王峰 AI × 经营 约一次诊断

总目 / 文章 / WF-0363

文章 · 存档

共享CFO深度解读财政部《会计基础工作规范》:从制度到系统的全链条升级

财政部9月3日发布了关于征求《会计基础工作规范(征求意见稿)》和《填制会计凭证、登记会计账簿操作指南(征求意见稿)》意见的函 财办会〔2025〕34号。

这是根据2024年7月发布的征求意见情况修改完善后,再次形成《会计基础工作规范(征求意见稿)》和《填制会计凭证、登记会计账簿操作指南(征求意见稿)》。

过去三十年,会计基础工作的底层逻辑经历了三次跃迁:从手工到电算化,从电算化到信息化,如今正迈向“智能化+治理”的第三阶段。

财政部此次对《会计基础工作规范》及配套操作指南的系统修订,实质上是在统一底层规则、补齐信息化治理与内控短板、为AI时代的数据可信与审计追溯打地基。

这不仅是一轮“修文改章”,更是一次以制度牵引流程、以流程规范系统、以系统反哺治理的结构化重构。

顶层定位:把“会计基础”从“事务环节”提升为“治理要素”

新《规范(征求意见稿)》旗帜鲜明地把单位负责人的第一责任压实到“会计基础工作与会计资料的真实性、完整性”上(不再只是会计条线的内部事),同时明确会计机构、会计人员依法履行核算与监督职责,并要求建立健全内部控制、对会计软件与会计资料的信息化合规提出明确要求。

换句话说,从“把账做对”升级为“用对机制+系统把账做对”。

这对企业治理的三个直接影响是:

1)CEO/CFO需对数据闭环负责(含凭证、账簿、报表、归档);

2)内控与IT成为会计基础工作的必要前置条件;

3)系统选择与参数治理将纳入可问责范围。

组织与岗位:信息化情境下的“三道闸”与“一把钥匙不能开所有门”

在岗位设置上,征求意见稿将岗位划分与不相容分离原则前置,并新增“会计软件管理岗位”,明确其只负责用户维护/授权等系统管理工作,不得在所管理软件上处理具体业务;

同时强调出纳不得兼任稽核、档案保管、债权债务登记、会计软件管理,严禁一人办理货币资金全过程业务,严禁一人保管全部支付印章/密钥。

这些条款共同构成了信息化时代的“三道闸”:岗位分离—权限分离—介质分离。

即便“一人多岗”,也必须切断“审批-经办-记账-保管”的串联路径。

对中小企业尤为关键,因为“人少岗多”最容易形成单点失守。

更进一步,文件提出计划性轮岗;不具备轮岗条件的单位应以专项审计等方式强化关键岗位风控——这是把“轮岗”从理想做法升级为必须有替代方案的控制要求。

对国有及控股/主导地位单位,还明确了任用回避(负责人、总会计师的亲属不得担任会计主管或出纳/软件管理),这将显著降低“家族式控制”下的舞弊风险。

交接与可追溯:把权限、介质、数据都纳入“移交清册”

与以往相比,本次对“交接”的信息化要求极具操作性:

- 交接需编制移交清册,覆盖凭证、账簿、报表、印章、库存现金、有价证券、支票簿、票据、文件、电子数字证书、密钥、会计软件数据载体等;

- 会计软件中的数据交接必须在实际操作状态下进行,新增接管人员账号并授权;同时关闭移交人员权限,严禁“接着用原账号”。

这套做法相当于把“人—账—物—权—证—数”统一纳入交接责任链,既满足审计追溯,也为AI/自动化流程运行提供干净的权限边界。

电子凭证与自动化:承认“机器生成”,强制“规则可追溯”

配套的《操作指南》是将“制度语言”翻译为“操作动作”的关键文件。它对电子原始凭证与--机制记账凭证(系统自动生成)--给出可落地的要求:

- 符合标准、来源可靠、程序规范、要素合规的电子原始凭证可直接入账与归档;

- 对系统自动生成的机制记账凭证,要求嵌入审核规则且规则可查询、可校验、可追溯,设立与变更须履行审签程序并留档;

- 系统要保证电子凭证接收、生成、传输、存储各环节的安全可靠,并能发现篡改、阻止重复入账。

这意味着,企业可以合理拥抱自动化/AI,但必须把“自动化规则”本身纳入被审计对象,并接受变更控制。AI工具不是“黑箱”,而是“可解释的会计工序”。

账簿与对账:把“第三方平台资金”纳入账实核对

在账簿与对账方面,征求意见稿将传统的“现金、银行”扩展为“第三方支付平台”,要求定期核对平台账户资金与明细账余额;

