This article was created in collaboration with our expert, Paweł Jędrzejewski, Senior Systems & Business Analyst at DataRiseLab.
Summary:
Companies often look outside the organization for data and AI competencies, even though they already hold some of the most important knowledge in-house: within their finance teams. Controllers, accounting, and FP&A understand what the numbers mean and the exceptions in their processes. They also know where discrepancies in data interpretation can arise and when a result actually makes business sense. This isn’t about controllers or accountants learning to code. But their involvement in a project from the very start increases the likelihood that the tools you implement will genuinely support the company’s day-to-day operations and translate into real results.
Companies looking to build up their data and AI capabilities usually do one of two things: train their existing employees or recruit new specialists. That’s a natural direction, but it often overlooks the knowledge and experience that already exist within the organization.
In many organizations, the resources needed to work effectively with data are closer than they might seem. Very often, they sit within finance – more specifically, in controlling, accounting, FP&A, and management reporting. This is exactly where the knowledge lives about where the numbers come from, what they actually mean, when they can be compared, and where it’s easy to draw the wrong conclusions when someone interprets data without understanding the underlying process.
This matters enormously, because a number can be technically correct and still lead to the wrong business decisions. Revenue can mean the invoiced amount, revenue recognized under accounting standards (Polish Accounting Act, IFRS, or US GAAP), the value of orders received, or the funds that actually landed in the account. Margin can factor in discounts, logistics costs, and annual bonuses – or leave some of them out. Even “customer” can mean the payer, the recipient, or a single record in the system. Each of these values can be calculated correctly. The problem is that not every one of them will answer the question the company is actually trying to answer.
And this is precisely where the crucial role of finance in data and AI projects begins.
Why does AI need finance?
AI can generate an answer in seconds. It can summarize data, flag a deviation, draft a comment on a result, or help with forecasting. But AI doesn’t inherently know whether a given number actually means what the business wants to read into it.
If a model receives data without context, it will work with whatever it sees in the structure of a table, a report, or a system. It won’t automatically understand that a particular margin doesn’t include annual bonuses, that the figure in a management report differs from the accounting result, or that two departments use the same KPI to mean different things.
That’s exactly why data can’t be treated as a purely technological topic. In AI projects, data architecture alone isn’t enough. You also need definitions, business rules, quality control, and ownership over how a result is interpreted. And those competencies very often sit precisely within an organization’s finance departments.
The problem arises when finance joins the project too late
In many organizations, the scenario plays out the same way. The Data team builds a model. IT integrates the systems. The business orders a dashboard, a report, or an AI solution. Only at the very end does the question come up:
Who’s going to check whether the result is correct?
Most often, the answer is: controlling. Controlling is where the questions land – about the numbers, the definitions, the differences between reports, and how a KPI is calculated. That’s usually when it turns out the data doesn’t account for important adjustments, the customer hierarchy is out of date, or two departments define the same metric differently.
Technically, the solution may work correctly. The problem is that it was built on rules no one had agreed on together beforehand.
For the business, that means one thing: instead of making effective decisions faster thanks to a newly implemented solution, they end up untangling definitional and conceptual chaos.
The biggest barriers aren’t always technological
In a global report by ACCA i CA ANZ covering 1,600 finance professionals, the most frequently cited barriers to better use of data were:
-
- data quality issues (42%),
- a lack of the right skills (42%),
- the difficulty of integrating multiple data sources (40%).
Source: Enabling finance insight: Bridging skills and data gaps for AI-enabled finance | ACCA & CA ANZ
This matters, because these barriers very often occur together. The organization has data. It has systems. It has tools. Sometimes it even has AI solutions in place. But it doesn’t have agreed definitions, clear ownership of the data, or a shared way of validating results.
In that situation, one more tool doesn’t solve the problem. It can only deepen it.
If a company doesn’t know which number is the right one, AI won’t magically make it more reliable. What it can do, very quickly, is describe that number in a convincing way. And that is one of the biggest risks for managers making business decisions.
Finance should be at the table from the start
Finance teams shouldn’t show up only at the stage of final sign-off on a report. They should be involved earlier in the project – when the company defines the problem, designs the data, and decides how results will be checked.
A controller can help pin down which decision the solution is meant to improve, which metrics actually influence that decision, and which exceptions and limitations need to be taken into account. Accounting can map out the relationships between the general ledger, source documents, period adjustments, and reports. FP&A can connect the financial result to prices, volumes, headcount, production, inventory, or customer behavior.
