> **来源：[研报客](https://pc.yanbaoke.cn)** # BCG Executive Perspective: Applying Agentic AI in the Finance Function for Transformative Impact ## Introduction Agentic AI is now a top priority for CFOs, transitioning from a productivity tool to a strategic imperative. With the ability to reason, plan, and execute multistep workflows with minimal human input, the focus for CFOs in 2026 is not just on whether AI will reshape finance, but how quickly and broadly they can act. While many finance functions are still in the "copilot" stage, early adopters are already reaping significant benefits. ## Core Content and Key Insights ### The Strategic Imperative of Agentic AI - **CFOs’ view**: 88% of CFOs consider AI essential, and 96% expect transformative impact within five years. - **Future vision**: The 2030 goal is "always-on finance," with real-time closing, continuous forecasting, and autonomous optimization. - **Role evolution**: Agentic AI shifts finance from number producers to value architects, enabling faster, better, and more efficient operations. ### Value Unlock from Agentic AI - **Current capabilities**: AI agents are already being used in FP&A, accounting, and treasury for multistep workflows. - **Value realization**: Early adopters have unlocked up to 30% value, with long-term potential exceeding 50%. - **Key areas**: - **Planning, budgeting, and forecasting**: 20%-30% efficiency gain - **Reporting and business intelligence**: 30%-40% efficiency gain - **General accounting**: 15%-25% efficiency gain - **Finance operations**: 20%-30% efficiency gain - **Expert functions**: 5%-15% efficiency gain ### Finance's Unique Design Requirements - **Controls and security**: Finance must maintain high standards (e.g., SOX, GAAP, IFRS), including segregation of duties. - **Accuracy and audit readiness**: Transparency and explainability are essential to ensure reliable AI use. - **Data access and consistency**: Finance must ensure a single source of truth and maintain control over calculation logic. - **Change management**: Finance teams are typically conservative and slow to experiment with AI, requiring support and training. ### AI Design Considerations - **Model integrity**: Ensure compliance with automated error and risk thresholds. - **Human oversight**: Define agent objectives and implement human-in-the-loop workflows. - **Explainability**: Avoid black box models; allow human override. - **Rules and triggers**: Clearly reference policy documents and implement guardrails. - **Encouraging experimentation**: Provide tools, certifications, and cross-functional teams to support innovation. ## Use Cases and Impact Stories ### Case Study 1: AI-Driven Finance Transformation - **Before**: Manual, siloed processes limited strategic output and decision speed. - **After**: AI-first finance teams saw: - **15%+ FTE capacity reduction** - **90%+ reduction in reporting time** - **25%+ more accurate forecasting** - **Proactive insights** enabling cost savings and avoidance. ### Case Study 2: FP&A Insights Agent - **Before**: Slow, fragmented reporting and analysis. - **After**: Enabled: - **80%+ faster ad hoc analysis** - **2x–4x faster reporting generation** - **Enhanced data-driven decision making** with real-time P&L impact modeling. ### Case Study 3: Touchless Procure-to-Pay Process - **Before**: Manual, error-prone, and slow invoice processing and vendor onboarding. - **After**: AI-enabled process: - **80%+ touchless invoice entry** - **~45% FTE capacity unlocked** - **15+ days reduction in onboarding time** - **~50% reduction in onboarding errors** ## Getting Started: AI Transformation Journey To capture the full value of agentic AI, CFOs must reimagine processes, tech, data, operating models, and governance simultaneously. Key steps include: 1. **Process reimagination**: - Redesign end-to-end workflows with AI in mind. - Clarify agent execution and human intervention points. 2. **Tech ecosystem redesign**: - Deploy LLM and coding platforms to enable citizen development. - Layer orchestration platforms to shift to agentic workflows. - Assess existing tech stack and build new tools where needed. 3. **Data foundations**: - Build a business context fabric with data relationships and driver trees. - Unify taxonomies and clean master data using GenAI. 4. **Operating model redesign**: - Redefine roles, OKRs, and incentives to align with AI goals. - Refresh talent and upskill teams for future capabilities. 5. **Governance**: - Define guardrails for responsible AI (e.g., no-AI zones). - Establish frameworks for AI roadmaps, prioritization, and investment. ## Conclusion Agentic AI is rapidly transforming the finance function, enabling it to move from manual, siloed processes to autonomous, real-time operations. The shift requires a comprehensive approach, including process redesign, technology integration, data management, and governance. CFOs who act early and strategically can unlock significant efficiency gains and elevate finance to a strategic value-creation role.