Matej Trbara, Co-Founder, Farseer
Finance teams have spent years building out their technology stacks. And yet, humble Excel spreadsheets remain deeply embedded in how finance actually operates, connecting systems and filling the gaps between them.
In many organisations, this has created an environment that works, but only because finance teams make it work. People know which numbers need reconciling, where the latest assumptions sit, and why the same metric might look different depending on where it is being reported.
Now AI is being introduced into that environment, with the promise of faster forecasts, instant analysis and better decision-making. But AI is also becoming an exposure machine for finance. Connect it to a fragmented finance stack and it won’t make the inconsistencies underneath disappear – it will surface them faster.
For CFOs, that changes the question. It’s not simply how quickly they can adopt AI, or even whether their data is clean enough. It’s whether the financial model beneath that data, from the definitions, assumptions and business logic that determine what the numbers actually mean, is consistent enough for AI to rely on.
Clean data isn’t the same as a consistent financial model
Finance has no shortage of data. The harder problem is establishing a consistent understanding of what that data means and how it should be used.
Financial planning depends on definitions, assumptions and relationships that aren’t necessarily captured in the underlying data itself. Something as familiar as revenue can be defined differently across systems or reports. Forecasts may use different assumptions. Operational plans might not connect cleanly to the financial model.
The problems start to emerge more clearly as AI enters the picture.
Ask an AI assistant for your expected EBITDA margin in 2027 through an FP&A platform and it might give you one answer. Ask another assistant connected to the data warehouse and you could get another. Neither answer necessarily means the AI has malfunctioned. The two systems may simply be using different assumptions, definitions or versions of the forecast.
This is why “clean the data” isn’t a sufficient AI strategy for finance. A warehouse can contain accurate, well-structured data while the organisation still has multiple definitions of annual recurring revenue or different logic for how headcount translates into costs.
Those aren’t simply data-quality problems. They are questions about how the organisation defines and models its business. AI will inherit those inconsistencies unless finance resolves them first.
AI exposes the spreadsheet glue holding finance together
If these inconsistencies already exist, why haven’t they stopped finance teams from operating? The answer, in large part, is people.
Finance professionals have become extremely good at navigating fragmentation. They know that one figure needs adjusting before it reaches the board. They know which spreadsheet contains the latest forecast assumptions. They understand why the number in the ERP doesn’t quite match the one in the planning model.
Spreadsheets are often the glue that makes this possible. They sit between systems and allow teams to reconcile data, adjust assumptions and encode business knowledge that doesn’t exist anywhere else.
This isn’t simply a technology problem. It is a people and process problem. Much of the context required to understand financial information remains implicit: inside spreadsheet formulas, manual workflows or the knowledge of individual employees.
AI challenges that model because we increasingly expect technology to do more of the interpretation itself. But AI doesn’t automatically inherit the context that finance teams have spent years applying manually. When it starts working across fragmented systems, the inconsistencies that people previously reconciled in the background become visible in its answers.
In that sense, cracks in financial infrastructure that might have been hidden before are now being exposed by AI.
AI readiness starts underneath the AI
Becoming “AI-ready” actually requires finance leaders to look beneath the AI layer, at the conflicting definitions, assumptions and business logic that AI can amplify at pace.
The goal shouldn’t be to eliminate every spreadsheet or force the entire finance function into a single system. Modern finance will continue to depend on multiple tools. What matters is whether those tools are working from a consistent understanding of the business.
That requires finance teams to make more of their implicit knowledge explicit. Core metrics need agreed definitions, business logic needs to be governed rather than duplicated across different systems and spreadsheets, and assumptions need clear ownership. Operational drivers such as headcount, volumes or pricing need to connect consistently to their financial consequences.
This is the distinction between having clean data and having a reliable financial foundation. Technology can provide the infrastructure, but finance still needs to decide which definitions are authoritative, how planning processes should work and where responsibility for maintaining that logic sits.
AI can only accelerate what sits underneath it
There is understandable pressure on CFOs to establish an AI strategy now. But the value of that strategy will depend less on how quickly another layer of technology is added and more on whether AI has a consistent financial model to work from.
With that foundation in place, AI can help teams interrogate forecasts, understand variances and explore scenarios without requiring someone to manually reconcile the context each time. It can give finance teams faster access not just to information, but to a shared understanding of how the business works.
Without it, faster isn’t necessarily better. AI can retrieve conflicting definitions faster, apply inconsistent assumptions faster and produce contradictory answers faster. Instead of improving financial decision-making, it risks scaling the inconsistencies that finance teams have spent years working around.
Finance teams have long compensated for fragmentation through people, processes and spreadsheets. AI is now exposing just how much hidden logic those workarounds contain. The opportunity for CFOs isn’t simply to put AI on top of the existing stack. It’s to make the financial logic underneath it consistent enough that AI accelerates the right things.