SAP Embeds AI Agents to Automate 80% of Routine Finance Workflows

Finance Reimagined: How AI is Automating Every Finance Function With SAP — Ai4

Ai4

Deploying embedded AI agents to automate routine close and compliance tasks protects CFOs against looming talent retirements while shifting accountants into strategic supervisory roles.

AI-assisted summary of Finance Reimagined: How AI is Automating Every Finance Function With SAP — original source. Check the original for full context.

Source published: 2026-08-18

Detailed notes

## Background Published on YouTube by the Ai4 channel on August 18, 2026, this presentation features Mary Sibley, RVP of SAP Cloud ERP Customer Advisory. According to the page description, the session explores how SAP Autonomous Finance embeds AI agents into financial workflows to help organizations navigate escalating regulatory complexity, economic volatility, and talent constraints while maintaining human-in-the-loop governance.

## TL;DR Mary Sibley presents a framework for transforming CFO organizations through SAP Autonomous Finance, addressing macro headwinds such as e-invoicing mandates, economic shocks, and an impending demographic retirement cliff. Central to this vision is SAP's copilot, Joule, which orchestrates specialized AI agents across key finance domains—including receivables clearing, intercompany reconciliation, and financial close reporting—while keeping seasoned accountants in the loop to validate outputs. Sibley highlights that up to 80% of routine finance processes can be automated, urging enterprise leaders to bypass unwieldy multi-year IT initiatives in favor of targeting discrete operational bottlenecks with immediate ROI.

## Atmosphere & style The presentation carries a pragmatic, seasoned tone that blends executive advisory with relatable empathy for the grinding realities of traditional accounting work. Sibley draws effectively on her personal background as a former corporate controller, grounding high-level artificial intelligence concepts in relatable scenarios like eliminating stressful month-end late nights and mentoring junior staff. The pacing is clear and structured, transitioning smoothly from high-level macro pressures to hands-on dashboard demonstrations and actionable next steps for enterprise adoption.

## Key takeaways - [00:04](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=4s) Enterprise finance AI has evolved from traditional machine learning and predictive forecasting to generative AI and collaborative agentic systems that will reshape the CFO organization. - [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) Mounting regulatory requirements like global e-invoicing, macro volatility, and an impending demographic retirement cliff make legacy manual finance operating models unsustainable. - [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) A human-in-the-loop framework ensures experienced professionals validate agentic outputs, capturing both hard operational ROI and soft returns like talent retention and morale. - [06:41](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=401s) SAP Autonomous Finance utilizes Joule as a natural language orchestrator to coordinate specialized domain agents across core business data structures. - [08:54](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=534s) Autonomous finance agents are deeply embedded into enterprise ERP logic, business rules, and controls rather than serving as superficial chatbot overlays. - [08:54](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=534s) The objective of autonomous finance is capacity creation and strategic C-suite advisory, not head-count reduction in finance departments. - [12:12](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=732s) The suite shifts organizations away from rigid, periodic cycles toward continuous dynamic planning, real-time tax optimization, and continuous financial close. - [14:30](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=870s) Agentic workflows for receivables clearing, intercompany reconciliation, and journal entry accruals transform accountants from manual data processors into strategic review managers. - [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) Enterprise proof points demonstrate tangible outcomes, including daily continuous internal control testing at Pfizer and a 75% reduction in scheduled reporting packages. - [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) Enterprise AI adoption relies on transparent governance and comprehensive audit trails that record every data point, agent proposal, and human approval. - [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) Finance leaders should initiate AI adoption by surveying staff on high-friction manual tasks, piloting a single targeted use case, and avoiding bloated multi-year IT programs.

## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=4s) Evolution of AI and the Future of the CFO Organization

Mary Sibley introduces her vision of the future finance organization by reflecting on her career journey. Starting out in public accounting before becoming a corporate controller, she recalls the grueling reality of long weekends spent working late to close the books and manage year-end reporting. Joining SAP in 1994 ("don't do the math," she jokes), she developed an enduring appreciation for enterprise software and has watched artificial intelligence evolve inside the platform across decades. As Sibley notes, artificial intelligence is far from a brand-new phenomenon; enterprise finance has progressed steadily through traditional machine learning, predictive analytics, and generative AI, arriving at the current frontier of intelligent agents and assistants. Looking ahead five years, she emphasizes that the office of the CFO will be fundamentally reshaped as emerging agentic capabilities redefine how financial work is executed.

- **Career Path**: Transitioned from public accounting and corporate controller roles into SAP in 1994. - **AI Evolution**: Advanced across distinct waves from machine learning and predictive forecasting to generative AI and multi-agent coordination.

[02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) Macro Pressures, Talent Dynamics, and the Human-in-the-Loop Model

Finance teams are operating under mounting external pressures that make traditional operating models unsustainable. Sibley highlights accelerating regulatory complexity—such as mandatory e-invoicing frameworks implemented across more than 60 countries—alongside macroeconomic shocks like sudden tariff shifts, geopolitical conflicts, and raw material shortages that force organizations into rapid damage control. At the same time, the workforce is facing a demographic cliff: a massive percentage of seasoned professionals will retire within the next five years, while incoming early-career talent expects purposeful work rather than tedious manual tasks.

``` MACRO FORCES SHAPING THE MODERN FINANCE ORGANIZATION ┌──────────────────────┐ ┌──────────────────────┐ ┌──────────────────────┐ │ Regulatory & E-Inv. │ │ Macro Volatility │ │ Talent Constraints │ │ 60+ countries, rapid │ │ Tariffs, supply chain│ │ Seasoned departures, │ │ compliance updates │ │ shocks, market shifts│ │ desire for value work│ └──────────┬───────────┘ └──────────┬───────────┘ └──────────┬───────────┘ └────────────────────┐ │ ┌────────────────────┘ ▼ ▼ ▼ ┌─────────────────────────┐ │ Autonomous Finance & │ │ Human-in-the-Loop │ └───────────┬─────────────┘ │ ┌─────────────┴─────────────┐ ▼ ▼ ┌─────────────────────┐ ┌─────────────────────┐ │ Hard ROI: 80% Tasks │ │ Soft ROI: Retention,│ │ Automated / Assisted│ │ Morale, Satisfaction│ └─────────────────────┘ └─────────────────────┘ ```

Addressing this transition requires preserving deep institutional judgment while upskilling the next generation. Sibley stresses that as organizations adopt agentic AI, the "human in the loop" becomes more critical than ever. Experienced professionals evaluate AI outputs using decades of contextual knowledge, making mentorship and "buddy" pairing structures essential for developing young accountants. Furthermore, board conversations are shifting from theoretical AI discussions to demanding hard accountability: boards are asking finance departments what tangible returns they achieved from their token investments. Sibley notes that while hard dollar savings are vital, soft returns—such as improved morale, job satisfaction, and reduced turnover—are just as critical, especially when up to 80% of routine finance business processes can be fully or partially automated to eliminate persistent backlog items from employee to-do lists.

[06:41](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=401s) Introducing SAP Autonomous Finance and Joule

To tackle data fragmentation and operational friction, SAP frames its forward-looking strategy around "autonomous finance." Sibley presents a vision where finance professionals no longer spend late nights manually assembling fragmented data across disconnected systems to model scenarios or close the books under stringent sustainability and compliance demands. Central to this paradigm is Joule, SAP's natural language generative AI copilot, which coordinates underlying teams of specialized AI agents to monitor performance, analyze risks, forecast scenarios, and keep operations compliant from a single source of truth.

