From Explanatory Chatbots to Execution-First Finance AI Agents
Agentic AI execution in finance refers to software agents that not only analyze data or answer questions but also perform ERP-adjacent accounting tasks, such as posting journals or resolving invoice errors, while following defined policies and leaving a usable audit trail for controllers and auditors. This marks a break from first-generation AI chatbots, which focused on explanation, search, and summarization. Finance AI agents aim to reduce manual work by acting inside or alongside ERP automation, handling recurring workflows like accruals, reconciliations, and tax checks under human oversight. At events such as Sage Future, vendors and CFOs are debating how far this execution should go and which safeguards are required. The core tension is clear: the more autonomous accounting systems become, the more finance departments must treat AI as part of their control environment rather than a conversational helper.
SAP’s Autonomous Finance Roadmap Puts Governance on the Clock
SAP’s Autonomous Finance portfolio shows how fast execution is moving ahead of governance. The company plans four Joule-powered assistants for general availability in Q2: Financial Closing Assistant, Tax and Compliance Assistant, Billing Assistant, and Accounts Receivable Assistant. These agents focus on concrete accounting execution, from journal validation and intercompany reconciliation to e-invoicing error resolution and collections workflows. Cash and Treasury Assistant sits in a mixed status with both Early Adopter Care and general availability labels, while Financial Planning Assistant is expected in Q3 and a Governance Assistant in Q4. This staggered rollout means CFOs may test autonomous accounting systems in production before the full AI governance finance layer is ready. According to SAPinsider, finance teams now face a “planning gap between early benefits and the controls needed to scale those workflows safely,” turning adoption into a sequencing decision as much as a technology one.
Nominal and the Push for Agentic Performance Alongside ERP
Nominal’s approach at Sage Future highlights why execution is becoming the central design goal. CMO Stephanie Montelius draws a sharp line between AI chatbots that explain and finance AI agents that execute. In her view, conversational tools may answer questions but often leave operational workloads unchanged. Nominal instead focuses on agentic AI execution that sits alongside ERP systems, follows standard operating procedures, and performs finance tasks such as intercompany processing and reconciliation with human approval steps built in. This ERP-adjacent model appeals to high-volume, multi-entity businesses that struggle with repetitive accounting and complex intercompany networks. By staying ERP-agnostic, Nominal aims to plug into existing ERP automation without forcing a replatforming project. The discussion at Sage Future underlined that trust now depends as much on determinism and procedural compliance as on model accuracy or conversational finesse.
Execution Changes Risk: Why Agentic AI Needs Stronger Controls
When AI agents touch core accounting flows, every action becomes both operational work and control activity. A closing assistant that posts journals and resolves reconciliation issues shapes the audit evidence that controllers and external auditors will later rely on. A tax assistant that monitors legal changes and supports IFRS or BEPS Pillar Two tasks influences statutory reporting outcomes. SAPinsider notes that CFOs must evaluate autonomous tools “less like productivity software and more like systems of record extensions,” asking whether each agent decision leaves durable evidence instead of a black-box trace. Agent readiness therefore depends on process maturity: organizations with clean data, defined approval chains, and consistent exception handling are better placed to adopt autonomous accounting systems safely. Without that foundation, agentic AI execution risks amplifying weak controls and widening the gap between mature and still-standardizing finance environments.
From AI Explainers to Accountable Co-Workers in Finance
The shift from explanation to execution forces finance leaders to rethink how AI governance, GRC processes, and ERP automation fit together. Conversational copilots can be governed like decision-support tools; execution-first finance AI agents need role definitions, segregation of duties rules, and change management aligned with existing control frameworks. SAP’s timeline underscores the challenge: execution-capable assistants arrive before a dedicated Governance Assistant, so CFOs must design interim oversight models that can later scale. At the same time, platforms like Nominal show that agentic AI execution can be framed as performance management rather than opaque automation, with agents following standard operating procedures and keeping humans in the approval loop. For finance departments, the strategic question is no longer whether AI can explain data, but how to make autonomous accounting systems accountable co-workers inside controlled ERP environments.
