AI in practice: ‘It’s already here’
Artificial intelligence (AI) is moving so quickly that before you know it, it's turning up everywhere, reports John Stokdyk.
Technology has accelerated to the point where when you first encounter a new concept, it has probably passed you by and become part of the furniture. For proof, just take a look at how AI has already infiltrated accountancy. During 2016, AccountingWEB members devoted considerable attention to AI and how it was likely to shape the future of the profession. But software developers and many accountants did more than just debate the issue. They started experiments to make the future happen.
“AI is already here – there are firms doing it, and not just the Big Four,” said Gordon Roxon, sales director for MindBridge, a supplier of automated audit and fraud detection tools to accountancy firms.
Roxon was attending this week’s Alternative AI for the Professions conference in London, where MindBridge clients from Kingston Smith and Kreston Reeves rubbed shoulders with representatives from KPMG, EY, Grant Thornton and numerous law firms and consultancies.
The impact of AI on the professions and jobs was a constant talking point, but the real benefit of the event was to see how many different ways AI was being applied – from extracting data from legal contracts for due diligence assignments to fraud detection, audit and basic customer service.
Wired magazine editor Gregg Williams proclaimed AI as the “new cloud”, but with so much of it about and spreading so quickly, it is worth keeping in mind his prediction that it will become an accelerator of other technologies.
Currently there are two main strands of AI: first, the classic data analysis scenario where automated pattern-checking and machine learning algorithms identify anomalies and underlying trends within large data sets. And where better to apply such technology than in automating the audit?
But that’s quite a limited view. According to KPMG audit partner Nick Frost, when undertaking an audit for a listed consumer goods supplier, he wants a more holistic view and won’t just stop at just analysing the financials. The contemporary auditor will explore other risk factors to obtain a holistic view of the organisation’s health: unstructured third party data from social media, sentiment analysis, public domain datasets and comparative analysis.
The other end of the AI spectrum revolves around robots and agents designed to interact with humans and respond to their questions and instructions. In terms of data processing and logic, understanding and interpreting human communication is as big a challenge as automating analytics.
No major accounting software house can show its face in public without making a pronouncement or demonstrating its software agent, as we have seen during the past year from UNIT4, Sage, Xero and QuickBooks.
Automation is also becoming commoditised in a process that Wired’sGreg Williams termed, “AI as a service”. Microsoft, Amazon, Google and Facebook are building these capabilities into their platforms and making them available to developers of all sizes.
Microsoft’s Raymond Hounon explained how the two branches of AI might start to converge in a typical accounting and finance setting, when using Microsoft’s Power BI analysis tools. While reviewing accounting data, an analytic bot might flag up unseen patterns, or prompt the user to refine their analysis through a series of context-sensitive questions.
Machine learning on the audit front line
Back on accounting’s front line, Kingston Smith partner Becky Shields said that clients are already asking audit firms about their use of technology during the tender process. The Big Four are spending billions on developing proprietary tools, so unless firms like Kingston Smith embrace similar capabilities with commercialised AI products such as MindBridge, audit will become two tier again, with the Big Four holding sway at the top end, she warned. “If we don’t have these solutions, we’re in danger of losing work,” she said.
Getting to grips with AI is hard work – as well as researching the market, Kingston Smith had to work extensively with MindBridge to train its software.
Getting data into a state in which machines can start applying their learning is a major hurdle for AI, but the structured nature of financial system data makes life easier for accountants than lawyers. At Kingston Smith, the MindBridge program can import a full transactional log from accounting ledger systems such as QuickBooks Online, Sage Intacct and NetSuite and will then carry out an analysis according to the auditor’s preferred risk controls – including time of journal posts, high values, sequence gaps, last three digit analysis and so on.
Answering a query about the impact of AI on jobs, Shileds put forward a view that many at the AI conference shared. While the machines will be able to eat up rote-like low value work, they will still need to be guided by their human masters. “The knowledge jobs will go, but the wisdom jobs will remain,” she said.
You might also be interested in
AccountingWEB’s interim Editor in Chief has been with the site since 1999 and returned to the editorial hot seat in March 2020 to lead the hunt for a long-term successor... Send a DM if you're interested! When not tending to the needs of AccountingWEB members and geeking out on their technology habits, he devotes much of his time to his oddball...