``` JOULE (AI COPILOT) Natural Language Interface & Context Orchestrator │ ┌───────────────────────┼───────────────────────┐ ▼ ▼ ▼ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ Receivables │ │ Intercompany │ │Journal Entry │ │ Clearing │ │ Reconciliation│ │ & Financial │ │ Agent │ │ Agent │ │ Analysis │ └──────┬───────┘ └──────┬───────┘ └──────┬───────┘ └───────────────────────┼───────────────────────┘ ▼ FINANCIAL MANAGEMENT SUITE Contextualized Data & Enterprise Business Rules ```

[08:54](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=534s) Demystifying Autonomous Finance: What It Is and Is Not

Sibley clarifies common industry misconceptions about autonomous finance. First, Joule is not merely a superficial chatbot, nor is the agentic layer an external wrapper bolted onto enterprise systems; it is deeply embedded within the core SAP software, natively understanding business data structures, controls, and enterprise rules. Sibley shares her own practical experience building analytical dashboards with Joule entirely through natural language—commanding it to establish baseline parameters, upload real-time data, slice and dice by multidimensional attributes, and export board-ready PowerPoint presentations or PDFs at the click of a button.

Crucially, autonomous finance is not about reducing headcounts in the office of the CFO. Instead, it reallocates human capital so finance leaders can focus on high-impact strategic advisory within the C-suite and finally clear long-standing operational backlogs. Sibley strongly cautions against treating AI as an unwieldy, multi-year IT transformation. Instead, organizations should identify targeted points of operational drain—where senior staff spend unnecessary hours on low-value grunt work—solve those specific use cases to achieve early wins, and methodically scale from there.

[12:12](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=732s) Architecture of the Autonomous Finance Suite

The functional layout of autonomous finance mirrors the structural domains of the modern CFO office, anchoring intelligent agents directly into daily operating controls and enterprise data structures:

``` OFFICE OF THE CFO: AGENTIC SUITE ┌──────────────────────────────────────────────────────────────────┐ │ Revenue Management & Billing Automation │ ├──────────────────────────────────────────────────────────────────┤ │ Dynamic Planning & Continuous Budgeting │ ├──────────────────────────────────────────────────────────────────┤ │ Treasury, Working Capital & FX Policy Management │ ├──────────────────────────────────────────────────────────────────┤ │ Governance, Risk, and Compliance (GRC) │ ├──────────────────────────────────────────────────────────────────┤ │ Taxation Strategy, Dynamic Planning & Bottom-Line Optimization │ ├──────────────────────────────────────────────────────────────────┤ │ Accounting & Continuous Financial Close │ └──────────────────────────────────────────────────────────────────┘ ```

Sibley shares her view that periodic budgeting cycles will soon give way to continuous, dynamic planning driven by real-time data feeds. In taxation, delivering cleansed data faster directly expands strategic tax planning opportunities, delivering immediate bottom-line impact. In accounting and close operations, intelligent automation targets the stress and sleepless nights historically associated with rigid month-end and year-end deadlines.

[14:30](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=870s) Deep Dive: The Financial Close Assistant and Agentic Workflow

A concrete walkthrough demonstrates how the SAP Financial Closing Assistant and agentic AI streamline the close process through a simulated persona, "Avery the close accountant":

``` AGENTIC FINANCIAL CLOSE WORKFLOW ┌──────────────────────────────────────────────────────────────────┐ │ 1. RECEIVABLES CLEARING AGENT │ │ • Analyzes & prepares clearing proposals │ │ • Avery reviews matching logic and executes batch posting │ ├──────────────────────────────────────────────────────────────────┤ │ 2. INTERCOMPANY RECONCILIATION AGENT │ │ • Compares AP/AR across all entity pairs │ │ • Flags discrepancies with detailed reasoning │ │ • Avery validates accuracy and approves adjustment entries │ ├──────────────────────────────────────────────────────────────────┤ │ 3. JOURNAL ENTRY ACCRUALS (via Joule Copilot) │ │ • Avery requests accrual proposals for Company Code 1710 │ │ • Joule generates compliant journal entry for one-click post │ ├──────────────────────────────────────────────────────────────────┤ │ 4. FINANCIAL STATEMENT ANALYSIS AGENT │ │ • Auto-evaluates Financial Statements Review Booklet │ │ • Synthesizes KPIs: Gross Margin, Fixed Expense Ratio, ROE, │ │ ROA, Current Ratio, and Depreciation across Company Code │ └──────────────────────────────────────────────────────────────────┘ ```

Operating from a centralized dashboard, the accountant acts as the final decision-maker. After the receivables and intercompany agents complete matching and reconciliation, Avery engages Joule via conversational prompts to generate accrual proposals for company code 1710. Once posted, Avery utilizes the Financial Statement Review Booklet alongside the financial analysis agent to rapidly interpret core performance metrics, shifting the accountant's role from manual data entry to strategic financial oversight.

[17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) Real-World Enterprise Proof Points, Governance, and Call to Action

Sibley provides concrete enterprise examples showing how AI delivers immediate operational value across diverse business functions:

- **Pfizer**: Deploys AI to continuously validate internal controls across every business entity on a daily basis, replacing periodic spot-checks with continuous governance. - **Reporting Optimization**: An enterprise eliminated 75% of scheduled reporting packages by empowering authorized business users to generate their own ad-hoc reports on demand using natural language. - **Treasury & FX Management**: Organizations set strict enterprise rules for foreign exchange handling and automated cash transfers, utilizing agents to enforce and execute policies autonomously.

Sibley underscores that enterprise adoption requires enterprise-grade auditability and governance. Every action taken by an AI agent—including the specific assistant involved, the exact data accessed, and the human who confirmed the transaction—is recorded within SAP's audit logs, ensuring strict compliance, transparency, and bias prevention.

To conclude, Sibley offers a pragmatic three-step call to action for finance leaders: 1. **Survey Senior Talent**: Convene experienced finance staff to identify their most frustrating, time-consuming, and low-value tasks. 2. **Select One Use Case**: Focus strictly on a single high-friction problem rather than initiating an expansive multi-year project. 3. **Engage in Targeted Workshops**: Collaborate with peers, partners, and SAP through hands-on AI workshops to map existing out-of-the-box system capabilities or design custom agentic solutions.

## Notable quotes > [00:04](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=4s) don't do the math — Mary Sibley

> [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) human in the loop — Mary Sibley

> [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) up to 80% of routine finance business processes can be fully or partially automated — Mary Sibley

> [06:41](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=401s) autonomous finance — Mary Sibley

> [08:54](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=534s) Joule is not merely a superficial chatbot, nor is the agentic layer an external wrapper bolted onto enterprise systems — Mary Sibley

> [08:54](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=534s) autonomous finance is not about reducing headcounts in the office of the CFO — Mary Sibley

> [12:12](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=732s) periodic budgeting cycles will soon give way to continuous, dynamic planning driven by real-time data feeds — Mary Sibley

> [14:30](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=870s) Avery the close accountant — Mary Sibley

> [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) Every action taken by an AI agent—including the specific assistant involved, the exact data accessed, and the human who confirmed the transaction—is recorded within SAP's audit logs — Mary Sibley

## Mentioned Mary Sibley, Ai4, SAP, Joule, SAP Autonomous Finance, SAP Financial Closing Assistant, Financial Statements Review Booklet, Pfizer, PowerPoint

## Video description Mary Sibley - RVP, SAP Cloud ERP Customer Advisory, SAP

Finance organizations are confronting escalating pressures driven by regulatory complexity, economic volatility, talent constraints, and growing expectations for AI-enabled transformation. Traditional operating models are increasingly unable to absorb these demands at scale. This session showcases SAP Autonomous Finance as the new paradigm that embeds AI agents in all financial processes. This approach emphasizes a human-in-the loop model, where AI optimizes execution while preserving governance and human judgment. Attendees will examine why the cost of inaction now exceeds the risks of transformation and gain a practical framework for initiating an AI-enabled autonomous finance journey.

