Published on the 02/08/2017 | Written by Donovan Jackson

Grand plans for AI, machine learning and automation starts with minor improvements…

There is a tendency to think in grand terms when it comes to data analytics, machine learning, AI. After all, this is supremely complex stuff. It’s on the bleeding edge of the bleeding edge. It’s state-of-the-art, revolutionary. It’s so big a deal that the right adjectives are hard to find.

Except, said Xero’s Head of Data Innovation Sam Daish, a lot of the truly useful advantages can be found by focusing on the little things that cause daily annoyance, interruptions or derailments of the efficiency train. “There is a spectrum in terms of ‘intelligence’; when an interaction with software feels smart and helps you, it could rest on sophisticated algorithms, data science and machine learning – or it could sometimes be really simple, like the email feature which notices that you’ve forgotten the attachment.”

There is a lesson in Daish’s observation. It doesn’t have to be the grand things which deliver great improvements. In the accounting profession, where Xero plies its considerable trade, small checks and balances can have big results. Identifying when and where information is entered incorrectly, or omitted, can have major implications for the small business owner, or for their accountant.

“At the heart of identifying how we can improve Xero is identifying what a joyful experience is for those using the system,” Daish confirmed. “So, we focus on identifying common mistakes or barriers that we can remove – and there are many examples of them in accounting and other number-oriented systems. With little prompts and targeted help, it makes things smarter and easier to work with. Sophistication isn’t necessarily complex; instead, you want machines to focus on what they are good at [detail] which helps humans to avoid making mistakes.”

Xero, said Daish, is focused on using analytics and data driven insights in a couple of key areas. It wants to make accounting easier and more pleasurable for small businesses; it wants to give accountants better tools to look after small businesses. And it wants to identify what makes small businesses successful (this is the grander scheme of things).

“Nobody gets into business to do more accounting, so perhaps counter-intuitively, we are working to make it possible for our customers use Xero less,” quipped Daish.

Because, he added, the software should work for the business owner rather than the other way around: drop in the invoices and bills and let the system handle it; the machine learns how you use the system, how items are accounted for and invoices allocated and coded, and knows which metrics you find interesting.

Machine learning, said Daish, makes personalisation possible at scale. And often, the outcomes of that are small things, like appending the right codes to invoices. It doesn’t take a genius to understand that removing many minor niggles from one’s day can add up to a decent productivity boost.

Secondly, he said, AI is being applied to make accountants more effective, addressing their bugbears, which are slightly chunkier than the niggles faced by the business owner.  “We want to equip accountants with tools which help them advise the small businesses they work with, allowing them to examine businesses in an automated way to provide better advice.”

Benchmarking is the low hanging fruit, but Daish said there are more possibilities; moving further up the sophistication ladder, Xero is looking into using data to find out, for example, the characteristics of a successful business. “We can’t do that today, but we’re working towards it. If we can provide tools like this to accountants, it becomes a powerful tool to shape productive conversations with business owners. It puts them in a position of knowledge to advise, by being able to look at a very granular level at how a business is operating over time, compare with others, and contextualise that with what the business is trying to achieve.”

Crucially, said Daish, the big data and analytics which must stand behind such insights, will be turned into ‘actionable suggestions’ for small business owners – rather than insights which make you go ‘wo that’s interesting, but how do I use it’.

With a quarter of small businesses failing within their first three years, using AI and machine learning to make owners more efficient (and needing to worry less about accounting) has obvious appeal. When their accountant is equipped to see how that business is tracking against its peers, there is every possibility advice and guidance can be provided to put the company back on track. And, ultimately, when it becomes possible to know what makes for a successful small business, with actionable suggestions from AI, perhaps the high failure rates will become a thing of the past.