software is a novel experience.)

JAX can also summarise invoices by date, status or contact, provide cashflow insights, and pull help content from Xero Central without breaking a workflow.

Under the hood, a system called JAX Assure was designed to filter and check data before feeding it to large language models, with the aim of improving accuracy and reducing hallucinations.

The reality has fallen short of expectations. Accountants in public forums report three frustrations. First, JAX is slow. It often hangs on a “generating response” message while an accountant could have clicked through to the answer in seconds.

“I asked JAX what the income from a particular contact was… That’s what it should be able to do, but it couldn’t.”

Kylie Parker, Lotus accountants

Its accuracy is still unreliable. In one test conducted by this writer, a query to “show me all paid invoices over $1,000 in the past 12 months” produced a list that included an invoice for $550.

And third, its usefulness is limited. That same test reported there were 18 qualifying invoices, yet JAX displayed only the first five, including the incorrect $550 one, and gave no link to the full list.

One challenge is that chatbots in accounting software will inevitably be compared to ChatGPT and its peers, which operate with fewer restrictions and are comparatively far more powerful. This means that actions which would be simple for ChatGPT are not possible with JAX.

“Yesterday I had a client that was asking for information about his prior income as part of his legal claim for insurance. I asked JAX what the income from a particular contact was each year (for 13 years) to see what it produced. That’s what it should be able to do, but it couldn’t. I just pulled up an account transaction report instead and edited it in Excel,” says Kylie Parker from Sydney-based Lotus Accountants.

MYOB: The cautious path

MYOB has taken a much more gradual approach to adopting AI than Xero. It has instead concentrated on upgrading existing workflows.

Yonatan Bley, MYOB’s general manager of product and product marketing, says the company has deliberately focused on banking and invoicing, where automation can deliver the biggest time savings.

One of the most visible changes is enhanced OCR for receipt capture. The long-standing Snap and Track tool has been upgraded so it no longer just records totals, but can now extract richer data from receipts, including GST amounts, merchant or vendor names, and even individual line items. This reduces the amount of manual reconciliation and coding required.

MYOB has also rolled out improved AI-driven transaction categorisation by shifting to large language models to improve accuracy. The LLM analyses past user behaviour and refines its predictions as accountants accept or reject its suggestions.

Another step forward is AI mapping in client accounting, built in partnership with Belgian software Silverfin, a business intelligence reporting tool. Rather than manually mapping a client’s chart of accounts into MYOB’s client accounting system, the AI does it automatically. Bley says this feature can save practitioners up to an hour and a half per client during setup.

Looking ahead, MYOB’s roadmap includes experiments with autonomous agents that can perform specific accounting tasks within controlled parameters. Bley says the company is rapidly prototyping new features, with a strong focus on automating repetitive processes and listening closely to customer feedback to guide what gets released. 

Intuit: An army of AI agents

Intuit is taking the most aggressive pro-AI stance of all the software vendors. It has released what it calls a “digital team of agents” to tackle different business functions, from categorising transactions to chasing payments and even drafting customer emails. The pitch is that these agents form a kind of virtual workforce working alongside the business owner.

Intuit has wrapped existing automation with genuinely new capabilities and is selling it as a broader narrative: the AI workforce. Each agent is designed to mimic a human role inside a business:

  • Accounting Agent: automatically categorises transactions, reconciles books, and detects anomalies, reducing manual work and catching errors earlier.
  • Payments Agent: monitors cash flow, optimises invoicing and collection, and helps ensure bills are paid on time.
  • Customer Agent (currently in beta): sources leads from emails, drafts personalised responses, and tracks opportunities in the sales cycle.
  • Finance Agent: analyses financial data, creates forecasts, highlights budget variances and pinpoints their source.
  • Project Management Agent (beta): generates project estimates from documents and automates setup, with the aim of ensuring profitability.

But some of this looks more like a rebrand than a revolution. Receipt processing, for instance, is being promoted as part of the new Accounting Agent. As Aaron Patrick, better known as the QuickBooks Chap, pointed out in a YouTube review of an agent extracting details from a receipt to create an expense: “Part of the Accounting Agent is this autofill. We’ve kind of seen this already, right?”