[Source](https://www.youtube.com/watch?v=_IVz8a6xsZQ)

中文简报

SAP嵌入AI智能体以自动化80%的日常财务流程部署嵌入式AI智能体自动化日常结账与合规任务，不仅能帮助CFO应对迫在眉睫的人才退休潮，还能推动财会人员向战略监督角色转型。 Mary Sibley提出了通过SAP自治财务（SAP Autonomous Finance）转型CFO组织的框架，以应对电子发票强制令、经济冲击以及即将来临的人口退休断崖等宏观阻力。该愿景的核心是SAP的Copilot产品Joule，它跨越应收账款清账、内部交易对账和结账报告等核心财务领域协调各专业AI智能体，同时保持资深会计人员进行人机协同验证。Sibley强调，高达80%的日常财务流程均可实现自动化，并敦促企业领导者跳过冗长庞大的多年期IT项目，转而聚焦于能够带来即时ROI的具体运营瓶颈。企业财务AI已从传统的机器学习和预测分析演进为生成式AI及协作智能体系统，并将重塑CFO组织架构。 (0:04) 全球电子发票等日益繁重的监管要求、宏观波动性以及即将到来的人口退休断崖，使传统的纯人工财务运营模式难以维系。 (2:23) 人机协同（Human-in-the-loop）框架确保资深专业人士对智能体输出进行审核验证，从而兼顾硬性运营ROI与人才保留、团队士气等软性回报。 (2:23) SAP Autonomous Finance利用Joule作为自然语言编排器，协调跨核心业务数据结构的专业领域智能体。 (6:41) 自治财务智能体深度嵌入于企业ERP逻辑、业务规则及控制流程中，而非仅作为浮于表面的聊天机器人插件。 (8:54)

Reading summary

Mary Sibley of SAP argues that mounting regulatory mandates, macroeconomic volatility, and demographic turnover are forcing CFO organizations to adopt embedded agentic AI to automate routine workflows while keeping human judgment at the center of financial governance. Macro Pressures and the Human-in-the-Loop Model (2:23) Finance departments face accelerating external pressures, including mandatory e-invoicing frameworks across more than 60 countries, unpredictable supply chain disruptions, and tariff shifts. Simultaneously, organizations confront an impending demographic cliff, with a substantial portion of experienced personnel retiring within the next five years while incoming junior staff expect purposeful work rather than manual data entry. Sibley contends that addressing this transition requires preserving institutional knowledge through a human-in-the-loop framework. While up to 80% of routine finance tasks can be partially or fully automated, experienced professionals are needed to evaluate AI outputs and mentor early-career accountants. Sibley also notes that corporate boards are demanding tangible accounting of AI investments, requiring finance leaders to track both hard cost savings and soft returns such as improved employee retention and morale. Embedded Agentic Architecture via SAP Joule (6:41) SAP frames its response to fragmented systems around autonomous finance, coordinated by Joule, an integrated natural language generative AI copilot. Rather than functioning as a surface-level chatbot or an external wrapper, Joule is embedded directly into core SAP enterprise software, operating with native visibility into business data structures, controls, and enterprise rules. According to Sibley, autonomous finance aims to redirect human capital toward strategic business advisory rather than cutting finance headcounts. Sibley cautions organizations against launching unwieldy, multi-year IT transformations, recommending instead that leaders identify specific operational drains—such as manual dashboard compilation—and solve targeted friction points using embedded conversational agents. Agentic Workflows Across the Financial Close (14:30) The practical mechanics of autonomous finance are illustrated through a simulated close accountant, Avery, utilizing the SAP Financial Closing Assistant. Specialized agents execute heavy data-matching operations independently: a receivables clearing agent prepares proposals, and an intercompany reconciliation agent flags discrepancies across entity pairs with contextual reasoning, leaving human accountants to validate and execute the adjustments. Accountants then prompt Joule to draft compliant journal entry accruals for specific company codes, allowing one-click batch postings. Once entries are recorded, a financial statement analysis agent evaluates review booklets and synthesizes performance indicators such as gross margin, return on equity, and liquidity ratios, shifting the accountant's core responsibility from mechanical compilation to executive analysis. Enterprise Deployments, Governance, and Incremental Adoption (17:07) Sibley cites real-world enterprise deployments to demonstrate operational impact across finance functions. Pfizer uses automated agents to perform continuous daily validations of internal controls across entities rather than relying on periodic manual audits. In reporting, another organization eliminated 75% of scheduled reporting packages by enabling staff to generate ad-hoc reports via natural language, while corporate treasuries deploy agents to enforce cash transfers and foreign exchange policies automatically. To address enterprise compliance and mitigate bias, every automated action—including the specific agent used, the accessed data, and the approving human reviewer—is recorded in SAP audit logs. Sibley advises finance leaders to adopt AI incrementally by surveying experienced staff to identify low-value tasks, selecting a single friction point for initial deployment, and evaluating capabilities through focused workshops. Sibley's model depends on organizations successfully shifting senior staff from transactional processing to supervisory oversight while maintaining data integrity across core systems. The discussion leaves open how organizations with fragmented legacy software outside unified environments like SAP will maintain comparable auditability and multi-agent coordination.

