The Automation Gap Is a Finance Workflow Problem, Not a Spreadsheet Problem
Recent fintech research has put a useful number on an old frustration: UK businesses are still losing several days a month to manual finance administration. The headline is easy to interpret as a spreadsheet indictment. In practice, spreadsheets are rarely the root cause. The deeper problem is that finance work is a chain of handoffs, exceptions, approvals, and evidence checks that no single…
A recent study reveals that UK businesses lose several days each month to manual finance administration. The research suggests that spreadsheets are not the primary culprit, but rather the complex process of finance workflows. Engineers building financial products should focus on the workflow rather than simply replacing a spreadsheet with a dashboard.
Consider a typical supplier payment review: someone downloads a bank statement, another checks an invoice, a third verifies if the purchase was approved, and a manager reviews any exceptions in a chat. The actual accounting entry may take only seconds, but the coordination required can consume hours. This illustrates why "automation rate" is not a reliable metric on its own. Even if 98% of transactions are automatically imported, manual issues with the remaining 2% can create bottlenecks and duplicate work.
A robust finance workflow should address four key questions for every material decision: what source records were considered, what rule or model produced the suggestion, who reviewed or changed it, and what should happen if the same event occurs again. The fourth question regarding future occurrences is often overlooked. To handle exceptions effectively, start by designing the exception path first, rather than optimizing the normal path and adding exceptions later.
A useful exception record includes fields such as an event ID for deduplication, the amount and currency of the transaction, the candidate account, the system's confidence level, evidence supporting the decision, the reason for the exception, and the current state. These fields provide essential information for reviewers to accept, edit, reject, or request more evidence. Each action should be recorded as a separate event to maintain an audit trail and enable reproducible debugging.
When integrating AI and rules engines, it's crucial to separate three layers: observed facts (transactions, invoices, contracts, and timestamps), proposed interpretations (suggested matches, categories, accruals, or anomaly flags), and accepted decisions (reviewed results that impact the ledger or reporting). This separation allows for model improvements without rewriting business history and enables precise questioning about financial decisions.
Metrics to measure the success of automation projects should focus on the minutes returned to the finance team, such as median time from exception creation to resolution, the percentage of exceptions with complete source evidence, duplicate events prevented by idempotency, number of follow-up messages per close item, percentage of accepted proposals later reversed, and time spent preparing explanations for reviewers or auditors.
These metrics reveal whether the system is genuinely reducing coordination and improving efficiency. The real opportunity lies in building a system where every important number has a source, uncertain interpretations have owners, and decisions can be clearly explained without reconstructing lengthy chat histories.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
