Founded in 2019, the original goal of data solutions firm Engine B was to create a data platform that automates testing and analysis with standardised data from all ledgers, from any client system. That job has been completed. However, the pace of technological advances, particularly in generative artificial intelligence (AI), means that Engine B is now working on its next project to further transform audit and accounting.
Those that have indulged in ChatGPT or similar AI models will undoubtedly confirm just how impressive AI currently is. But the version that most people are playing with is version 3.5. Next generation AI, or version 4.0, is even more extraordinary: it has the ability to create a ‘mental model’ of a problem and help solve it.
In Microsoft’s words, this version, Copilot “combines the power of large language models with your data in the Microsoft Graph and the Microsoft 365 apps to turn your words into the most powerful productivity tool on the planet”.
Shamus Rae, Chief Executive Officer and co-founder of Engine B, says: “The creation of a mental model is profound from a technology point of view, especially as it happened by accident. Nobody knows how it’s done – it wasn’t trained to do that. It’s unbelievably impressive. Copilots are the kind of things that sit on top of the GPT model to help you do things like code or create a presentation.”
Copilots rely on good quality data, which is predictable and understandable, and on knowledge graphs. What Engine B is now working on is to build up quality data and a series of knowledge graphs to allow accounting firms to create their own copilots.
“So that’s what we're doing – working with different accounting firms to come up with ideas for copilots,” Rae says.
The potential for this technology is huge and Rae expects to see AI co-pilots appearing at firms by the end of the summer, saving them thousands of hours.
For example, in the audit process called related parties where auditors check to see if somebody in the business has a connection to a client or a supplier that isn’t appropriate, Rae fed into Engine B’s copilot: ICAEW’s technical sheet on related parties, Engine B’s house data for that situation and its client list. It then ran a related party check and the co-pilot came back with results in less than two minutes.
“GPT has also passed the American CPA exam with an 85% pass rate. So, it’s a reasonable accountant already,” Rae says.
Like most advanced AI tools, the cost of using a GPT copilot will initially be pricey because firms will want to have their own private copy of GPT so that their data is safe and kept segregated. Nonetheless, Rae says as more and more people adopt the technology the price will come down.
“There’s a lot of fear about data security, but actually if the firms have their own copy of GPT running, the security is fine. It’s not cheap, but worth it. We’ve been waiting for GPT to unlock the next stage and it’s now unlocked with co-pilot,” he says.
Engine B, Rae will tell accountants at the ICAEW Annual Conference, will be building co-pilots for things like related parties, transfer pricing, due diligence and groupings, among others, that will save firms “days and days of audit time”.
“If you’ve got 1,000 medium-sized audits, you’ll probably save 1,000 days. So, the impact is quite big. And obviously for large audits the impact is huge.”
Are you ready for the revolution?
Find out more