精炼阅读摘要

SAP副总裁Mary Sibley指出，企业财务正从传统的分析与生成式AI迈向由智能体驱动的“自主财务”（Autonomous Finance）阶段；通过将AI副驾驶Joule与多专业智能体深度嵌入核心ERP系统，企业能够在应对监管剧变与人才断层压力的同时，将多达80%的常规财务流程实现自动化或辅助处理，使财务团队转型为战略业务顾问。外部冲击与人才断层倒逼运营转型 (2:23) Sibley指出，当下的财务团队面临严峻的外部挑战，包括超过60个国家推行强制电子发票等合规复杂度的攀升，以及突发关税调整、地缘政治冲突和原材料短缺等宏观经济震荡。与此同时，财务部门正遭遇人力资源结构断层：大量资深专业人员预计将在未来五年内退休，而新入职的年轻从业者则不愿承担枯燥的重复性劳动。这种脱节使传统的运营模式难以维系。为应对这一转变，Sibley强调必须坚持“人在回路”（Human-in-the-Loop）机制。资深员工数十年的业务洞察力对于核验AI输出结果不可或缺，因此企业需要建立导师结对制度以培养下一代会计人才。面对董事会对AI实际投资回报（ROI）的质询，Sibley表示，除了可量化的财务节省外，员工士气提升、留任率增加等软性回报同样关键；在自主财务模式下，高达80%的常规财务流程可被全部或部分自动化，从而彻底清理长期积压的基础事务。嵌入核心系统的自主财务架构 (6:41) 针对系统断裂与数据孤岛导致的低效，SAP提出了“自主财务”构想。Sibley澄清了行业误区，明确指出自主财务并非简单的表面对话机器人或外挂式包装，更不是为了裁减CFO部门的员工编制，而是将AI原生嵌入SAP核心系统，使其具备理解业务主数据、风控逻辑与企业规则的能力。该架构以自然语言生成式AI副驾驶Joule为中枢，协调底层的专业智能体团队，覆盖收入管理、财资与外汇、治理风险合规（GRC）、税务策略以及连续关账等领域。 Sibley表示，自主财务能够使传统的定期预算编制转向基于实时数据流的连续动态规划；在税务管理中，更快速交付的清洗数据能直接扩大战略税务筹划空间，为企业底线利润带来积极影响。她建议企业切勿将AI视为耗时数年的庞大IT重构，而应精准识别资深员工耗费过多精力的具体业务瓶颈，通过解决单个用例快速获得早期成效并逐步扩展。智能体协同驱动的财务结账流 (14:30) 在财务关账场景中，Sibley通过模拟会计师“Avery”的日常操作展示了智能体工作流的落地路径。在这一流程中，应收清账智能体（Receivables Clearing Agent）首先分析账目并提出清账建议，由会计师审核逻辑并执行批量过账；跨公司对账智能体（Intercompany Reconciliation Agent）自动比对所有实体间的应收应付账款，标明差异并给出详细分析，等待人工校验与调账确认。随后，会计师通过自然语言向Joule发出指令，为特定公司代码（如Company Code 1710）生成合规的预提费用凭证并实现一键过账。在报表分析阶段，财务报表分析智能体自动评估“财务报表审查手册”（Financial Statements Review Booklet），综合提炼毛利率、固定费用比率、净资产收益率（ROE）、总资产报酬率（ROA）、流动比率及折旧等多维关键指标。整套流程中，会计师的角色从机械的数据录入者转变为依托控制台做最终审批与战略分析的决策者。企业落地实践、审计合规与实施路径 (17:07) Sibley列举了多项企业实际应用成效以证实智能体的价值：辉瑞（Pfizer）已部署AI每日对所有业务实体进行内部控制的持续验证，取代了传统的定期抽检；某大型企业授权业务人员使用自然语言按需生成临时报告，直接削减了75%的固定排期报告包；在资金与外汇管理方面，企业通过预设严格的业务规则，使智能体能够自动执行合规的外汇操作与资金划拨。针对企业关切的安全合规问题，Sibley强调SAP系统对所有AI智能体的活动实施全流程审计追踪。系统会完整记录参与任务的具体助理、调用的底层数据以及进行人工确认的操作人员，确保操作透明、防范偏差并满足监管审计要求。为此，她向财务主管提出三步走建议：首先调研资深业务骨干，找出最繁琐耗时的痛点任务；其次仅挑选一个高摩擦场景作为突破口，避免启动冗长复杂的整体工程；最后与同行、合作伙伴及SAP联合开展实战工作坊，借助开箱即用能力或定制方案完成交付。 Sibley的阐述展现了以Joule与智能体为核心的自主财务愿景，主张通过人机协同释放资深财务人员的战略潜能并化解代际人才断层。然而，虽然报告展示了辉瑞等企业的内控持续验证与报告自动化成果，但针对复杂的全球多法人实体在更广泛业务链条中完全依赖智能体协同的长期容错率与实施成本，演讲并未展开详尽的量化评估。

