The AI tool transforming Moore Kingston Smith’s audit offering (Part 1)

Becky Shields, partner at top 20 firm Moore Kingston Smith, is passionate about the potential of technology in the accounting profession. In her 13 years at the firm, she has focused on IT, and now chairs an IT committee at the ICAEW as well as playing a key role in Engine B, the consortium run by ex-KPMG luminary Shamus Rae. It looks at how the digital revolution impacts on professional services. For three years, Moore Kingston Smith has used the AI MindBridge software for its audits. Here, in this two-part interview, she shares her thoughts on how the technology has improved the offering Moore Kingston Smith can give its clients, the future of audit, and how close we are to real time advisory.

How did the partnership between Moore Kingston Smith and MindBridge come about?

When I became a partner, I was asked to look at data analytics. And we just researched what was out there. AI was a buzzword, which meant that when it comes to talking to clients about it, it was always going to be an easier sell than some of the others that were less AI-focused. There are some products specifically for the audit market, like MindBridge and Inflo. But we also looked at things like IBM Watson and other platforms out there, because that’s been tweaked for the audit market, but we’re much broader than the audit market. In the end, we had a matrix of things we wanted it to do, and MindBridge ticked most of those boxes.

How did you go about implementing it?

We did a trial and six months later, we signed a contract to use it on all of our audits. There was a team empowered to go and find something, and when we found it, we rolled it out. We don’t have year-long trials like some firms. If you can invest that much time into a trial, it’s either working or it’s not, very quickly. There are very specific reasons that we will grant someone [in the firm] with a dispensation. And that will be because the data simply cannot be extracted in a sensible format. it’s not just that we have a ‘one firm’ approach. It’s about our clients – it doesn’t matter whether you’re big or small, you’re going to get the same access to the same quality.

How did it compare to the way you did audits before using AI?

Before, we would use old-fashioned random sampling techniques. So you’d have your population and we would just pick a random sample. Whereas what has changed is we put all of the transactional information in, and then it tells us what it deems to be the riskiest transactions and we sample those. It’s a targeted, focused sample. And obviously, that’s got to be better than the needle in the haystack approach.

Did it require much training or changes in the way you worked internally?

Even though we’re a big client for MindBridge, apparently we’re one of their lowest users in terms of their training and support. It picks your sample, and then you’ve still got to go and do the same audit work. Also, regardless of whether it finds fraudulent transactions or not, it always gives me something to talk to my clients about.

Can you give an example?

There are about 30 control points it looks for. One of the control points, for instance, is weekend postings. Sometimes clients working over the weekend is expected, sometimes it’s not. If the client works all weekend all the time, you can say I’m not going to put much weight on this weekend posting, because that’s just normal. You can make it really bespoke to your clients. I had a client where it wasn’t normal for them to work at the weekends. Even though there were no issues with the transactions, they weren’t fraudulent at all, it did highlight that their finance team were having to work over the weekend, and they weren’t aware of that. That extra strain, even though it hasn’t resulted in misstatement – and misstatement can be fraud or error – when you’re working over the weekend because you’re stretched, that’s when you’re more likely to get errors arising. And they said, okay, we’ll look into it. Where it’s flagged a transaction because of the detail, the narratives that they have used, in instances it’s either been blank, or it’s been insufficient, or the words have been things like ‘miscellaneous’. You can put those keywords in, and say these are the words that I want to look for. We can talk to our clients now, before MTD gets rolled out any further, and say this is what HMRC is looking for – can you get your teams to be a bit more specific?

What are the advantages of using MindBridge Ai, compared to before?

It’s about getting insights from data, and the value of that to a business to help it thrive. And therefore, it’s important to me to educate my clients on improving the quality of their data search so they can get that insight. So those insights are part of the product as well, which just from another old sort of audit file, I wouldn’t have had.

What has the client reaction been like?

There’s probably a natural assumption that most people are using things like this already and I think if you surveyed most business owners, they would be surprised at the lack of adoption of technologies like this. But they do appreciate that although they’re SMEs, that technology is available to them, even though they’re not being audited by the Big Four.

What difficulties have you had, and has it saved you any time?

We told MindBridge at one of their conferences two years ago that ingestion was the problem. This is a problem across the whole profession. We all have our clients’ data, and it is disparate data, it’s in different locations. It’s not rationalised. It’s in whatever format it comes out of the client system in. We knew when we started on this journey that would be an issue. GDPR came in right in the middle of our implementation, so people were a lot more aware of what was happening with their data. Part of the AI element of MindBridge is that it learns from the whole population, all the transactions, to help it define what is a normal transaction. We pushed them quite hard to make sure they had all the security certificates we required. For one client, we had to just install an on-premise, isolated instance of it; they wanted to use that technology, but they didn’t want their data anywhere near anybody else’s. We made that work.