From the Desk: TRUST. Accounting and Artificial Intelligence.

I think AI could be the most discussed topic in the world currently and to be completely honest, I’m struggling to define it, so I asked ChatGPT for a definition. “Artificial intelligence is the ability of a machine or computer program to simulate aspects of human intelligence in order to perform tasks, learn from experience, and adapt to new information.” I think this definition is so interesting, particularly with saying “the ability of a machine to simulate aspects of human intelligence”. When the “simulation” of human intelligence isn’t perfect, the consequences depend entirely on why you’re using it. In the accounting industry, there are concerns about being able to trust AI’s ability to perform tasks and it’s clear why. In accounting, being almost right is the same as being wrong.

Most AI tools I use, and maybe you do as well, are built for general purpose use. Think ChatGPT, Claude, Gemini. They’re trained to “simulate” human intelligence across a wide range of tasks and give fast responses. These are “black box” AIs, meaning they take a prompt and give an answer, however how they got to that answer is unknown. AI like this doesn’t work for accounting. These AIs occasionally hallucinate and give faulty responses that would be detrimental to a business’s chart of accounts, financial records, or audit trail.

However, I’ve learned that Sage’s approach to AI is different, not only with the technology but more importantly, the philosophy behind it.

It all started back in 2017. AI was embedded in the background of Sage software, detecting anomalies, reading invoice data, and learning from user corrections. Sage spent the better part of a decade learning what accountants truly needed in an AI. They need an AI that catches human mistakes, handles mind numbing tasks, and most importantly, stays completely transparent about everything it’s doing. Then, in 2024, Sage Copilot was launched.

Sage Copilot is a “glass box” AI, meaning that every result is traceable to specific rules or original data sources and it is fully transparent with how it came to its answers. This allows users to easily check the accuracy of results and allows them to have confidence in its answers, making it genuinely useful for accountants.

With most general AIs being “black box”, you may get useful answers, but you can’t fully trust it. That’s where Copilot differentiates itself and becomes truly useful for accountants. Trust. Trust in its answers, trust in what it does with user data, trust in its transparency.

So in my opinion, accounting doesn’t need AI to do all the work or be completely revolutionary, it needs AI to be trustworthy.

Love Nikhil, Summer Intern, nuvemXP