Detailed notes — zh

## Background 本内容由 Ai4 频道于 2026 年 8 月 18 日发布在 YouTube 平台上。视频记录了 SAP 云 ERP 客户咨询区域副总裁 Mary Sibley 的主题演讲，主要探讨了在监管趋严、经济波动与人才断层的多重压力下，SAP 如何通过自主财务（Autonomous Finance）及人机协同（Human-in-the-Loop）模式，将 AI 智能体深度嵌入财务全流程并驱动企业转型。 ## TL;DR 演讲首先剖析了财务部门面临的法规剧增、宏观不确定性及资深人才退休等严峻挑战，指出高达 80% 的常规财务工作具备自动化潜力。随后，Mary 辨析了自主财务的核心内涵，强调 AI 智能体与 Joule 助手并非简单外挂，而是深度融入业务底座且保持严格审计追踪。通过展示公司间对账、预提分录过账及财务 KPI 智能诊断等实操演示，以及辉瑞等企业的落地案例，她呼吁财务团队从最耗时、最痛苦的单一场景切入，以小步快跑的方式开启 AI 转型之旅。 ## Atmosphere & style 整场演讲充满务实、亲切且引人入胜的行业共鸣感。演讲者结合自身从基层会计师做起的真实经历，用幽默风趣的语言精准击中财务人熬夜结账与应对海量报表的痛点，迅速拉近与观众的距离。现场节奏紧凑干练，既有底层技术逻辑的清晰梳理，又有流畅直观的产品实操演示，最终在计时器精准归零时幽默收尾，展现出专业自信且富有号召力的现场感染力。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=4s) 财务技术演进已从早期的机器学习、预测分析与生成式 AI，迈入以智能体（Agents）和数字助手（Assistants）为核心的新阶段，未来五年 CFO 办公室的工作模式将被彻底重塑。 - [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) 企业正面临多国合规法规（如电子发票）频更与宏观地缘风险激增的双重冲击，倒逼财务团队提升敏捷响应与损失控制能力。 - [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) 财务领域面临严峻的人才代际断层，既需要应对资深员工五年内集中退休的挑战，也需借助“人在回路（Human in the Loop）”机制与导师带教沉淀专业经验。 - [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) 财务业务流程中高达 80% 的工作具备通过 AI 实现全自动或部分自动化的潜力，能够极大释放核心人员被待办清单侵占的精力。 - [06:40](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=400s) 自主财务（Autonomous Finance）的智能体并非简单的“外挂层”，而是必须深度嵌入业务底座，完全理解底层数据结构与企业内置控制规则。 - [06:40](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=400s) SAP 副驾驶产品 Joule 不仅是聊天机器人，更是调度多智能体协同运作的中枢，可通过自然语言交互自动搭建仪表板并生成多维分析报表。 - [06:40](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=400s) AI 赋能并非为了削减财务编制，而是推动财务组织职责升级以强化战略影响力；企业应避免漫长的多年期大项目，通过聚焦痛点单点突破快速滚动。 - [12:12](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=732s) 智能体全面覆盖收入规划、司库管理、GRC、税务与关账等 CFO 核心业务，推动预算全面动态化并消除熬夜结账的痛点。 - [15:20](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=920s) 在端到端结账流程中，智能体自动执行应收核销与公司间对账并给出推导建议，人类专员仅需把关异常与最终审批过账，实现高效人机协同。 - [15:20](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=920s) 财务分析智能体可自动拆解固定支出比率、毛利率、ROE、ROA 及流动比率等核心 KPI，秒级绘制企业财务健康全景画像。 - [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) 标杆企业实践表明，AI 可实现全组织每日内控巡检（如辉瑞案例），并能精简多达 75% 的固定报表制作，赋能业务人员权限内自助制表。 - [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) 智能体时代的 AI 治理强调全流程审计追踪（Audit Trails），精确记录智能体操作与人工确认人，确保合规透明、杜绝算法偏见。 - [19:09](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1149s) 财务团队开启 AI 之旅的三步行动倡议：排查资深骨干最耗时的低价值痛点、聚焦单一场景先行试点、协同生态伙伴开展 AI Workshop 评估实施路径。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=4s) 财务职能的历史沿革与 AI 的演进演讲者 Mary Sibley 首先确认后排观众能听清自己的声音，随后展开关于未来财务组织如何在 AI 驱动下发生重塑的分享。她结合自身的职业履历指出，自己早年从公共会计师做起，后来担任财务总监（Controller），对每逢月末、年末周末加班结账的煎熬深有体会。她于 1994 年加入 SAP，并调侃大家“别去算工作年份”，在此期间亲历了技术与业务的演变。 AI 并非新鲜事物，技术路线经历了从早期的机器学习、预测能力、生成式 AI，再到如今令人兴奋的智能体（Agents）和数字助手（Assistants）阶段。她坚信，五年后的首席财务官办公室（Office of the CFO）将发生天翻地覆的变化，因为技术能力的飞跃彻底改变了具体的工作性质。 ```mermaid flowchart LR accTitle: AI 发展阶段演进图 A["机器学习 (Machine Learning)"] --> B["预测分析 (Predictive Capabilities)"] B --> C["生成式 AI (Generative AI)"] C --> D["智能体与助手 (Agents & Assistants)"] ``` - **演讲者背景**：公共会计师出身，曾任企业 Controller，1994 年加入 SAP。 - **本次分享主线**：剖析外部冲击因素、阐释 SAP 对自主财务（Autonomous Finance）的构想，并提出三点行动呼吁（其中仅有“半点”直接与 SAP 软件相关）。 --- ### [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) 外部冲击与人才断层挑战企业面临的合规压力与日俱增，以电子发票（e-invoicing）为例，已有 60 个国家推出并频繁更新法规，带来庞大的合规复杂度。同时，关税突增、地缘战争、原材料短缺等无法预料的宏观事件频发，迫使财务团队必须迅速反应，将业务损失降至最低。在人才层面，劳动力市场呈现两极分化： 1. **老龄化与资深员工流失**：大量经验丰富的“老兵”将在未来五年内集中退休； 2. **新生代诉求转变**：年轻一代渴望产生实质性影响，拒绝重复枯燥的纯体力式记账工作。在 Agentic AI（智能体 AI）时代，“人在回路”（Human in the Loop）的价值愈发突出。资深员工依赖数十年的实操历史做出精准拍板；为了防止代际断层，企业内部必须建立“导师帮带机制”（Buddy Process），将经验快速传授给年轻员工。同时，董事会关注的焦点正在从探索技术转向严苛的投资回报率：“财务部，你们在 AI 上的投资到底带来了什么回报？那些消耗的 Token 换来了什么成果？”除了直接的硬性资金收益（Hard Dollars），员工士气改善、工作满意度提升、离职率降低等软性价值（Soft Dollars）同样不可忽视。 - **自动化潜力**：财务业务流程中多达 80% 的工作有望通过 AI 实现完全或部分自动化。 - **工作减负预期**：把长期卡在待办清单（To-do list）上的繁杂工作交给智能体，释放核心精力。 --- ### [06:40](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=400s) 视频展播与“自主财务”内涵辨析在播放的概念展示片中，系统揭示了财务团队长期面临的痛点：数字时刻变动、数据散落在各孤立系统中、跨系统拼凑报表导致结账心力交瘁；同时 ESG 可持续性披露与外部监管步步紧逼。SAP 呈现了基于副驾驶产品 Joule 的工作景象：无需在多系统中频繁切换，自然语言提问即可调度后台助手与 AI 智能体群，实现场景预测、风险洞察与合规监控。 Mary 重点辨析了外界对“自主财务（Autonomous Finance）”的认知误区： ```mermaid flowchart TD accTitle: 自主财务架构逻辑 User["财务人员 (自然语言交互)"] --> Joule["统一管理枢纽: Joule"] subgraph SAP_Core ["SAP 业务内核 (底层嵌入)"] Joule --> Orchestrator["智能体协同与治理"] Orchestrator --> Agent1["应收核销智能体"] Orchestrator --> Agent2["公司间对账智能体"] Orchestrator --> Agent3["税务/资金/预测模型"] end SAP_Core -.-> Context["理解企业底层数据结构 & 遵循内置控制规则"] ``` 1. **Joule 不只是聊天机器人（Chatbot）**：它是调度智能体与助手的中枢。Mary 分享了自身的实操体验——她只需用口语告诉 Joule：“我需要这个仪表板；我想设一个 baseline（底线）；我想一键上传数据并生成 PPT 或 PDF 汇报，同时按多个维度切片分析数据”，系统随即自动完成搭建。 2. **智能体绝非外挂层（Not a layer on top）**：它必须深度嵌入业务底座，完全理解业务数据结构与企业内部合规控制规则。 3. **并非削减财务编制**：重构并非裁撤 CFO 办公室，而是推动职责升级。CFO 应聚焦在 C-suite 中发挥战略影响力，带领组织抵御外部风险。 4. **拒绝数年期庞大工程**：AI 落地不应设计成漫长冗余的多年期项目，而应排查当前企业内部痛点（消耗资深骨干精力最多的低价值环节），单点突破后迅速滚动作业。 --- ### [12:12](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=732s) CFO 职能覆盖面与结账场景实操 SAP 的智能体与助手体系严格对齐 CFO 办公室的核心业务架构： | 业务板块 | 核心价值与未来变革趋势 | |:--- |:--- | | **收入管理与规划** (Revenue & Planning) | 规划与预算将彻底动态化，打破传统的按季度、按年度周期性预算模式。 | | **司库管理** (Treasury) | 深度覆盖营运资本管理（Working Capital）与外汇风险对冲（FX Management）。 | | **治理、风险与合规** (GRC) | 实时嵌合多国监管规则，保障审计痕迹合规。 | | **税务管理** (Taxation) | 税务部门的核心诉求是“数据越快到位，筹划空间越大”，每一元税款节约都将直接沉淀为净利润。 | | **会计与财务结账** (Financial Close) | 消除“熬夜关账创伤”，实现平稳自动化关账。 | 现场进一步通过视频演示了财务结账场景中 SAP Financial Closing Assistant 的实操细节：财务人员 Avery 坐在“智能体结账团队”的驾驶位，仪表板实时呈现： - **应收账款核销智能体（Receivables Clearing Agent）**：自动分析账项并准备好预提核销清单，Avery 确认无误后一键过账，节省数小时的手工核对时间。 - **公司间交易对账智能体（Intercompany Reconciliation Agent）**：紧接着自动对接下一个对账环节，推动自动化连贯结账。 ### [15:20](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=920s) 人机协同的端到端自动化：从内部对账到凭证过账与财务分析在现场演示的业务场景中，智能体系统已经跨所有法律实体对自动完成了公司间应收与应付账款（Intercompany Receivables and Payables）的自动比对，并智能标记出需要人工介入的异常项目。财务人员 Avery 下钻查看首个标记项，系统中的 AI 智能体不仅完成了数据匹配，还给出了条理清晰的匹配依据推导过程。经过快速核对确认智能体的判断完全无误后，Avery 随即批准了匹配对，并一键接受系统针对其余未平项所给出的调整提议。完成对账后，Avery 切回主工作台查看当前智能体的运行状态，此时活跃的智能体仅剩两个。Avery 随即调出日记账分录智能体（Journal Entry Agent）以处理权责发生制预提费用（Accruals）。他唤出 SAP 的生成式 AI 助手 Joule，针对公司代码“1710”发起对话并提供上下文数据，确认下一步操作指引。Joule 迅速生成了完备的日记账分录提议；Avery 审核确认无误后直接点击“过账”（Post）。随着基础机械工作被 AI 全数搞定，Avery 终于从繁琐的数据录入中解放出来，拥有了充足的时间投入真正创造价值的财务分析工作。随后，Avery 打开针对公司代码 1710 的财务报表审查手册（Financial Statements Review Booklet），要求 Joule 对关键业绩指标（KPI）展开深度诊断。财务分析智能体随即响应，迅速汇总并全面拆解了包括固定支出比率（Fixed Expense Ratio）、毛利率（Gross Profit Margin）、净资产收益率（ROE）、总资产收益率（ROA）、流动比率（Current Ratio）以及总折旧费用在内的各项核心指标，为企业整体财务健康状况绘制了一幅一目了然的全景画像。至此，主控制面板上的所有待办事项圆满收官。 ```mermaid flowchart TD accTitle: Avery 与 AI 智能体协同闭环流程 subgraph S1["内部往来对账"] A["智能体自动对账比对"] --> B["标记异常项与推导说明"] B --> C["Avery 审核确认并接受调整"] end subgraph S2["凭证预提与过账"] D["调用 Joule 交互对话"] --> E["生成公司代码 1710 记账提议"] E --> F["Avery 审核并执行过账"] end subgraph S3["深度财务健康分析"] G["开启财务报表审查手册"] --> H["财务分析智能体拆解 KPI"] H --> I["输出 ROE/ROA/毛利等健康全景"] end S1 --> S2 --> S3 ``` - **覆盖核心财务指标**：固定支出比率、毛利率、ROE（净资产收益率）、ROA（总资产收益率）、流动比率及累计折旧费用。 - **人机协同模式（Human-in-the-Loop）**：AI 承担数据比对、推理建议与草案编制，人类专家把控确认、审批与最终过账决策。 --- ### [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) 企业实战案例与 AI 治理：辉瑞、自主报表与审计追踪 AI 技术在企业财务中的落地并非空中楼阁，演讲者列举了三个代表性的典型应用场景。首先是制药巨头辉瑞（Pfizer）的实践：辉瑞利用 AI 每天对全组织范围内的内部控制（Internal Controls）进行无死角排查，而非依赖传统的周期性抽检，这为企业风控合规带来了前所未有的“安全感与确定性”。第二个案例则触及财务部门日常的一大痛点：一家企业成功精简了 75% 的固定报表制作。演讲者幽默地调侃道：“我们大家都干过这种事对吧？悄悄压下一份报表不发，看看究竟会不会有人打电话来催着要。”而借助 AI，只要业务人员拥有合规权限，即可自主生成所需报告，实现了数据报表能力的真正向下赋权。第三个场景则是外汇（FX）与现金调拨管理，企业预先设定外汇头寸和资金划转规则，交由智能体自动监控并严格执行。针对企业普遍担心的合规问题，演讲者着重强调了底层的控制机制。SAP 向来以严密、无可挑剔的审计追踪（Audit Trails）闻名，而在 Agentic AI（智能体 AI）时代，系统将对智能体的一举一动实施完全透明的审计追踪：究竟是哪一个数字化助手或智能体执行了操作？当时环节中是哪一位人类专员点击了“确认”？整个链条全量留痕、有据可查。与此同时，SAP 强调杜绝算法偏见，确保系统的负责任（Responsible AI）与可信赖运营。 - **辉瑞（Pfizer）内部控制自动化**：实现跨所有组织架构的每日自动化内控合规巡检，取代低频次的人工周期抽查。 - **报表降负 75%**：由人工周期性集中制表转变为员工在权限框架内由 AI 赋能的自助式报表生成。 - **智能体审计追踪**：精准记录智能体操作明细、关联的具体 AI 助理身份及对应的人工核准确认人（Human in the loop）。 --- ### [19:09](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1149s) 财务团队的行动建议：从痛点切入开启 AI 之旅面对这一波变革浪潮，演讲者为在场的财务管理者提出了清晰务实的三步走行动倡议（Call to Action）。第一步，把团队召集起来，特别是去问一问部门内最资深的业务骨干：“你们最讨厌日常做哪些工作？哪些事情是你们觉得毫无价值感却又最耗费时间的？”通过这种方式，梳理出一份高耗时、低价值的日常业务痛点清单。第二步，不要贪多，从中精选出“仅仅一个”（Just one）最痛的业务场景作为突破口。第三步，积极与行业同行、技术伙伴展开对话。当企业准备就绪时，SAP 非常乐意为企业量身定制一场 AI 研讨工坊（AI Workshop），共同梳理业务诉求，评估是将需求直接匹配映射到标准系统中现成的 AI 智能体功能，还是量身构建专属的定制化 AI 解决方案。在演讲计时器精准跳至 00:00 时，演讲者笑着说道：“我刚好卡在零分零秒结束，时间把握得刚刚好！”她向在场的所有观众表达了感谢，并热情邀请大家前往展台做进一步面对面的深入探讨与交流。 - **行动第一步**：内部调研骨干，摸排最耗时、最缺乏价值认同的重复性痛点工作清单。 - **行动第二步**：克制聚焦，锁定单一场景先行试点，避免大而化之。 - **行动第三步**：协同生态伙伴开展 AI Workshop，评估标准开箱即用方案与客制化方案的实施路径。 ## Notable quotes > [00:04](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=4s) 别去算工作年份 — Mary Sibley > [02:23](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=143s) 财务部，你们在 AI 上的投资到底带来了什么回报？那些消耗的 Token 换来了什么成果？ — Mary Sibley > [06:40](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=400s) 我需要这个仪表板；我想设一个 baseline（底线）；我想一键上传数据并生成 PPT 或 PDF 汇报，同时按多个维度切片分析数据 — Mary Sibley > [12:12](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=732s) 数据越快到位，筹划空间越大 — Mary Sibley > [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) 我们大家都干过这种事对吧？悄悄压下一份报表不发，看看究竟会不会有人打电话来催着要。 — Mary Sibley > [17:07](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1027s) 究竟是哪一个数字化助手或智能体执行了操作？当时环节中是哪一位人类专员点击了“确认”？整个链条全量留痕、有据可查。 — Mary Sibley > [19:09](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1149s) 你们最讨厌日常做哪些工作？哪些事情是你们觉得毫无价值感却又最耗费时间的？ — Mary Sibley > [19:09](https://www.youtube.com/watch?v=_IVz8a6xsZQ&t=1149s) 我刚好卡在零分零秒结束，时间把握得刚刚好！ — Mary Sibley ## Mentioned Mary Sibley, SAP, Ai4, Joule, Pfizer (辉瑞), Avery, SAP Autonomous Finance, SAP Cloud ERP, SAP Financial Closing Assistant, Receivables Clearing Agent, Intercompany Reconciliation Agent, Journal Entry Agent, Financial Statements Review Booklet ## Video description Mary Sibley - RVP, SAP Cloud ERP Customer Advisory, SAP Finance organizations are confronting escalating pressures driven by regulatory complexity, economic volatility, talent constraints, and growing expectations for AI-enabled transformation. Traditional operating models are increasingly unable to absorb these demands at scale. This session showcases SAP Autonomous Finance as the new paradigm that embeds AI agents in all financial processes. This approach emphasizes a human-in-the loop model, where AI optimizes execution while preserving governance and human judgment. Attendees will examine why the cost of inaction now exceeds the risks of transformation and gain a practical framework for initiating an AI-enabled autonomous finance journey. [Source](https://www.youtube.com/watch?v=_IVz8a6xsZQ)