并重申总账-明细账-日记账、账证、账表的“四相符”,同时规范手工记账的编号、划线、红字冲销、结转等细部动作。

对全面线上化的企业而言,“多钱包/多平台”资金正式进入硬性对账范围,这对电商、直播、SaaS订阅与跨境业务尤为重要。

会计监督:把“拒绝办理权”与“内部制度执行监督”写到明处

文件明确会计机构、会计人员的监督职责与拒绝办理权:

- 对不真实、不合法原始凭证,有权不予接受并报告;

- 对审批手续不全、未纳入统一核算、违反制度的财务收支,有权退回/拒绝/纠正;

- 对内部会计管理制度的违反行为,可制止、纠正并报告;

- 对账实不符的情形,要求查明原因并作出处理。

这将“监督”从“道德倡议”变成“制度授权”,为会计人依法履职提供了明确边界与保护。对于AI批量生成凭证或自动处理收付的场景,也意味着必须保留人类复核与干预机制。

修订动因:从“法治短板”到“数智融合”的系统回应

《起草说明》交代了修订的四个核心动因:

1)承接会计法修改与“进一步加强财会监督”的顶层要求;

2)支撑“有效市场+有为政府”的数据基础;

3)响应信息技术广泛应用带来的智能会计、财务共享、机器人自动化等新形态;

4)通过“分离附件、另立操作指南”的方式增强灵活调整能力,便于以后随实务更新。通俗地说:把不变的写入规范,把会变的写进指南,为未来与AI、区块链、物联网等技术的持续融合留足制度接口。

与AI/智能财务趋势的“自动对齐”

1. 电子凭证直入与自动生成凭证的审计化

AI和RPA生成的凭证、分录、报表,不再是“试验品”,而是在“规则受控+日志可追溯”的条件下被制度承认(前提是来源与要素合规)。这为凭证自动化、分录引擎、月结自动化的规模化落地扫清了合规障碍。

2. 权限治理成为AI落地的前提件

新增的“会计软件管理岗位”与“交接时关停原账号/增开新账号”的规定,使权限治理成为AI/系统项目成功的第一步——没有清晰的人-岗-权映射,智能化只会放大风险。

3. 第三方平台纳入对账闭环

对电商/平台企业而言,AI自动对账引擎(对接银行、支付、ERP、DTC后台)将成为“账实相符”的关键工具——制度已把“该核对的对象”明确列为必核。

4. “人机共治”的监督模型

制度强调会计监督与拒绝办理权,这意味着AI/自动化不可替代人类的合规判断与职业怀疑。技术层面,应将“AI输出—人类复核—日志留痕”设计为标准流程。

共享CFO给你的企业落地“4×3路线图”:四条主线 × 三个月达标

为帮助企业快速对齐,这里给出一个可执行的“4×3”路线图(4条主线,3个月达标):

主线A:组织与岗位(第1月完成制度稿,第2月执行)

- 梳理会计岗位矩阵与不相容分离清单;明确会计软件管理岗位职责;

- 建立轮岗/替代控制与专项审计计划;

- 对国有/控股单位落实任用回避与亲属关系排查。

主线B:流程与凭证(第1-2月制度+配置,第3月试运行)

- 制定电子原始凭证管理细则:来源、标准、查验链、重复入账防控;

- 为机制记账凭证配置审核规则与变更审批流程,所有规则可查询、可追溯;

- 建立错账红字冲销/更正与分割单应用规范,统一摘要与编号规则。

主线C:权限与交接(第1月梳理,第2月切换,第3月固化)

- 全量清查人-岗-权,关闭共享账号;

- 制定交接清册模板,把印章、密钥、证书、数据载体、账号权限纳入交接;

- 建立账号接管—原账号关停的标准作业程序(SOP)。

主线D:对账与监督(第1月立项,第2月上线工具,第3月评估)

- 把第三方平台资金纳入月度/季度对账;

- 上线自动对账/异常预警与审计追踪日志;

- 实施“拒绝办理/退回补正”台账与管理例会通报机制。

共享CFO给管理层的三点提醒

1)把会计基础当作“数据治理工程”来管:把制度条款落实为可配置的系统参数、可审计的日志、可核验的报表。

2)从“项目思维”转向“运行思维”:机制记账凭证、电子凭证与权限治理不是一次性上线,而是持续复核与变更控制的运营事项。

3)AI是加速器,不是救火队:AI/RPA让效率指数级提升,但前提是规则清楚、数据干净、权限清晰。制度已把“前置条件”列给你了,做不到,别急着上AI。

这轮修订的价值,在于用“可操作的制度语言”把智能化时代会计基础工作的边界、路径与抓手讲清楚了:岗位-流程-系统-监督四位一体,既守住底线,也为AI与自动化留下了高速通道。企业真正要做的,是把这些条款翻译成系统参数与团队行为。当制度与系统“互相支撑”时,AI才可能成为会计与经营管理的可信引擎。

