Kaleidoscope has warned finance teams to review how they use artificial intelligence tools at work. The warning follows research by the financial modelling software company based on a survey of 170 finance professionals.
The survey found that finance staff are adopting AI tools while still relying heavily on spreadsheet-based workflows. This combination can raise questions about data handling, version control and confidence in outputs.
Michael Gould, Founder of Kaleidoscope, said many teams are using AI to speed up modelling and reporting, but may not fully understand the implications of uploading sensitive financial information to external platforms.
According to the survey, 42% of finance teams rely exclusively on spreadsheets without a dedicated modelling platform. Another 45% spend significant time manually updating data, while 44% spend a large amount of time checking for errors.
These findings suggest many organisations are adding AI tools to existing processes rather than changing the systems that underpin financial modelling and planning. That approach can leave teams working in fragmented environments that are harder to govern and validate.
Gould said an immediate concern is whether finance workers understand how AI providers store and process the data they enter. He urged teams to look beyond basic product features and review data retention and governance terms before using these tools with company financial information.
"The number one technical skill financial planning and analysis (FP&A) professionals should be investing in right now is AI. Something has shifted very rapidly over the last few months in terms of what these tools can do, and finance professionals need to understand both the opportunities and the risks. What it can do now might look rubbish compared to your own work, but that might not be true in six months' time. The pace of change is extraordinary," Gould said.
He said this shift brings a growing risk that staff may share confidential information without realising how it could be retained or used by AI providers.
"If you stick all your company finances into an AI tool and you haven't got a subscription with a zero data retention policy, you've just provided the AI tool with learning data," Gould said.
He also called on organisations to set clearer internal rules for the use of AI across finance functions.
"Finance teams should ensure their organisation has a comprehensive AI data policy covering retention, processing and governance across all platforms in use," Gould said.
Pressure to modernise
The survey points to wider strain on finance teams as businesses become more operationally complex. It found that 72% of respondents were very or extremely interested in adopting more specialised modelling tools, rising to 91% among professionals working with highly advanced models.
For many teams, the main difficulty is no longer just building models but managing them across multiple systems, stakeholders and workflows. The report said, "Change management, not model creation, is the core friction," and identified version control, scenario modelling, data cleaning and tracing downstream impacts as key operational problems.
This suggests finance teams are under pressure to support decision-making across more of the business while still spending considerable time maintaining existing models.
"Finance should be enabling the decision-making process across the business. The challenge is that many teams are spending too much time maintaining models and not enough time analysing the future states of the business," Gould said.
He argued that many finance tools do not reflect how businesses actually operate.
"Businesses do not operate in rows and columns. They operate through products, customers, people, supply chains and operational decisions. The tools underneath finance need to reflect that complexity," Gould said.
Trust in outputs
The findings also raise questions about how finance teams validate AI-generated work. While AI may help produce formulae or automate parts of spreadsheet work, greater speed does not necessarily increase trust in the numbers if the underlying model remains difficult to check.
That concern is especially relevant in finance, where errors can affect planning, forecasting and executive decision-making. In organisations that still depend on spreadsheets and manual updates, adding AI may create another layer that needs scrutiny rather than removing existing weaknesses.
"An AI tool that generates 100,000 cell formulae in a spreadsheet sounds impressive, but how do you know it's done it right? That question of trust is going to become increasingly important for finance teams," Gould said.