The AI tool transforming Moore Kingston Smith’s audit offering (Part 2)

How did you solve the problem of extracting data from multiple sources?

We decided to build our own data platform to rationalise all our client information. We do that in-house, using ETL tools – extraction tools that connect to over 350 different accounting systems. We are on the journey to real-time advisory. Clients will post a transaction and if they’ve got one of our ETL tools, it comes straight into our system.

Is there a big gap between the mid-sized firms such as Moore Kingston Smith in terms of technology adoption, and the Big Four?

Undoubtedly there is a gap in terms of the spend. When we put a budget together that’s affordable for a client, we do have to carefully consider what they are getting out of it. The Big Four can spend a lot more money on this. Through Engine B, we found out that KPMG have 25 different ETL projects going on, and that’s just the data extraction. But they’re far less agile than we are. We have a massive disadvantage, but I think there are lots of advantages as well, there are some much smaller firms out there who, in terms of audit (less so with the analytics generally) but in terms of the advisory are way ahead of the Big Four. SMEs still need access to AI tools and the enhanced benefits they bring, and not just companies who use the Big Four.

Has MindBridge spotted any instances of fraud?

Fortunately, in the period that we’ve been using it, there haven’t been any frauds, but we have tested it with our forensics accounting team by pushing data through that we’ve known to be fraudulent, and it would have picked it up. That’s how we validated it. Once you spy a fraud, you can then cover the whole population wherever you need to look for another transaction with these properties.

Do regulators need to do some work to catch up with technology?

There is a new ISA [International Standard of Audit] coming out, which touches on data analytics, and I hoped it was going to move towards ‘these sorts of processes you should do’, but it’s still very much can do rather than should do. We had a roundtable discussion with the regulators and most firms in the top 20. And we talked about some of the technology that was available to us, and what the barriers are of implementing those in terms of the regulatory guidance. Sadly, at the time, the FRC were very honest and said, within the programme of the things we’ve got to write in the next two years, it isn’t going to hit the table. That’s such a long time in terms of the technology and the speed its advancing. There is nothing in the ISAs that precludes you from using these technologies. But the thing that makes people nervous, if you’re looking to move towards whole population sampling, and identifying outliers, well, if by virtue of the fact that that is an outlier, does that mean you’ve got to test all of them? Can you cluster them and say, right, all of these are the same type of outliers that are going to test one of those? That’s the problem – you are very much out on your own. Nobody wants to be penalised publicly, and that can stifle innovation. That’s why I’m happy to talk honestly about our experiences, because if we’re not out on a limb, and everybody else is doing it, then as a group we’re a lot more likely to get the regulators to react.

Would you say the industry is behind in comparison to others?

If you compare it to law, their use cases for AI are a lot more talked about and proven in terms of reading leases, contracts, all those sorts of things. There are some well-publicised use cases for it, which is very good, considering the data they’re dealing with is unstructured. Our data is largely structured, if not semi-structured. We ought to be a lot further ahead.

And would you say Moore Kingston Smith is ahead compared to other similarly-sized firms?

Definitely. Others are adopting these technologies, but they haven’t solved the data layer.

What about other countries?

The reason why some of the European countries are farther ahead in some of their analytics offerings is because everyone’s reporting in the same format. When you’re performing analysis, it’s a lot easier. There’s a whole load of mapping that has to go on in the UK, because everybody’s system is different. Whereas the technology platforms in all these other European countries and the US information is largely standardised – the AICPA have made big inroads into that.

What do you predict we’ll see next with technology in accounting?

Real-time advisory, and that data insight is going to be huge. As some of the low-level compliance falls away, the challenge we will have is bringing people in, in year one, and helping them to understand the systems and what can go wrong when that low-level stuff has been done for them. People skills will have greater focus as technology does some of the other jobs for us.
EQ, going forward, is going to be just as valid as IQ. Accounting will require a broad skill set – being book smart alone is no longer going to cut it.