Before the generative AI era, many of these functions were simply ‘automations’. The main difference now is that a conversation with the software triggers them.

“I don’t see AI replacing accounting. I see AI accelerating accounting.”

Lil Roberts

Generative AI is making it easier to push automations into more areas of accounting software, faster. Patrick was especially struck by the project management agent’s ability to set up projects from documents: “I’ve never seen anything like that,” he said.

The company’s marketing pushes the ledgerbot idea even further. QuickBooks agents are described as “digital teammates”, available around the clock to do the repetitive work so that business owners “only need to approve the work”. When they can’t answer a question, Intuit promises a “handoff” to trusted human experts. It’s a textbook ledgerbot pitch.

But not everyone is convinced. Lil Roberts, founder of Xendoo, a US-based accounting firm that uses Intuit’s competitor Xero, says she has seen clients get into trouble when they relied on QuickBooks’ AI. “QuickBooks AI is more cavalier, more aggressive. In the early days of their AI, we would have customers come to us and say that they turned on QuickBooks AI and it wrecked their books, and now they’re using a spreadsheet,” she recalls.

Roberts says she is less concerned that AI will replace accountants than that it can accelerate mistakes. “I don’t see AI replacing accounting. I see AI accelerating accounting,” she says. AI agents give business owners more power to act directly in their books. For those who are diligent, that may be helpful. But for those who aren’t, the software can create bigger errors, faster, leaving accountants to clean up the mess.”

Sage: The war on the month-end close

“We’ve held up the monthly close as a bit of a boogeyman. We said, seven years ago, our goal was to get rid of it. That was pretty provocative,” says Aaron Harris, Sage’s CTO, who opened Sage’s centre of excellence in AI when he joined the company in 2018.

That ambition – to kill the close – runs through everything Sage is doing with AI. On the enterprise side, most of the heavy lifting takes place in the company’s ERP, Sage Intacct, where anomaly detection, compliance monitoring and accounts payable automation are already well established. The software can read invoices, categorise them, match them to purchase orders and flag duplicates.

Harris groups these capabilities into three themes: continuous accounting, continuous assurance and continuous insights. The idea is to automate the workflows, continuously check for errors, and then analyse real-time data to help finance teams make faster decisions.

Sage is claiming that its AI investment has produced results.  Harris says that a medium-sized business taking two or three weeks to close their books can now do it on Sage Intacct in two or three days.

Sage has also rolled out its chatbot, Sage Copilot, to over 50,000 UK businesses, says Harris. It is available for users of Sage Intacct and its small business software, Sage Business Cloud Accounting. (Sage Business Cloud Accounting is not available in Australia.)

Business owners can ask Copilot for insights on cash flow or outstanding invoices, and it will surface reminders about upcoming tax returns or late receivables.

It can also draft and send email reminders to customers with overdue bills, predict which invoices are most likely to be paid late, and help owners prioritise collections.

However, reliability has been a concern. In January 2025 Sage quietly took Copilot offline after a customer discovered that the assistant had returned invoice data from other businesses. Sage described it as a “minor issue involving a small amount of customers” and said no invoices were exposed.

It is also telling that there is very little feedback – positive or negative – about Sage Copilot in online forums. For a tool positioned as central to Sage’s AI strategy, the silence suggests adoption has been limited.

Still a dream – for now

The promise of a ledgerbot that runs the books on its own is still more fantasy than fact. In fact, accountants are arguably as secure as ever.

The AI that has made its way into accounting software bears a resemblance to the ledgerbot ideal. Yet in practice, it does something less dramatic. It speeds up repetitive work and gives accountants more time for higher-value tasks. Instead of replacing the profession, it helps them serve clients better.

The hype is loud, but the day-to-day impact is still modest – and accountants remain firmly in charge.


Explore a mix of expert-led sessions, hands-on workshops and meaningful networking opportunities at IPA’s National Conference 2025 which has been designed to inspire, challenge, and energise you. Register now.

Share This