共享CFO能为你提供的服务

免费诊断沟通(0元,20分钟)

预约一次20分钟的共享CFO深度沟通,快速识别企业在财务、业财、财税、资金、预算、战略落地、管理分析等环节出现薄弱状态的深层原因,帮你厘清问题的根源与优先级。

轻量落地诊断(低价体验,协助落地)

在一次小范围、低成本的专项诊断中,我会为企业绘制专属的问题剖析与落地路线图,明确未来三个月能看得见成果的具体动作。

共享CFO陪跑(短期见效 → 长期共创)

当企业通过轻量方案初见成效后,可进入共享CFO陪跑服务——以阶段性目标为导向,用最少成本完成从制度、流程到AI治理的一体化落地,持续提升管理层的经营掌控力。

Deep Analysis of “Accounting Basic Work Standards”

Shared CFO Deep Analysis of the Ministry of Finance's “Accounting Basic Work Standards”: From Rules to Systems, a Full-Chain Upgrade

On September 3, the Ministry of Finance issued a letter (Cai Ban Hui [2025] No. 34) to solicit opinions on the “Accounting Basic Work Standards (Draft for Comments)” and the “Guidelines for Preparing Accounting Vouchers and Registering Accounting Books (Draft for Comments).”

This new draft was revised based on the feedback collected since July 2024 and represents another round of refinement.

Over the past thirty years, the foundation of accounting work has gone through three major transformations: from manual bookkeeping to computerized accounting, from computerization to informatization, and now, toward the third stage — “intelligent + governance.”

This latest revision of the Accounting Basic Work Standards and its supporting guides essentially aims to unify underlying rules, fix gaps in information governance and internal control, and lay the groundwork for data reliability and audit traceability in the AI era.

It's not just a “document update,” but a structural reconstruction — using systems to enforce governance, processes to standardize operations, and institutions to guide systems.

Top-Level Redefinition: Turning “Accounting Basics” from Transactional Work into a Governance Element

The new draft clearly places the first responsibility on the organization's head for ensuring the truthfulness and completeness of accounting work and records — no longer just an internal accounting matter.

It also defines that accounting departments and staff must perform both bookkeeping and supervision duties according to law, strengthen internal controls, and ensure compliance for accounting software and information systems.

In short, it's an upgrade from “doing the books right” to “using the right mechanisms and systems to keep the books right.”

The direct effects on corporate governance are:

CEOs/CFOs must take full responsibility for the data loop — vouchers, ledgers, reports, and archives.

Internal control and IT now become preconditions for solid accounting work.

System selection and parameter management will be included in accountability.

Organization and Roles: Three Layers of Control and “One Key Can't Open All Doors”

In staffing design, the draft emphasizes segregation of duties and incompatibility rules, and introduces a new “accounting software management role.”

This role only manages system access and user authorization, and may not process transactions in the same system.

It also bans cashiers from holding any position related to auditing, recordkeeping, debt/credit registration, or software management. No one is allowed to handle the entire cash process or keep all payment seals or encryption keys.

These clauses form the three gates of information-age accounting:

Role separation → Permission separation → Medium separation.

Even when “one person holds multiple roles,” approval, transaction, recording, and custody must be cut apart.

This is critical for small and medium enterprises (SMEs), where “few people, many jobs” often leads to single-point failure.

The draft also requires planned job rotation, and if rotation isn't possible, firms must apply special audits as a substitute risk control.

For state-owned or controlling entities, conflict-of-interest rules are reinforced — the head or chief accountant's relatives may not serve as accounting supervisors, cashiers, or software administrators.

This significantly reduces fraud under family-style control structures.

Handover and Traceability: Including Permissions, Media, and Data in the Transfer List

Unlike previous versions, this draft's digital handover rules are highly practical:

Transfers must include a handover list covering vouchers, ledgers, reports, seals, cash, securities, checks, bills, files, digital certificates, keys, and accounting software data carriers.

Data handover must occur within the system — create new user accounts for successors, revoke the former user's access, and strictly ban “using old accounts.”

This method unites people, books, assets, permissions, certificates, and data into one chain of accountability — satisfying both audit traceability and AI automation security boundaries.

Electronic Vouchers and Automation: Accepting “Machine-Generated,” Requiring “Traceable Rules”

The Guidelines for Preparing Vouchers translate system language into operational actions.

They define clear, practical requirements for both electronic original vouchers and system-generated vouchers:

Electronic vouchers that meet standards, have reliable sources, proper procedures, and compliant elements can go directly into accounting and archiving.

