Xero folds AI receipt capture into its plans as Making Tax Digital nears

On 17 February Xero said it would build AI-powered data capture and extraction into its platform for UK customers, with the rollout taking place over the first two weeks of March. Customers on all business edition plans will be able to photograph receipts in the Xero Accounting App, email documents to a dedicated address or drag and drop them on the web, at no extra cost.

The timing is deliberate. From 6 April, sole traders and landlords with more than £50,000 of income from self-employment and property must use recognized software to keep digital records and send quarterly updates to HM Revenue & Customs (HMRC) under Making Tax Digital (MTD) for Income Tax. HMRC’s own announcement on 5 February put the number of sole traders and landlords in scope from April at 864,000.

For anyone choosing software for a small business, the more interesting question is not whether AI can read a receipt. It is what changes when AI arrives as a feature inside a tool the business already pays for, rather than as a product it decides to buy.

What Xero is adding

Xero says the feature uses large language model technology to read receipts, sales invoices and landlords’ rental statements in under 20 seconds and turn them into pre-populated digital records. It then suggests which existing bank transactions match those records, for the customer to review and approve. The company claims the technology provides “an enhanced level of accuracy”, a claim that cannot yet be tested independently.

Kate Hayward, UK managing director at Xero, tied the launch directly to the tax change. “Many businesses are still trying to understand the requirements around Making Tax Digital, but with AI the good news with Xero is that data entry doesn’t have to be done manually any more, and users can worry less about mistakes,” she said.

Xero describes the March release as the first set of capabilities, available in the UK before other markets, and says further AI-powered bookkeeping automation and insights will follow in the coming months.

Why bundling matters

Government research suggests much business AI use already comes through off-the-shelf tools. A survey of 3,500 UK private sector businesses with at least five employees, commissioned by the Department for Science, Innovation and Technology (DSIT), carried out by IFF Research and Technopolis Group between February and May 2025 and published on 28 January, found that 16% were using at least one AI technology. Among micro businesses the figure was 14%, against 36% for large ones.

Of those using AI, 85% used natural language processing and text generation, which the report attributes to the availability of off-the-shelf generative tools. Administration was among the most common areas of use, cited by 72% of current and prospective adopters.

The spending data is more telling. Just under a third (31%) of businesses using or planning to use AI reported no AI-related spending in 2024, and 27% of those already using AI said they had spent nothing. The report suggests that could reflect AI built into existing solutions or free tools that need no dedicated budget. Xero’s move fits that pattern, with the AI absorbed into the subscription rather than sold separately.

The same survey found that the most common reasons for not adopting AI were a lack of identified need and limited AI skills and expertise. A statutory deadline supplies the need. A feature that works from a phone photo lowers the skills bar. If bundled AI is going to raise adoption among the smallest firms, compliance software is a likely place for it to start.

The case for caution

None of this makes AI a requirement for MTD. HMRC stresses that free software options are available and describes the quarterly submissions as “light-touch” updates that are “not extra tax returns”. Those who genuinely cannot use digital tools can apply for an exemption. Businesses joining in April 2026 will not receive penalty points for late quarterly updates for the first 12 months, and under the new regime a £200 penalty applies only once four points are reached.

The evidence on productivity is also mixed. In the DSIT research, three-quarters (75%) of AI users reported improved workforce productivity, but 77% had not yet seen any change in revenue, and the authors caution that the figures are self-reported.

Accuracy is the obvious risk with any extraction tool that feeds a tax record. Xero’s design keeps a person in the loop, asking customers to review and approve suggested matches, and that step matters more than the speed of the read. The DSIT survey found 84% of businesses using AI applied at least some human checking to its outputs. Small firms should not be among the minority that do not.

There is a structural point too. As accounting platforms fold AI capture into the base price, the value of separate capture and data entry tools is likely to come under pressure, and more of a small firm’s financial records may end up running through a single vendor. That is convenient. It also makes switching harder later.

What small firms should do now

For businesses inside the April 2026 threshold, the priority is to choose software and sign up; features come second. For those already on Xero, the March rollout is worth testing on a month of real receipts and invoices before the first quarterly update is due on 7 August 2026.

Buyers should check four things:

• whether the extracted records match the source documents often enough to trust;
• who reviews and approves entries, and when;
• what happens to uploaded documents and where they are stored;
• whether they are still paying for a separate capture tool that the platform now covers.

The lesson extends beyond accounting. For most small businesses, AI is likely to arrive as a default setting in software they already use. The firms that benefit will be the ones that test it against their own records, keep the human check and stop paying twice.