Play audio
Oracle’s recent industry analyst summit demonstrated significant progress in embedding artificial intelligence in its business applications—including those for finance, HR and supply chain management. This amplifies the already broad and deep functionality developed over decades. Oracle also continues to expand and extend its collaborative efforts with financial institutions, a form of network-enabled extended process facilitation that I expect will grow in importance for business software providers because it can be a source of value to customers and a form of stickiness for the application.
Last year, nearly every software provider was adding AI and generative AI (GenAI) capabilities to products in the form of assistants and copilots. 
This year, the emphasis is on agents and agentic capabilities. Like most advances in business technology, the age of agents is being proclaimed ahead of their actual arrival and practical accessibility, but they will be ubiquitous shortly. ISG Software Research asserts that by 2028, all business software providers will have incorporated agents or agent-like devices to improve productivity by streamlining processes, limiting errors and reducing the need for training.
Currently, Oracle’s agents do not fit our definition of agentic systems, but they are a useful first step to facilitate adoption by customers and are likely to evolve over the next few years into true agentic systems. I will use Oracle’s terminology in this research note. Truly autonomous agentic systems are still somewhere in the future, and user-defined guardrails will be an essential part of system designs for years to come. This does not diminish the transformative potential for agent-like technology in the near term.
Oracle’s stated design objectives for its office of finance software are aligned with industry and market trends. It sees AI, generative AI and agents as potentially eliminating low-value repetitive work to improve productivity and shorten process cycles
while increasing process quality. ISG Software Research asserts that by 2028, almost all providers of ERP software will have incorporated AI to reduce workloads and speed processes and reduce errors. Agents are useful in this regard because they enable business software providers to extend and amplify the existing capabilities of applications, as well as allow partners and customers to create extensions that suit specific needs.
Beyond the productivity gains supported by this extended functionality, agents can potentially harness predictive and generative AI more effectively and efficiently than end users. This is accomplished by designing in the best choice of algorithms or canned sets of prompts that reliably deliver useful results using the optimal amount of compute resources. Customers train predictive and generative AI systems on their data to increase the speed and accuracy of business planning, handle cash flow forecasting and management, and guide process execution faster with less training and streamline role-based searches. Oracle allows customers to use whatever language models they wish for their specific use cases. Customer-trained machine learning (ML) makes deeper, contextually aware anomaly detection possible, highlighting issues in accounting operations and possible resolutions. For instance, rather than identifying an amount that appears high or low, the system will note an amount that is out of range for the specific customer, transaction, location, time or purchasing unit. This reduces false positives and negatives—a source of frustration—and provides a better context for clarifying interactions.
Oracle has ERP AI agents available today, with many more on this year’s roadmap. In its presentations, the provider highlighted examples of how agents increase staff productivity. While their individual powers may seem trivial, agents represent meaningful incremental improvement when multiplied by millions of process motions. For example, Oracle’s software has a Document IO Agent designed to streamline and automate document-intensive business tasks such as processing invoices and payments. It can ingest, parse and characterize unstructured information in multiple languages, automatically converting incoming sales quotes and invoices into ERP system entries faster, more completely and more reliably than is typical for a human operator. Its Payments Agent helps identify the best use of working capital, accelerating those vendor payments that have the biggest discounts and are strategic from a supplier management standpoint, for instance. Ledger and Reconciliation Agents provide accountants with real-time observation and automated reconciliation to support a continuous accounting approach.
Agents also make it easier for financial planning and analysis (FP&A) groups to use predictive forecasting and planning while making it more accessible to operating managers and executives. Individuals can relatively quickly perform ongoing revenue and expense forecasting, project cash flows and identify cash flow or receivable issues among many use cases. There are also procurement and sustainability policy advisors that facilitate adherence to company requirements.
Moreover, Oracle’s AI Agent Studio, released earlier this year, enables customers, third-party software providers and partners to create, customize and implement individual agents and orchestrated agent teams. These can address an industry- or company-specific need ranging from the simple to the highly complex. These agents are built by adapting prebuilt templates with natural language prompts to facilitate complete and valid journal entries, quotes in sales motions or merchandise returns. To avoid duplicative effort and reduce the potential for technical debt, users can employ a modular approach to designing and constructing agents, using orchestration to handle a process using multi-agent collaboration.
Oracle also updated its progress in its collaborative banking arrangements that support payment facilitation. This form of network-enabled extended process facilitation is an emerging source of value to customers because, like agents, it boosts staff productivity by eliminating the need for work that is best handled by computers. Enterprises can compress process cycles and reduce the potential for errors in manual data entry and the downstream efforts needed to correct those mistakes. It also improves situational awareness by providing near real-time visibility. Commercial card services, including virtual cards, make it easier for enterprises to control and track spending. In countries with value-added tax (VAT) e-invoicing requirements, corporations can reduce the cost and risk of compliance.
From a “legacy” bank’s perspective, embedding its services in an ERP system provides a low-cost way of acquiring potentially high-margin digital finance customers and thwarting fintech rivals. Straight-through debit-based payment facilitation delivers lower-cost services to customers while creating opportunities for profitable trade finance and investment services.
The advances made since last year’s analyst event confirmed to me that, in the race to innovate, larger, incumbent providers have a potential advantage. This goes against the software industry disruption narrative that asserts that smaller, newer companies are better able to exploit new technology faster than established providers. In the past, the new players were able to reconceive how new technology could accomplish business tasks, unburdened by legacy code and incompatible sales or go-to-market models. In this meme, the old guard are dinosaurs, ready for extinction. That’s not the case with AI, GenAI and agentic AI because the technology amplifies the capabilities of the existing software, and new arrivals would have to replicate intellectual property developed over decades to be competitive.
Regards,
Robert Kugel
Fill out the form to continue reading.