Only that combination makes it possible to move from data to designing a solution that delivers real business value for the company.
This is especially important in AI projects. If the goal is forecasting sales, detecting deviations, automating reporting, or supporting management decisions, the model has to work with data that carries clear business meaning. Otherwise, it will analyze numbers without necessarily answering the right question.
Does every controller need to know how to code?
The ACCA report points out that finance leaders don’t expect accountants or controllers to be widely reshaped into people who write code. What they need instead are people who can make sensible use of new tools, understand what no-code and low-code solutions can do, and know how to work with IT and Data teams.
At the same time, 72% of respondents rated their generative AI skills as basic or non-existent: 38% rate themselves at a basic level, and 34% report no experience at all.
Source: Enabling finance insight: Bridging skills and data gaps for AI-enabled finance | ACCA & CA ANZ
This reveals a real skills gap. But it doesn’t mean that finance should build the entire data architecture, write the AI models, and design the technology infrastructure on its own.
A joint team creates far greater value. Finance owns the meaning of the numbers, the control rules, the business context, and the logic behind the metrics. Technology specialists own the architecture, integration, security, and scaling of the solution. The ACCA report emphasizes that effective work with data and AI requires exactly this blend of finance, analytical, engineering, and technology competencies.
It’s at the intersection of these areas that context emerges – the kind you can’t recreate through access to data alone.
Example: before you automate a process, straighten out its logic
This is well illustrated by an example described in the ACCA report.
One organization in the real estate and hospitality sector ran its budgeting process on an Excel file containing around 50 linked tabs. Simply opening the model took 20–30 minutes, manual consolidation took several days, and different teams were working on different versions of the data.
The finance team rebuilt this process and created a solution handling over 700,000 rows of data and 15 separate business models. Opening time was cut to under two minutes, and consolidation was automated.
The most important lesson from this example, though, isn’t about the tool itself. The team didn’t start by building a dashboard. First, it split one enormous file into smaller business plans, standardized how data was entered, added the necessary analytical dimensions, and built a shared data model. Only then did it automate consolidation and reporting.
Technology sped up the process, but first finance had to straighten out its logic. That’s a crucial distinction. The value didn’t appear because a new solution was implemented. It appeared because the way of working with data had been organized beforehand, and the rules of the process were agreed on together.
The biggest opportunity for finance
This is exactly where the biggest opportunity lies for controllers, accountants, and FP&A teams.
It’s not about creating the same files faster. Nor is it about adding yet another tool to a process that’s already unclear today. It’s about finance taking an active role in defining:
-
- what exactly the metrics in use mean,
- where the result could be wrong,
- who is responsible for interpreting it,
- which decision the analysis is meant to support,
- when the result can be trusted.
This shifts finance from being the team that “checks the numbers at the end” to being a partner that helps build the foundations for data, analytics, and AI.
And that’s an entirely different level of impact on the organization.
Where can you start?
You don’t need a huge transformation program right away. It’s enough to pick one process where finance today acts as the “human interface” between systems. It might be manually merging sales files, reconciling margins, updating the forecast, or preparing a management report.
Before starting to automate, it’s worth answering five questions together:
-
- Which decision do we want to improve or speed up? Not which report we want to build, but what someone should be able to do better because of it.
- How exactly do we define the key metrics? You need to identify the source, the calculation rule, the owner, and the most important exceptions.
- Which parts of the process are done manually today? It’s worth separating work that requires business judgment from copying, reconciling, and moving data around.
- Who will check the result, and how? Validation shouldn’t appear only after the solution has been built.
- How will we know the project delivered value? It could be a shorter process time, fewer corrections, greater forecast accuracy, or a faster response to a deviation.
A conversation like this usually makes it quick to establish whether the company really needs AI, simpler automation, a better data model, or simply a shared definition of a metric. It also lets finance build its skills on a real task, instead of starting with yet another general technology training.
Finance doesn’t have to build AI. It should help build its foundations
The controller of the future doesn’t have to be the best programmer in the company. But they should know what the solution is meant to calculate, which decision it’s meant to serve, and when its result can be trusted.
Finance doesn’t have to build the model on its own. But it should be at the table before anyone starts building it. Because in a world of AI, the advantage won’t go to the organizations with the most tools, but to those that can connect technology with the right business context.
And that context very often starts precisely with finance.