BGL managing director Ron Lesh said his firm has been experimenting with artificial intelligence for some time, particularly machine learning, and has discovered that it has plenty of potential for simplifying the process of transaction matching for SMSFs.
“With transaction matching we’ve got all these rules and the problem is that it is all really dependent on having a rule, and having data there, and finding stuff and matching stuff, but what happens if you didn’t need a rule? What happens if the whole idea of data input rules as we’ve got them at the moment disappear from transaction matching?” he explained.
“It disappears because you don’t need it anymore, so what we’ve been playing with is taking something that’s thousands of lines of code in a computer sense and bringing that down to 15 or 20 lines of code, where the machine learns using AI.”
A simple example of this is with a bank transaction, he explained, where the computer looks at the patterns, analyses those and comes up with suggestions and instead of creating a rule, the machine actually learns and the next time it actually auto-processes.
“So the whole idea of you having to say put it here or put it there — it actually doesn’t need you to do that,” said Mr Lesh.
“We are heading to zero touch transaction matching. Can you imagine that with SMSFs where you don’t have to actually match a transaction? Where the computer looks at the transactions that are there and says ‘I’ve seen all of these transactions before and in the past they’ve gone here so therefore this one has to go here, and I can give you 95 per cent confidence that this is where this has to go but if I can only give you 30 per cent confidence then I’m not going to post it automatically for you, I’m going to let you tell me, and then the next time this transaction comes through I know exactly where this goes’.”
Artificial intelligence, he said, can also be used for predictive compliance with audits and may be able to analyse the SIS rules around SMSFs and the tax rules related to superannuation.
“[Imagine] if the software actually looked at the data and said, ‘okay you’ve got a problem here, you’ve got a problem there, you’ve got this here — this is something you need to look at’,” he said.
Mr Lesh said the industry could start to see some predictive compliance technology coming soon.