System-generated vouchers must embed audit rules that are visible, verifiable, and traceable. Rule creation or change requires approval and documentation.

Systems must ensure secure receipt, generation, transmission, and storage of vouchers and prevent tampering or duplication.

In essence, automation and AI are welcome — but their rules themselves become auditable objects.

AI tools are not “black boxes” but must act as “explainable accounting procedures.”

Ledgers and Reconciliation: Including Third-Party Platform Funds

The draft expands reconciliation from “cash and bank” to “third-party payment platforms.”

Companies must regularly reconcile platform balances with ledger accounts, and ensure consistency between general, sub, and journal ledgers, vouchers, and reports.

It also standardizes manual bookkeeping details like numbering, red corrections, and carry-forward entries.

For fully digital firms — e-commerce, live streaming, SaaS, or cross-border — this means “multi-wallet/multi-platform” funds now fall under mandatory reconciliation, a major compliance shift.

Accounting Supervision: Clear “Right to Refuse” and “Internal Oversight”

The document formalizes the supervisory power and refusal rights of accountants:

They may reject false or illegal source documents and report them.

They can return, refuse, or correct transactions with incomplete approvals or rule violations.

They can stop and report breaches of internal accounting rules.

They must investigate and resolve mismatches between records and actuals.

This shifts “supervision” from moral duty to legal authority, giving accountants clear protection to perform their duties — and ensuring AI-generated vouchers or automated payments remain subject to human review and intervention.

Why This Revision: From Legal Gaps to Intelligent Governance

The drafting notes explain four main drivers:

Aligning with the revised Accounting Law and national goals for stronger financial oversight.

Providing reliable data foundations for both market and government governance.

Adapting to new realities — smart accounting, financial sharing, and robotic process automation (RPA).

Separating the fixed rules (in the standards) from adjustable ones (in the guide), leaving room for future tech like AI, blockchain, and IoT.

In short: Keep the stable parts in the rules, the evolving parts in the guide.

Natural Alignment with AI and Smart Finance

Auditable AI-Generated Vouchers

AI and RPA-produced vouchers, entries, and reports are now formally accepted, as long as they're rule-controlled and traceable — clearing compliance barriers for automation.

Permission Governance as a Prerequisite

The new accounting software role and account-transfer rules make access control the first step of AI success. Without clear person–role–permission mapping, automation only magnifies risk.

Third-Party Reconciliation Loop

For e-commerce and platform businesses, AI-driven reconciliation across banks, payments, ERPs, and DTC backends becomes essential — now a mandatory compliance item.

Human–AI Joint Governance

Since accountants retain the right to refuse or supervise, AI cannot replace human judgment. Systems should be designed around “AI output → human review → logged trace.”

Shared CFO's 4×3 Roadmap: Four Tracks × Three Months to Compliance

Track A: Organization and Roles (Month 1–2)

Build the accounting role matrix and incompatibility checklist.

Define the accounting software manager's duties.

Create rotation or audit plans and enforce conflict-of-interest checks.

Track B: Process and Vouchers (Month 1–3)

Write electronic voucher management rules: sources, validation, anti-duplication.

Configure audit rules for system-generated vouchers; all must be traceable.

Standardize correction, red entries, and numbering.

Track C: Permissions and Handover (Month 1–3)

Audit all accounts and close shared logins.

Create a unified handover checklist including seals, keys, data, and permissions.

Establish a “new account in, old account out” SOP.

Track D: Reconciliation and Supervision (Month 1–3)

Include third-party funds in monthly reconciliation.

Launch auto-reconciliation and anomaly alerts.

Keep “refusal and correction” logs and management reports.

Three Reminders from the Shared CFO to Management

Treat accounting basics as a data governance project — convert rules into system settings, audit logs, and verifiable reports.

Shift from project thinking to operational thinking — automation and permissions need ongoing review, not one-time setup.

AI is an accelerator, not a firefighter. Efficiency comes only when rules are clear, data is clean, and access is controlled.

This revision's true value lies in using “operable institutional language” to define the boundaries, paths, and tools of intelligent accounting — integrating roles, processes, systems, and supervision.

When policies and systems reinforce each other, AI can finally become a trusted engine for accounting and management.

What Shared CFO Can Offer You

Free 20-Min Diagnostic Call

A short talk to identify weak points in finance, business-finance integration, tax, funds, budgeting, and strategy, and clarify causes and priorities.

Light Implementation Diagnosis (Low-Cost Trial)

A focused review that maps key problems and gives a 3-month actionable plan for visible improvement.

Shared CFO Coaching (Short-Term Results → Long-Term Co-Creation)

After initial success, move into guided execution — completing system, process, and AI governance integration at minimal cost, while enhancing management control and business insight.