In Murderbot (2025), a robot gains the ability to act unrestrained by its human creators and their rules.
At a glance
- The dream of a fully autonomous “ledgerbot” has not yet been realised.
- Sage aims to “kill the close,” but its AI Copilot has had a quiet launch.
- Xero’s JAX chatbot has been criticised for being slow, inaccurate, and of limited use.
- Intuit’s “AI workforce” risks accelerating errors, making accountants more essential than ever.
Accounting has long been seen as one of the professions most susceptible to automation. The logic is straightforward: accounting rests on rules and repeatable processes. It should be a perfect target for smart software.
And now the AI revolution has given us powerful new chatbots.
In Apple’s 2025 show Murderbot, a robot gains the ability to act unrestrained by its human creators and their rules. Surely it’s only a matter of time before an AI-powered “ledgerbot” gains the same capability, reconciling bank accounts, processing expenses and running payroll on its own?
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?
Sage: Killing the 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.
“The monthly close is a very personal and intimate part of accountants’ lives. It’s been very useful in creating sort of an emotional appeal to the story.”
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 claims 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 a chatbot, Sage Copilot, to more than 50,000 businesses in the UK using Sage Intacct and its small business accounting software, Sage Business Cloud Accounting, Harris says.
Business owners can ask Copilot for insights on cash flow or outstanding invoices, and it will surface reminders about upcoming VAT 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.
To make this work in the messy, real world of small business accounting, Sage has worked to improve accuracy and reliability. It has trained and fine-tuned its own large language models, starting with European AI company Mistral, and has augmented them with accounting-specific knowledge, including material from the ICAEW.
“I asked JAX what the income from a particular contact was each year … That’s what it should be able to do, but it couldn’t.”
Kylie Parker, Founder, Lotus Accountants
But 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 such as LinkedIn or Reddit. For a tool positioned as central to Sage’s AI strategy, the silence suggests adoption has been limited.
Xero: Early promise, patchy results
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 accountants and business owners 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.
Its accuracy is still unreliable. In one test, 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’” reports Lotus Accountants founder Kylie Parker. “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.”
Intuit: The most ambitious play
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. Accountant Aaron Patrick – better known online as the QuickBooks Chap – points this 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?”
“I don’t see AI replacing accounting. I see AI accelerating accounting.”
Lil Roberts, founder, Xendoo
Before the AI era, many of these actions would have been called automations. The difference is that these automations are now triggered through a conversation with the software. Generative AI is also 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.)
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 – software that looks less like accounting software and more like a digital finance team.
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.
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 more secure than 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.
