TL;DR: Data gathering. Reconciliation. Model maintenance. Version control. These are the tasks that consume most of the finance function's capacity, and none of them produce understanding. Here is what the stack was supposed to fix, and why it did not.
McKinsey's research on digital transformation of the finance function is direct about the problem: you cannot be a strategic partner to the business if you are spending eighty percent of your time on reporting and reactive analysis. That is not a performance critique. It is a structural diagnosis.
The finance function at a typical mid-market company employs capable, trained, often highly experienced people. They work long hours, particularly around close and board cycles. Their output, the reports, the variance analyses, the forecast updates, is accurate and professionally presented.
And most of it is work the business never sees as insight. It sees the output. The slide. The commentary. The number in the board pack. What it does not see, and cannot benefit from, is the vast majority of the effort that produced that output: the data gathering, the reconciliation, the model rebuilding, the version control, the error checking. That work is invisible to the business because it produces nothing for the business. It is the overhead required to reach the analysis, not the analysis itself.
The question worth asking is not how to do that overhead more efficiently. It is why it represents eighty percent of finance capacity in the first place.
The Tool Investment That Did Not Change the Shape of the Work
Finance has been investing in technology for decades with the explicit goal of reducing the overhead and freeing capacity for analysis. ERPs replaced manual ledger entries. Dashboards replaced manual report distribution. Planning tools replaced sprawling Excel workbooks. Each generation of technology promised to push finance up the value chain.
The AFP's 2025 research found that one hundred percent of FP&A professionals still use spreadsheets at least quarterly. Gartner forecast that by 2026, more than seventy percent of finance organisations would have moved away from Excel as their primary planning tool. The gap between those two data points is instructive. The planning tools were adopted. Excel persisted alongside them. The overhead was not reduced. It was replicated in a new format.
This is the pattern that defines the last twenty years of finance technology investment. Each new tool genuinely improved the specific step it was designed for. ERPs made transaction processing faster and more accurate. Dashboards made data more accessible. Planning tools made budgeting less painful. But the work that sits between the data and the insight, the reconciliation, the model building, the version management, the error checking, remained largely unchanged. The tools optimised steps. The overhead structure did not change.
McKinsey's finance transformation research suggests that automating finance processes can free up thirty to forty percent of team capacity. That figure represents the improvement available within the existing workflow architecture. It is a real gain. And it still leaves the majority of finance time going to work that produces no direct strategic value.
Where the Time Actually Goes
Breaking down finance capacity in a typical mid-market finance team produces a picture that most CFOs recognise immediately and rarely make explicit.
Data gathering and extraction from source systems takes a significant share of analyst time. In a business running an ERP alongside a data warehouse and various operational systems, the same data often exists in multiple places in slightly different forms. Before analysis can begin, finance must establish which version is authoritative, extract it cleanly, and confirm it is comparable to the prior period basis.
Reconciliation follows. Data from different systems rarely agrees on the first pass. Timing differences, accrual treatments, intercompany eliminations, and reclassifications all require manual resolution before a comparison is meaningful. KPMG's research found that eighty-seven percent of finance teams report spending significant time reconciling data discrepancies rather than generating insight. This is not a data quality problem that technology has solved. It is a structural consequence of data living in multiple systems with different update schedules and accounting conventions.
Model maintenance is the third major category. In most mid-market finance functions, the planning and analysis models are built in Excel by specific individuals, and they require ongoing maintenance as the business changes. New product lines require new rows. Reorganisations require restructured hierarchies. Assumption changes require formula updates. The model that was built for last year's business shape is not quite right for this year's, and bridging that gap requires time.
Version control and error checking complete the picture. Finance teams routinely manage multiple versions of the same model, the original, the one the CFO commented on, the one updated after the leadership meeting, the final version that went to the board. Managing that version history and ensuring the right version is the basis for every subsequent analysis is administrative work that adds no analytical value.
The Incremental Improvement Trap
The response most organisations make to this picture is to invest in tools that make each category of overhead more efficient. Better data integration to reduce extraction friction. Automated reconciliation to reduce manual matching. More sophisticated Excel models to reduce fragility. Better version management to reduce confusion.
These investments produce real improvements. The reconciliation that took five hours now takes two. The model that broke every quarter has been replaced with something more robust. The version confusion that produced two incorrect board packs has been addressed with better controls.
But Gartner's finding that only fifteen percent of FP&A teams operate with a sustainable delivery model, despite a decade of this kind of investment, reflects the ceiling of the incremental approach. Making the overhead more efficient does not change what proportion of finance capacity goes to overhead. The eighty percent is still eighty percent. It just gets there faster.
The structural shift requires a different question. Not how to do the overhead more efficiently, but how to move the overhead out of the finance function's primary workflow entirely.
What Changes When the Overhead Is Carried by the System
The opportunity that AI-native FP&A represents is not faster reconciliation or smarter model maintenance. It is the movement of the overhead out of finance's cognitive work and into the system's background processes.
When data gathering is continuous rather than triggered by a request, finance does not start analysis by extracting data. The system has been ingesting and reconciling data from source systems as it arrives. When a question comes in, the data is already clean and comparable.
When actuals are monitored continuously rather than reviewed at period end, reconciliation differences are identified and resolved as they occur rather than accumulated into a month-end task. The five hours of reconciliation at close compresses because the discrepancies that would have been found then were found and addressed earlier.
When context is carried forward from cycle to cycle, model maintenance shrinks because the system remembers the adjustments made in prior periods and applies them consistently. The analyst who would have spent three hours rebuilding last quarter's adjustment logic spends thirty minutes confirming that the logic still applies.
This is not a vision of fully automated finance. It is a reallocation of where cognitive effort goes. The judgment, the interpretation, the commercial understanding, the strategic framing: those remain with the human. The overhead, the gathering, reconciling, maintaining, and version-managing, transfers to the system.
When that transfer happens, the eighty percent that goes to overhead shrinks. The twenty percent that reaches the business as insight grows. The finance function does not get bigger. It gets directed toward the work that actually matters.
The Practical Starting Point
The transition from an overhead-heavy to an insight-heavy finance function does not start with a transformation programme. It starts with identifying the single workflow where the overhead is most disproportionate to the insight produced.
In most mid-market finance teams, that workflow is the post-close variance explanation. The data gathering, reconciliation, and model building required to explain a monthly variance consumes two to three days of analyst time and produces a document that reaches the business a week after the period ended.
Start there. Move that specific workflow to a continuous model. Measure how much overhead compresses when explanation forms as events occur rather than after they close. When the answer to that question is visible on real data, the case for extending the model throughout the function makes itself.
The eighty percent does not have to be permanent. But it will stay permanent as long as the workflow is designed to create it.
Uptio connects to ERPs and transactional source systems and carries the overhead, so finance can carry the insight. It gathers, reconciles, and contextualises actuals continuously, and combines internal financial signals with external market intelligence, so finance teams can do the work the business actually needs. Learn how Uptio works.