Source quotes

[0:04] Mary Sibley: 别去算工作年份 [2:23] Mary Sibley: 财务部，你们在 AI 上的投资到底带来了什么回报？那些消耗的 Token 换来了什么成果？ [6:40] Mary Sibley: 我需要这个仪表板；我想设一个 baseline（底线）；我想一键上传数据并生成 PPT 或 PDF 汇报，同时按多个维度切片分析数据 [12:12] Mary Sibley: 数据越快到位，筹划空间越大 [17:07] Mary Sibley: 我们大家都干过这种事对吧？悄悄压下一份报表不发，看看究竟会不会有人打电话来催着要。 [17:07] Mary Sibley: 究竟是哪一个数字化助手或智能体执行了操作？当时环节中是哪一位人类专员点击了“确认”？整个链条全量留痕、有据可查。 [19:09] Mary Sibley: 你们最讨厌日常做哪些工作？哪些事情是你们觉得毫无价值感却又最耗费时间的？ [19:09] Mary Sibley: 我刚好卡在零分零秒结束，时间把握得刚刚好！

Source quotes

[0:04] Mary Sibley: don't do the math [2:23] Mary Sibley: human in the loop [2:23] Mary Sibley: up to 80% of routine finance business processes can be fully or partially automated [6:41] Mary Sibley: autonomous finance [8:54] Mary Sibley: Joule is not merely a superficial chatbot, nor is the agentic layer an external wrapper bolted onto enterprise systems [8:54] Mary Sibley: autonomous finance is not about reducing headcounts in the office of the CFO [12:12] Mary Sibley: periodic budgeting cycles will soon give way to continuous, dynamic planning driven by real-time data feeds [14:30] Mary Sibley: Avery the close accountant [17:07] Mary Sibley: Every action taken by an AI agent—including the specific assistant involved, the exact data accessed, and the human who confirmed the transaction—is recorded within SAP's audit logs

