Artificial Intelligence (AI) has forged significant inroads into the daily workflows of finance professionals, including Excel. From Microsoft Copilot to third-party tools like ChatGPT and Claude, there’s a growing set of capabilities now available to support, automate and accelerate spreadsheet-based tasks. However, with opportunity comes responsibility, particularly around data privacy, accuracy and readiness.

Internal or external tools?

AI tools in Excel generally fall into two broad categories: integrated solutions such as Microsoft Copilot, and external tools like ChatGPT, Claude or other large language models (LLMs).

Copilot, part of Microsoft 365, is built into Excel itself. It includes an embedded AI assistant that understands spreadsheet structure, generates formulas, highlights inconsistencies and even creates charts. It also offers a chat-like interface similar to ChatGPT, allowing users to ask questions in natural language.

External tools like ChatGPT and Claude can perform similar tasks via uploaded Excel workbooks or described data problems, but they come with a trade-off: data privacy. Uploading sensitive or proprietary Excel data to these platforms could mean your files are processed or even retained for future model training, depending on the provider's terms.

With Copilot, Microsoft states that your data remains within your organisational boundary and is not used to train its models. However, users should still confirm this with their organisation’s Microsoft license and compliance policies.

“Where Excel users might be concerned around AI is if you're using an external tool, such as ChatGPT or Claude. There's a good chance that whatever you upload could be used in their training for their large language model,” says Mark Proctor, Twenty Seven Solutions director and Excel Off the Grid founder.

“If you want to use those kinds of AI external tools, you have to be very careful about what's included in your Excel document. With Copilot, because it's baked into your Microsoft product and Microsoft says that they don't use anything you upload for its training, there’s a greater element of trust,” Proctor says. 

Using AI for formulae

Used correctly, AI can greatly accelerate Excel-based tasks for two key audiences: novice users and advanced users.

For example, Excel Copilot can help build formulas, create charts, summarise data and check for mistakes without needing to understand Excel’s more complex composition. A user can simply type, ‘create a bar chart showing monthly sales by region’, and Copilot can automatically interpret the existing data and produce an illustration. With supervision, this can be a useful learning tool.

Equally, expert users can benefit by using AI to speed up repetitive tasks, test new ideas or refine formulas. Even if it returns an imperfect result, experienced users can quickly identify and correct issues, saving time overall.

However, the risk lies in the “dangerous middle”, Proctor warns. That is, users who have some Excel knowledge but not enough to confidently corroborate AI-generated outputs. These users may copy, trust and implement formulas or analyses without testing them thoroughly, leading to errors and poor decision-making.

“There's the dangerous middle-ground of people who don't have enough knowledge of Excel but trust what's given them. They haven't interrogated it enough, they haven't tested it enough and therefore it gives them problems,” Proctor says.

This issue is not new. For years, Excel users have copied formulas from forums or templates without full understanding. AI just makes it easier and faster to generate potentially flawed content at scale, he warns.

Creating sentiment analysis for your spreadsheet

Qualitative insight such as sentiment analysis may be more common in marketing or HR, but finance professionals can now use AI tools to apply similar text approaches for creating meaningful narrative to support the numbers and explain trends.

AI can now support every stage of the spreadsheet lifecycle from creation and review to analysis and functions and narrative.

“When you're processing a huge amount of qualitative data, even if AI gets a few things wrong, it isn’t always a problem because you’re just trying to get a sense. So even with some false results, you can still get a feeling or a trend, which you wouldn't have been able to do at scale without AI,” Proctor says.

Using tools such as Power Query

Despite the potential, AI is only as good as the data it receives. In finance, data often arrives from disparate systems ranging from accounting or payroll to inventory or CRM – each with its own format, naming conventions and quality issues.

That’s where tools like Power Query become critical. Power Query helps users clean and aggregate data from multiple sources, preparing it for accurate analysis and AI interpretation. Yet, despite being built into Excel, many finance professionals aren’t using it.

“In the finance space, that's often a challenge. I'd say that we're probably at only 30% or 40% of Excel users using tools like Power Query to clean data, and clean that data fast. Otherwise, they tend to use more manual processes and therefore people's ability to use AI is then reduced,” Proctor says.

He explains the reason for this is habit. Many users believe they “know” Excel, but their knowledge may be based on tools and methods from more than 10 years ago. “Much of Excel education is passed down informally between colleagues, meaning new features like Power Query go undiscovered unless someone actively seeks them out or attends modern training.”

According to a 2020 user survey (based on a small sample), Power Query users reported saving between 22 and 89 working days per year, with an average of 76 days saved. While early adopters in that study may not have been finance professionals, even a conservative estimate of time saved represents a major efficiency gain, Proctor says.

AI-ready finance function

The key to realising AI’s potential in Excel is about thinking beyond debits, credits and charts of accounts. Users must understand data structures, quality and flows.

This doesn’t mean abandoning accounting principles, but rather blending accounting principles with data literacy and understanding how structured, clean data ensures faster, more accurate insights.

AI can’t consistently fix bad data. However, when finance professionals pair strong data preparation with tools like Power Query and Copilot, they unlock the ability to move from wrangling spreadsheets to delivering strategic insight.

AI capabilities now embedded in Excel, particularly through Microsoft Copilot, offer a transformative opportunity for finance teams. Used wisely, these tools can reduce manual work, improve accuracy and help teams focus on insight rather than process. 

Those who embrace these tools early and responsibly are well-placed to lead the next generation of data-driven finance.