Knowledge Hub

Oracle Fusion Claw brings enterprise AI to Indian banking

Financial institutions in India are preparing to deploy Oracle Fusion Claw applications, bringing agentic workflows, audit trails and human oversight into business-critical operations.

Banks and financial institutions in India are among organisations preparing to implement Oracle’s new Fusion Claw-powered applications, but the company has yet to identify a customer running the technology in production. As the applications enter a phased rollout, the challenge is shifting from demonstrating autonomous AI capabilities to establishing whether agents can execute business-critical processes reliably, securely and with appropriate human oversight.

Kaushal Kurapati, Group Vice President, Applications Development, Oracle, confirmed during an interaction at Oracle AI World 2026 in Las Vegas that major Indian banks and financial institutions had signed on. However, implementations were still underway. The applications are being introduced during Oracle’s 26D release period, covering October to December 2026.

At Oracle AI World on 26 October, the company unveiled 25 Fusion Claw-powered applications within a broader portfolio of 75 packaged Fusion Agentic Applications. Designed for finance, human resources, supply chains and customer experience, they aim to move enterprise AI beyond answering questions and recommending actions towards coordinating and executing complex operational workflows.

From AI Assistance to Business Execution

Fusion Claw is an agentic execution runtime designed to support enterprise processes extending over several hours or longer. It combines AI reasoning with deterministic calculations, simulations and transactional operations, while reassessing execution plans as business conditions change. Such processes frequently involve multiple applications, datasets, approvals and operational dependencies.

During a demonstration, Chris Leone, Executive Vice President, Oracle Applications Development, showed how an agentic application could coordinate shipment planning to improve load utilisation from 61% towards a target of 70%. The system evaluated operational constraints and prepared actions for approval rather than merely presenting recommendations. The demonstration illustrated the intended capability, not a verified customer productivity improvement.

The Claw-powered Customer Experience applications include Account Territory Growth Plan, Service Health Check, Strategic Field Demand Planning and CPQ Test-to-Production Setup and Configuration Migration. Oracle AI Agent Studio for Fusion Applications also allows customers and partners to develop and extend agentic applications using reusable agents.

Oracle introduced Fusion Companion, which enables employees to interact with Fusion Applications through supported conversational AI interfaces, including ChatGPT, and Fusion Agentic Home, a personalised interface for delegating tasks and monitoring progress. Together, these capabilities support Oracle’s ambition to transform enterprise applications from systems of record into what it calls systems of outcomes.

“The next era of enterprise AI is all about delivering software that increasingly does the work, while people define the objectives and set the policies,” Leone said during his presentation.

Sunil Wahi, Head of Applications, ASEAN and SAGE, Oracle, said enterprises were increasingly distinguishing between systems of record and systems of outcomes. He identified workforce productivity, manufacturing efficiency, supply chain operations and customer experience as areas attracting interest.

According to Wahi, agentic applications could shorten traditional business process re-engineering cycles by allowing organisations to visualise workflows and develop prototypes more quickly. For enterprises, the potential lies in reducing manual hand-offs and accelerating operational decisions, although these benefits remain to be demonstrated across the newly announced Fusion Claw portfolio.

How Much Autonomy Can Enterprises Allow?

Despite the expansion of agentic execution, Oracle acknowledges that businesses remain cautious about allowing AI to make consequential decisions independently. Kurapati said the degree of autonomy depended on the workflow, industry and financial exposure involved.

In finance, an agent could process invoices overnight, leaving employees to review and approve payments the following morning. Similarly, AI could summarise employee performance information, but organisations were unlikely to delegate promotion decisions entirely to agents. “It also depends on the transaction amount, the financial risk and the liability involved,” Kurapati said.

He explained that Oracle had introduced outcome receipts within Fusion Claw to record decisions, their timing, information sources consulted and changes made to systems of record. These provide an audit trail for governance teams and help organisations investigate disputed outcomes.

However, an authorised transaction can still produce an incorrect business outcome. Kurapati cited an agent approving a product return because it lacked access to the latest policy information, highlighting the importance of accurate business context alongside permissions and procedural compliance.

Kurapati also cautioned against equating longer-running workflows with fully autonomous businesses. “I would not say an entire business can be run end-to-end by AI,” he said. The advance, he explained, is that agents can undertake tasks lasting several hours or an entire day rather than being restricted to shorter interactions.

Security and Reliability Remain Critical

Kurapati described Fusion Claw’s security architecture, distinguishing it from OpenClaw, whose approach to longer-running tasks had influenced Oracle’s development. He said Oracle had not forked OpenClaw’s code. Instead, Fusion Claw uses restricted containers created for individual tasks and terminated after execution.

These environments have neither direct internet access nor direct access to the APIs of underlying systems of record. Actions are routed through agents subject to role-based access controls and permissions, limiting what they can access and execute.

Another challenge is ensuring consistent agent behaviour when underlying AI models change. Kurapati said Oracle had developed ATLAS, an agentic testing and lifecycle automation suite that allows customers to test agents using simulated datasets and rerun scenarios against updated models. Execution traces can reveal changes in tool calls, skipped steps or differences in behaviour.

Oracle also evaluates hallucination rates, execution consistency and susceptibility to prompt-injection attacks. Kurapati said the company did not combine these measures into a single trust score because enterprise AI deployments involve different categories of risk.

He added that approximately 15 agents were in staging and testing for every agent in production. The figure indicates substantial experimentation but does not establish how many newly announced Fusion Claw applications are operating in production.

The Test of Enterprise Deployment

For financial institutions in India, the question is how much decision-making authority can be delegated without weakening accountability, transaction controls or regulatory oversight. Processes involving payments and financial reconciliation require agents to use current business information, follow approved procedures and escalate decisions when necessary.

Oracle’s phased rollout means customers will receive the capabilities according to their release schedules before undertaking testing and implementation. Kurapati’s disclosure establishes Indian customer commitments, but not completed production deployments.

The developments also build on Oracle’s AI World 2025 emphasis on enterprise-specific AI reasoning, trusted private data and integrated information. The 2026 announcements represent a more concrete attempt to translate that strategy into packaged applications capable of executing operational work.

Yet technological capability and operational autonomy remain distinct. Oracle has demonstrated how agents can coordinate complex workflows, but independently verified evidence of widespread production adoption and measurable improvements across the new portfolio remains limited.

For enterprises, the next test will be whether Fusion Claw delivers sustained productivity gains while preserving human judgement, security and accountability. For Oracle, success will depend less on the number of applications announced than on how reliably they perform when customers entrust them with real business decisions.

Leave a comment

Your email address will not be published. Required fields are marked *

Popular Blogs
Related Blogs
Category Cloud

Subscribe to Our Blog

Stay updated with the latest trends in the field of IT