Xero’s getting smarter with coding and machine learning

12 September 2017

Share This

Hands up if you enjoy entering your expenses and receipts into your bookkeeping software… yep, not many of you. Data entry of expenses is one of the necessary evils of doing your own bookkeeping – but if you’re a Xero user, there’s good news. Xero is developing it’s AI and machine learning functionality to help you automatically code your transactions!

Yep, that’s right – the software is going to do some of the work for you. Some Xero users are now trialling a beta version that uses machine learning to enter your transaction codes for you automatically, taking some of the hassle out of your expenses admin.

Steve Ash, content consultant at CommsBreakdown, is one of the lucky people running the beta version. In this blog, he outlines why machine learning is about to take on a lot of the hard work out of getting your expenses and accounts in order.

Xero got a little smarter

As a freelancer, I rack up a fairly minimal amount of expenses and receipts. So it makes sense for me to do my own expenses admin and bookkeeping (if you’re a start-up or small business, with a larger number of expenses to input, you can save a lot of time by using the Receipt Bank app to do this)

I use the Xero mobile app to take a quick snap of any paper receipts (mostly coffee shop receipts) and then go into the desktop version of Xero to raise my bills, enter the transaction details and attach my scanned copy of the receipt.

It’s fairly quick to do, it keeps my records up to date and keeps my accountant happy (yes, happy accountants do exist, honestly!)

A few weeks back, however, I noticed something strange…

A ghost in the machine?

I spotted that when I entered the supplier name of my local coffee shop and began entering the details of the expenses, that my transaction code (code 429 for General Expenses) popped up automatically! Weird, I thought.

Then I entered a receipt for some stationery supplies bought at WH Smith. Once I’d selected WH Smith as the supplier, ‘Stationery’ as my description and ‘1’ as the quantity, the transaction code magically appeared again in the code field (Code 461 for Printing & Stationery) this time.

So, either there was a strange ghost in the (software) machine, or Xero has a new automation function that I’d not heard of. And, having spoken to the folks at Xero, it turns out that I’m running a new beta version that’s testing an AI-driven machine learning model.

Good news for me, and great news if you’re a Xero user with lots of bookkeeping to do.

The ever-present issue of human error

If you’re already a Xero user (and if you’re not, why not?) then you’ll know that Xero is very hot on keeping the product updated. And a key part of ongoing development in recent months has been building more and more automation into the software – automation that saves you (the business owner) time and makes your financial management more efficient.

Bookkeeping and data entry of expenses is one area where small businesses can use up a lot of the working week. But with human data entry comes the potential for human error, and that’s something Xero has been keen to reduce.

They analysed the use of their ‘Find and Recode’ function that’s used to amend incorrect coding of transactions. They found that globally across Xero, people had carried out 3 million find and recode operations – that’s a lot of human mistakes in coding to put right!

Teaching accounting to a robot

So, how is Xero providing a solution to the issue of human error in coding?

The developers at Xero HQ have been busy putting together a machine learning model that can learn how accountants and business users code their transaction, and then learn to perform this coding for them… automatically!

What they’ve come up with is an algorithm within Xero that’s 80% effective at automatically coding your transaction after learning from only four invoices! That’s highly impressive work, and a change in Xero that’s going to help millions of us to get our books done more quickly, more accurately and with as little hassle as possible

There’s a more detailed overview of Xero’s development of artificial intelligence and machine learning in this Xero blog post.

Author Profile Image

Steve Ash is a content writer and consultant at CommsBreakdown