For accountants who are unsure of assistive technology, such as Copilot, Ian Schnoor, Executive Director at the Financial Modeling Institute, offers simple advice: “All accountants and all finance professionals should start playing around and try to use Copilot in their spreadsheets.”

He describes embracing this technology as “kind of mandatory”: “I know it is new and it is scary and it's changing very quickly, but an accountant who refuses to start using Copilot today is similar to an accountant who refused to use a computer and a spreadsheet in the ‘80s or ‘90s.”

What changed in the latest CoPilot update

Upgrades to Copilot that went live in February 2026 have enhanced the use cases for the generative AI chatbot solution. As well as spreadsheet functions, Schnoor says the main use cases for Copilot include building audit files, automating workflows, and creating forecasting and budgeting tools. He explains that it uses a multi-LLM strategy, incorporating Anthropic's Claude and OpenAI's ChatGPT, rather than being an LLM itself.

“Until recently, Copilot in Excel was essentially like having an AI agent right in your Excel, and it could give you ideas and check your work, but it wasn't building anything up until February,” Schnoor continues. “Now Copilot can be used to build and check things – it's pretty wild because you can give it an instruction and then watch it build a spreadsheet for you.

“It can actually create multiple Excel files simultaneously, which a human cannot do; it can work in more than one file.”

Rather than viewing Copilot as a general assistant, Schnoor urges accountants to treat it as a custom architect for your daily operations. By mapping out existing workflows, accountants can systematically replace tedious manual entry and data migration with streamlined, automated sequences.

“What I actually encourage people to do is think about their own workflows, the things people do all day, every day – you've just been doing them for so long, it's just second nature,” he says. “I encourage people to dissect and map out the particular processes they engage in, and then start experimenting with how they could automate some of these things using Copilot.”

How do you choose Claude or ChatGPT?

In terms of Claude and ChatGPT interacting with Copilot, Schnoor says it is important to understand that Copilot “sits on top” of the LLMs. Accountants have the option of choosing whether to use Claude or ChatGPT with Copilot or let Copilot choose, based on which one it “thinks” is better for the task. 

Just as Schnoor encourages accountants to experiment with different tasks on Copilot so they become accustomed to the technology and its benefits, he says it is important to try the latest versions of Copilot and test out whether Claude or ChatGPT works best for their needs.

“I don't want to say which one is better, because they're better for different things – and they're changing so quickly that whatever we talk about today could be very different in one week from now or a month from now,” says Schnoor. “I'm reading about more people deploying a multi-LLM strategy, whether they are writing something or designing a visual space or building a spreadsheet. A lot of people build with one LLM and get another one to check it.”

Use ChatGPT for data, Claude for documents

In general, ChatGPT is a solid Copilot partner for advanced data analysis, as well as having excellent vision capabilities for scanning and interpreting handwritten or unclear documents. Claude is known for offering strong writing skills, especially for sensitive documents and emails, as well as great capabilities with spotting discrepancies in large quantities of data.

While the possibilities for Copilot via Claude or ChatGPT are exciting, Schnoor says the technology still needs to be approached with a degree of caution while keeping expectations realistic.

“Copilot’s capabilities as a really effective tool are still quite new and you must start experimenting to stay sort of on the curve, but be cautious and be aware,” he says. “We’re seeing a lot of people experiencing AI burnout. There's a lot of hype all over LinkedIn and social media that says, ‘Hey, build your model in five minutes when it used to take you eight hours.’”.

While Copilot has great time-saving potential, “it's not going to give you quite the extreme savings in most cases that some of the hype sellers are selling,” Schnoor says.

Human oversight is key

Schnoor stresses the importance of human oversight, combined with continuously valuing human skills, with regards to solutions such as Copilot.

“I strongly believe human oversight is going to be mandatory [for AI technologies] and the other thing that is more important than ever is human skills,” he says. “I believe the only way to challenge and push when [AI solutions] hallucinate and make mistakes and go in the wrong direction. The only way to understand and validate the work of an AI agent is to fundamentally have the skills.”

Schnoor compares this need for accountants to still oversee and apply their skills when using AI to the changing role of computer programmers: “Good programmers are doing very little programming anymore, but they need to be able to read, check and audit everything AI is doing and the only reason they can do this is because they already have strong programming skills.”

“The exact same thing is true for accounting and finance professionals,” Schnoor continues. “You can use AI to build budgets, models and other tools, but you should only be doing that if you already have the skills and the knowledge, because otherwise it's going to generate work that you have no capacity to evaluate.”