Since OpenAI launched ChatGPT on 30 November 2022, the furore around the potential of ‘co-pilots’ has been immense. Since then, it has seemed like every technology company is launching their own co-pilot, regardless of effectiveness.
“On the hype cycle, we’ve passed the peak now, and we’re heading down the curve as people become more sceptical about the benefits. I think we’re going to bottom out fairly soon,” says Ian Pay, Head of Data Analytics and Tech at ICAEW.
“Then it’s about finding those use cases, and we are starting to see all sorts of different organisations figuring out what those use cases are.”
But what exactly is a co-pilot, and how will it benefit accountants? You may be aware that it involves generative artificial intelligence (GenAI). Engine B, which specialises in data solutions for professional services, describes them as “a conversational interface that leverages large language models (LLM) to support users across various complex tasks and decision-making processes.”
Digital assistant
Effectively, a co-pilot is a digital assistant. Overall, GenAI co-pilots empower accountants to focus on high-value tasks that require human judgement and decision-making, while automating routine and repetitive tasks. By leveraging AI technology, the promise is that accountants can improve efficiency, accuracy, and client service, ultimately driving better outcomes for their organisations and clients.
“It’s a clever way of highlighting that this is not ‘the pilot’. It’s supporting you, not taking over,” Pay says.
Co-pilots can automate many day-to-day repetitive tasks such as data entry by extracting information from various sources such as invoices, receipts and bank statements. They can classify and organise documents based on predefined categories and analyse financial transactions and patterns to identify anomalies that may indicate fraud or potential risks, among many other capabilities.
“The big transition came when the likes of Microsoft, AWS and Google Cloud made it technically possible for any enterprise to have private and controlled access to pretrained models in the cloud, and start leveraging them without having to make a $100m investment. That was the big pivotal moment. Co-pilots come to life when we have a way to teach them what ‘good’ looks like, so having quality data has never been this paramount,” says Hugo Pinto, VP Client Services at Engine B.
Emerging use cases
Even as these AI co-pilots evolve into real-life use cases for accountants, they are unlikely to ever replace qualified accountants because of the need for human judgement and decision-making. What they will do though is radically reduce the amount of time it takes to review and analyse mountains of documents.
Engine B is currently developing six different co-pilots for use cases such as lease accounting. It is currently working with clients to test the models in alpha mode and hopes to launch the beta version this summer.
Pinto says its first co-pilots are currently taking an eighth of the time that it would take to do the same amount of processing as a human team, while highlighting the need of a human in the loop. “It’s potentially huge in terms of transformation for the overall industry,” he says.
To be effective, in the use cases Engine B is working on, co-pilots need to have at least 85% accuracy to match human performance. Currently, Engine B’s co-pilots are at around 90% accuracy. Its lease accounting co-pilot has reached 94% accuracy. And Engine B is catching nearly half of the remaining 6% of errors, Pinto says.
Engine B isn’t the only software developer that’s set to launch AI co-pilots later this year. Sage recently announced that it was poised to unveil Sage Copilot, its new GenAI-powered assistant, later this year.
Hot on its heels came global business platform Xero, which unveiled its AI vision in the form of Just Ask Xero, or JAX for short. Its new “GenAI-powered business companion” promises to complete tasks such as generating an invoice, editing a quote or paying a bill. Xero says a JAX beta will go live later this year. And QuickBooks already has a few AI components in one or two of their products that have been released in some markets.
Internal uses
The Big Four firms and Grant Thornton have all developed in-house bespoke GenAI co-pilots based on LLMs. As yet, these tools aren’t being used to support client-facing work, but later this year it’s likely we’ll hear news of beta version launches of these co-pilots, too.
Many other firms in the industry are also exploring AI technology to improve their services and stay competitive in the rapidly evolving landscape of accounting and finance.
“It’s not a surprise the internal use cases are the ones being explored primarily at the moment. Because it’s just a little bit safer; a little bit lower risk,” Pay says.
Despite all the co-pilot announcements, detail on pricing remains scarce. Microsoft charges around $30 a month per user for its advanced Copilot version 4, and it’s likely we’ll see similar pricing for other launches.
Development of GenAI tools is incredibly expensive. And besides the levels of investment and research involved, running GenAI tools requires masses of server power too. But even if one co-pilot costs around £300 a year to use and it improves productivity by 1%, it’s arguably worth the investment.
It is hoped that this year will prove to be the year when the hype around AI tech translates into concrete real life uses that will transform industries like accountancy and free up accountants to focus on value-added judgement, analysis and decision-making. It could yet prove to be a pivotal year for co-pilots, users and clients.
ICAEW's Generative AI Guide
Getting to grips with Gen AI
ICAEW's Finance in a Digital World is a suite of elearning modules to support members in understanding digital technologies and their impact on finance, including generative AI. Completing modules can count towards verifiable CPD hours.