The news
Oracle (ORCL) announced Oracle Fusion Claw on September 29, 2026, a runtime that lets AI agents inside its Fusion business applications carry out complex finance, HR, supply chain and sales work rather than only assist with it. It launched with 25 new agentic applications.
Oracle said each Claw task starts with a frontier AI model that reasons and plans, then hands the work to deterministic enterprise computation, meaning conventional software that calculates and executes transactions the same way every time. Oracle argued the split applies AI only where judgment is needed and keeps high-volume execution cheaper. Claw runs on Oracle Cloud Infrastructure and initially uses Google's Gemini and OpenAI models, with more planned. The launch brings Oracle's Fusion Agentic Applications portfolio to 75.
The examples show how far into core systems the agents reach. Ledger works across large volumes of accounting entries to reconcile information, investigate exceptions and improve readiness for the financial close. Workforce Staffing builds staffing plans that weigh skills, accreditations, availability, labor rules and costs, and re-plans as conditions change. SiliconANGLE described a case in which an employee's absence leads the agent to reopen the plan and carry out authorized changes. Shipping Consolidation and Account Territory Growth Plan cover logistics and sales territories.
Control runs through what Oracle calls an Enterprise Operating Envelope: a company's objectives, procedures, policies, permissions, risk thresholds, decision rights, approval requirements and escalation limits. After each task, an Outcome Receipt records the authority applied, evidence used, decisions made and transactions executed. Customers set the automation level, from quick assistance to full auto within delegated authority. Natalia Rachelson, an Oracle senior vice president, told SiliconANGLE, "Honestly, any task can be in full auto mode," though she said organizations typically start with human review.
SiliconANGLE reported that Claw comes at no additional charge with Oracle's agentic applications product, which is bought separately from standard Fusion subscriptions, and that customers pay for AI units consumed as work completes. It also reported that customers cannot yet pick their own lower-cost open models for Claw tasks. Oracle CEO Mike Sicilia framed the launch as a move from AI assistance to execution.
The numbers
- New Claw-powered agentic apps
- 25
- Total Fusion Agentic Applications
- 75
- Models supported at launch
- Gemini and OpenAI
- Automation levels
- Quick assistance up to full auto within delegated authority
Why CEOs should care
For CFOs and controllers, an agent that reconciles and investigates entries is working inside the general ledger, so the controls question comes before the productivity question. Before enabling Ledger beyond assist mode, set approval thresholds by amount and account, make sure the agent's identity cannot both prepare and approve the same entry, and confirm Outcome Receipts are retained and mapped to your audit and Sarbanes-Oxley (SOX) controls. Bring your external auditor in early, and pilot on a closed period where you can compare the agent's work with your team's.
For CHROs, Workforce Staffing touches hours, pay and compliance. An automated change to a schedule can trigger overtime, breach a union agreement or conflict with local scheduling rules. Require human approval for any change that affects pay or hours, decide how employees are notified when an agent reworks a roster, and test how the agent handles conflicting constraints, such as a coverage requirement that can only be met with overtime.
For CIOs and procurement, the cost model is consumption-based. Ask Oracle for AI-unit price sheets, spending caps and usage forecasts before rollout, and get clarity on data handling when tasks call Gemini or OpenAI models. Because customers cannot yet choose cheaper open models, per SiliconANGLE, budget for frontier-model pricing on high-volume workloads.
The bigger picture
Enterprise software vendors are moving from chat-style assistants to agents that change records, and Oracle's design reflects a well-known limit of language models: they can plan, but they are not reliable calculators. Splitting planning from execution, and producing a receipt for every run, is Oracle's answer to the audit question every finance and HR leader will ask. Accenture, Deloitte, KPMG and PwC all supplied supportive statements for the launch, a sign that the large consulting firms expect implementation work.
What’s next
Watch for published AI-unit pricing, the first customers willing to name processes running in full auto mode, and support for more models. Also watch how audit firms treat Outcome Receipts as evidence in financial-controls testing.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