Original transcript

[0:04] Hello everyone. Can someone in the back raise their hand they can hear me clearly? Beautiful. [0:09] All right. Well, I am so happy to have a chance to chat with you today. And I want to [0:17] give a vision of the finance organization of the future. And how that is impacted by AI. [0:26] And these are my views based on my personal work history. I started as a public accountant and then [0:34] became a controller. I remember those days of long weekends get closing the books and year-end. So, I [0:43] have that frame of reference. And I was fortunate to join SAP back in 1994. Don't do the math. [0:55] And when I say fortunate, I didn't realize I was going to fall in love with an organization, [1:00] but I have. And I've also had the opportunity [1:05] to see the solution embrace AI as it has evolved over time. Cuz AI is not new. [1:18] Machine learning has been around for a long time. Predictive capabilities. Generative [1:25] AI. And now we're at the exciting moment of agents and assistants. And what is that going to mean? [1:34] All right. If I learn to click once, not twice. Well, what I believe is if we go into the future [1:44] 5 years from now and we look at the office of the CFO it's going to look different. Because the work [1:54] will be different as part of it because of the capabilities of AI as it continues to evolve. [2:02] Okay. There are Today I want to talk about external factors. [2:08] I want to share with you our particular view. And then I have a call to action. [2:15] There's three parts of this call to action and only .5 relate to SAP. [2:23] So, well, let's take a look at external factors in here. Compliance is getting harder and harder. [2:36] And I'm getting drier and drier. [laughter] [2:40] But look how complex e-invoicing has become. 60 countries already. We're the constant changes. [2:50] There's that level of complexity. What else is happening? The macroeconomic events [2:58] that we don't see coming that impact us. And what does that cause us to to do as organizations? We [3:06] have to react. We have to react quickly so we minimize the damage of those external [3:14] things such as surprise tariffs or war, right? Or shortages of raw materials. [3:23] And we think about talent. This is one area that I think about a lot. If you look at labor statistics [3:34] and you look at your seasoned workforce, is that the appropriate word? Right? And how many [3:42] of those people will leave the workforce in the next 5 years? It's a huge percent. [3:48] Right? So, that's one factor. The other factor is this new generation that has come in is really [3:57] fascinating and they want to make a difference. Right? They want to make a difference. They don't [4:03] necessarily want to do tasks that they view as boring. Right? They want to make a difference. [4:12] Now, there's another piece that I've been thinking a lot about. As we move into a gentic AI, [4:20] the human in the loop becomes more and more important. Right? These are decisions that [4:27] have to be made. And when I make decisions, when I look at the output, I am calling on all these [4:36] years of history that I have in order to make the decision. So, I my thought process is as we bring [4:46] young people into our organizations, we're going to have to give them the skill. So, [4:53] I think there's going to be a buddy process that has to work. And that's one of the changes I think [4:59] organizationally that'll happen. And then, the board. Right? The board has been talking about AI [5:09] for a while, but I believe the conversation will change. It'll be, "Dear finance department, [5:19] what is the return that you got for your investment in AI?" "How did you spend those [5:26] token dollars in there and what do you have to show?" And I think that I don't have all [5:31] the answers there, but I think some of it is soft and some of it is hard dollars. Because if we can [5:38] improve our employees' morale, if we can give job satisfaction, if we can reduce turnover. [5:48] That may not be as quantifiable as some hard dollar savings, but I think it's super [5:54] super important. When we take a look at finance business processes, [6:03] we believe that 80% of the business processes can be automated with AI either fully or partially. [6:17] And the reason that I get excited about that is I don't know if all of you have to do lists, [6:25] but I always have things that get stuck on my to-do list. Wouldn't it be nice if my agent [6:32] could take care of some of the things on my to-do list so I could get through and really finish. [6:40] All right. So, those are factors within organizations, externally, [6:46] the evolution of AI itself. But what I'd like to do is share with you our view, [6:55] and we call it autonomous finance. It doesn't mean there's no humans in the loop, but the use [7:00] of agents and assistants within the finance process. So, let me run a very short video. [7:08] Economic uncertainty is keeping you up at night. It's hard to rest easy when the numbers keep [7:13] changing and everyone's looking to you to make sense of it all. But the information you need is [7:18] scattered across [music] systems, leaving your teams tied up manually piecing data together, [7:23] making it harder to plan ahead and close the books with confidence. [7:27] At the same time, compliance and regulatory pressures are building with sustainability [7:32] requirements bringing more scrutiny than ever. In critical moments like these, [7:37] you need to find answers fast. And now, with Jewel, all you have to do is ask. [7:43] No more switching between systems or chasing answers. Jewel instantly understands what you're [7:48] trying to achieve and builds the right context around it. [music] Behind the scenes, finance [7:53] assistants coordinate a team of AI agents to predict and model scenarios, monitor performance, [7:59] analyze risks and opportunities, and keep operations aligned with governance and compliance. [8:04] So, you can make the right decisions based on all the relevant data. Applications, contextualized [8:10] data, and industry-specific AI come together in a complete financial management suite. [music] Work [8:16] becomes more autonomous, moving forward faster, helping you focus on what matters, [8:21] leading your business through uncertainty towards [music] sustainable growth. You can [8:26] forecast and plan with confidence from a single source of truth. When your teams need alignment, [8:31] everyone's working from the same view, so you can spend less time chasing numbers and focus [8:36] on driving [music] the outcomes your business depends on. And when the day is done, you can [8:41] finally switch off. This is financial management for the AI era, autonomous, intelligent, [8:47] and ready for what's next. This is autonomous finance. This is the beginning of better. [8:54] So, one of the things that I think is equally important is what autonomous finance is not. [9:03] So, first of all, it's not a chatbot. Jewel is our it really manages the agents and the assistants, [9:16] and it interacts with me as a user in here. So, I can, in natural language, [9:24] tell it what to do, and I find myself talking to Jewel like a person. Hey, Jewel, [9:30] I need this dashboard. And I want to put a line in the sand. I'm famous for my line in the sand [9:38] and there I want to put a line in the sand. I want to be able to upload data when I want to at push [9:45] of a button. I want to be able to create a PowerPoint or PDF to show the results and [9:51] by the way, I want to take this data and I want to slice and dice it by many dimensions in here. [9:59] And that is how I created a dashboard with Jewel with with just using my vocabulary and I [10:06] think that is super exciting for all of us to be able to do and that's available today. Okay. So [10:14] Jewel is not a chatbot. It autonomous finance the agent is not a layer that sits on top. Right? It's [10:21] embedded within the solution itself understanding the data structure and understanding your rules. [10:30] Okay. I feel very strongly that it's not about reduction of the office of the CFO. [10:39] I think it's about shifting the responsibilities. You know, if we look at the role of the CFO [10:50] the view is to be able to be a key member of the C-suite to be able to influence and to be able [11:01] to help guide the organization through whatever it's facing it at the time. So it's you know [11:09] we could argue whether it's reduction or not, but I don't actually think it is. [11:15] I really think it's about getting to that to-do list all of those items on there and [11:23] leveraging the key responsibilities and the knowledge of your most senior folks. [11:31] And I believe AI should not be a multi-year program for any organization. I would argue [11:42] that you should sit down, take a look at your use cases. Where is there a drain [11:49] in your organization? Everyone has a different spot there. Where are your most senior people [11:57] spending time they wish they weren't? And that is the place to start. Be successful, [12:03] get that rolling, and then move on to the next area in there. To be very symptomatic. Okay. [12:12] So, in a nutshell, when we talk about autonomous finance, there's some There's two really key words [12:21] in there. It's your data, your business data, understanding the structure of your business data, [12:28] and working inside the controls that you have established for your organization. [12:37] The way we have structured our agents and assistants are the way the office of the CFO [12:45] typically has a structure in there. So, there is revenue management component in there, [12:53] planning. I So, this is Mary's personal opinion. I believe planning and budgeting over time will [13:00] become dynamic. I I think the days of quarterly or annual or periodic budgeting will really change. [13:10] Treasury, working capital management, foreign exchange management. [13:15] Government, risk, and compliance. Taxation. Let's take this up. [13:25] I think this is a really interesting area because if anytime I talk to a tax department [13:32] they always say if you can get the data to me faster I can plan better and every dollar of tax [13:43] right goes right to the bottom line. It's a a huge benefit in there. So I'm very excited [13:47] about the taxation portion of it. And then accounting close, I still have some trauma [13:53] from accounting closes that seem to me yes. I I see heads nodding in there. But there was some [14:00] there was some sleepless nights. That's why I like that little video because I that's what I [14:04] want to do at night. I want all CFOs to be able to go to sleep and not not worry about things. [14:10] All right, so that's the structure of here. I wanted to to double click on just one [14:17] and that's the accounting and financial close just to give you a sense of what it's like. [14:30] For many, the world of the financial close is tight deadlines, manual data crunching, and [14:36] high pressure. Today, that world is changed. The SAP Financial Closing Assistant now supercharged [14:44] with Agentic AI puts Avery our close accounting in the pilot seat of an Agentic team close. [14:50] Instantly, the dashboard comes alive and the receivables clearing agent is ready for review. [14:57] Avery reviews the list of proposed clearings that the agent has analyzed and prepared. [15:02] And he is satisfied with the proposed clearings and posts them saving hours of manual effort. [15:08] Impressed, Avery realizes the benefits of creating an Agentic team close. [15:14] And now is eager for the next agent review, intercompany reconciliation. The agent has [15:20] automatically compared intercompany receivables and payables across all entity pairs, flagging [15:26] items that require attention. Avery drills into the first flagged item. The agent has provided a [15:33] detailed reasoning for the matched pairs. A quick review confirms the agent's accuracy. [15:40] Avery approves the matched pairs and accepts the proposed adjustments for the remaining items. [15:46] Avery transitions back to the main dashboard to view the current agentic statuses. [15:51] Now only two agents are active. Avery taps the journal entry agent to investigate the accruals. [15:59] Avery starts a conversation with Jewel to assist him with the journal entry proposals [16:03] for company code 1710. Avery provides relevant data to Jewel and confirms Jewel's next step. [16:12] Jewel then successfully creates the journal entry proposal, which is ready for Avery to review. [16:18] Avery likes the proposal and clicks post. [16:22] Avery checks his dashboard, and since the agents have completed all of the work, he has time for [16:27] financial analysis. Avery opens the financial statements review booklet on company code 1710. [16:35] He asks Jewel to analyze the KPIs. Jewel responds and the financial analysis agent acts, [16:43] helping to summarize, analyze KPIs like the fixed expense ratio, gross profit margin, [16:48] return on equity, return on assets, current ratio, and total depreciation expenses, [16:55] giving you a good overall picture of the financial health of your business. [16:59] And just like that, we are back to the main dashboard with our work complete. [17:07] So, I wanted to give three quick examples because I've said AI's been around for for quite a while. [17:14] So, Pfizer uses AI to to check their internal controls across every organization every day [17:26] instead of periodically checking that. So, the a comfort area in there. The next example, [17:35] one organization was able to eliminate 75% of the reports that they were producing [17:43] uh and we all have done this, right? We hold back a report to see if anyone [17:48] actually calls to see if they if they need it. But, in this case because the community [17:56] as long as they have a the right authority can create create the report themselves. [18:02] Right? It's an empowerment down. And then the last example is around foreign exchange, [18:09] setting the rules around foreign exchange, saying the rules around cash transfers, [18:15] and then having the agent enforce the rules. I wanted to mention a two points on this slide. [18:24] We are famous for audit trails. Every transaction, every piece is tracked. So, with a Gentic AI [18:36] the agent activity is also tracked. What assistant, what agent did it? Who was the [18:42] human in the loop that said confirm in there? So, you have a full audit trail going through. [18:49] And we also feel very strongly about making sure the bias does not exist and making sure that the [19:00] the uh responsibility is maintained in there. Okay. My call to action, again, three things. This [19:11] is what I would propose. First of all, bring your staff together. And ask your most senior people [19:20] what they hate doing because they don't see value in it. What is the most time-consuming activities [19:27] in there? You so you'll have a list of activities that go through. Number two, pick one, just one [19:34] in there. Number three, talk to talk to peers, talk to partners. And then when you're ready, [19:44] we would love to talk to you. And what we would propose is an AI workshop where we would take a [19:51] look at what you want to accomplish, either map it to what exists in the system or talk about a [19:58] custom AI solution in there. And uh I finished right at 00. I just went Well well timed. So uh [20:09] uh my last to ask is first of all, thank you for spending time with me today, [20:13] and come by our booth. We'd love to chat with you. I hope you have a great conference.