The Efficiency Gains and Responsibility Boundaries of Innovative Technology in EPM_News_北京智达方通科技有限公司

In the past two years, with the proliferation of generative AI in data processing, enterprises have gradually realised that while automation improves efficiency, it also introduces new risks – seemingly precise models may output conclusions that no one is responsible for. So, which aspects of enterprise management should innovative technology actually take over, and which aspects must be retained for human decision-making? This article analyses the applicable boundaries of innovative technology in EPM, common misconceptions, and how enterprises should build sustainable intelligent processes, starting from actual business pain points.

I. The Real Value of Automation: From Data Drudgery to Report Generation

The most burdensome work for corporate finance and analysis teams is often not the analysis itself, but the cleaning, mapping, and integration of data. Take budget preparation as an example: different subsidiaries use heterogeneous ERP systems, cost centre coding rules vary, and numerous Excel adjustment tables are scattered across personal computers. This causes finance personnel to spend a significant amount of time on data preparation. At this stage, rule-based automation technology can significantly improve efficiency – it can quickly read multi-source data, identify anomalous mappings, perform standardised dimensional conversion, and generate initial calculation templates.

Similarly, in drafting fixed-format documents such as monthly management reviews and board materials, technology can quickly generate first drafts based on historical data, ensuring accurate numbers and coherent phrasing. The common characteristic of these scenarios is high volume, clear rules, and well-defined error recovery paths. The intervention of technology allows analysts to shift from basic data processing to review and correction, which is the core of EPM efficiency gains.

II. The Boundaries That Cannot Be Crossed: Tacit Judgement and Responsibility Attribution

However, when faced with uncertainty judgement and cross-departmental negotiation, the capability boundaries of current technology become clearly visible. A typical scenario is budget variance analysis: when a discrepancy arises between the sales department's forecast and the finance department's calculation, the reasons may involve temporary adjustments to sales incentive policies, changes in customer payment rhythms, etc. However, technology can only present numerical differences, unable to trace the root cause of the problem. In such cases, experienced business finance personnel are needed to coordinate and communicate.

Even more critical is the issue of responsibility attribution. Any formal forecast model or budget plan requires a specific responsible person to sign off and bear the responsibility of explaining variances to management. Technology models can calculate hundreds of scenarios, but they will not face performance demerits or accountability pressure for forecasting errors. These seemingly reasonable yet unguaranteed outputs actually increase review costs. Furthermore, during training, technology models tend to fit the optimal path, easily ignoring low-probability, high-impact events, or even producing "hallucinations" that deviate from reality – the model itself is unaware of such errors, which are often only discovered through manual intervention.

Additionally, the reason many enterprises are disappointed with EPM intelligence is often not because the models are insufficiently advanced, but because the chaos in the foundational data layer has not been resolved. For example, different departments may have different definitions of "sales revenue" – does it include tax? Are returns deducted? Does it include internal transfers? If these dimensional definitions, mapping rules, and calculation standards are not unified at the system's underlying level, no high-level model can produce trustworthy results. Ultimately, the numbers output by the tool cannot be aligned across departments, and analysts still need to spend significant time manually reconciling differences; automation does not truly save labour. Conversely, if resources are first invested in unifying dimensional definitions across entities and establishing consistent master data mapping rules, and then automation capabilities are introduced, end-to-end efficiency gains become possible.

III. Pragmatic Division Principles and the EPM Implementation Path

Based on the analysis above, enterprise EPM leaders can follow these principles for division of responsibilities: allocate technology to work steps constrained by data throughput, including data ingestion, mapping conversion, calculations based on clear business drivers, and the visual presentation of standardised KPIs. Allocate to finance professionals those steps involving uncertainty, requiring subjective judgement, and carrying responsibility, such as investigating the business reasons behind variances, responding to non-routine events, facilitating cross-departmental alignment, and signing off on and explaining final numbers.

Leading practice over the past two years shows that building an "AI-ready" reporting layer is more practical than procuring high-order models. This specifically includes: standardising cross-entity master data mapping, unifying core dimensional definitions, and cultivating a finance team capable of effectively validating and challenging automated outputs. From a system support perspective, enterprises need an EPM architecture that balances two capabilities: robust data integration and automation processing on one hand, and sufficient flexibility for manual intervention, adjustment, and final sign-off on the other.

The design of Intcube EPM follows this logic – its multi-dimensional database engine supports automatic extraction, cleaning, and mapping from heterogeneous data sources, significantly reducing manual workload in the data preparation phase. Simultaneously, for steps requiring professional judgement, such as budget preparation, forecast adjustment, and variance analysis, the system retains complete process traceability and manual override permissions. This ensures that all automatically generated numbers are traceable, challengeable, and correctable, with the final version always confirmed and submitted by the responsible person. This architectural concept of "automated processing, human responsibility" closely aligns with the actual division of responsibility needs in enterprises.

The role of innovative technology in the EPM field is not to replace humans, but to strengthen those repetitive work steps that originally had lower value. From data cleansing to draft generation, automation has truly compressed process time. However, analysing the business substance behind budget variances, cross-departmental communication and coordination, and taking responsibility for the final numbers must still be done by humans. For enterprises, the rational path is not to seek an all-powerful system, but to establish a clear human-machine collaboration interface – this is precisely the substantive progress in the current evolution of EPM from process automation to intelligent assisted decision-making, and it is a practical pathway for enhancing the value contribution of the finance function.