Oracle applications chief sees enterprise AI agents as task-specific helpers

During this week’s Oracle AI Summit in London, Steve Miranda, executive vice-president of Oracle applications development, spoke about how the company positions agentic artificial intelligence (AI).

The company recently unveiled Oracle Fusion Agentic Applications, and Miranda used his keynote presentation at the London conference to showcase Oracle’s approach to agentic AI in enterprise applications.

Oracle launched more than 20 agentic applications across enterprise resource planning (ERP), human capital management (HCM), supply chain management (SCM) and customer experience (CX). These include automated cash collections risk analysis, workforce scheduling with real-time gap detection and AI-driven sourcing from design data.

“What we’ve done is go beyond automation and deliver enterprise agentic applications, which use the AI to advise and recommend based on business rules and data,” he said.

Significantly, Oracle sees its applications as the systems of record within a larger ecosystem with Oracle and non-Oracle AI agents. 

One concern IT leaders often raise about the use of agentic AI to automate business processes is around the safeguards in place to prevent incorrect outcomes or undesirable race conditions where the AI’s action is detrimental to the business.

Miranda said Oracle’s approach has been to build agents that work at a fine task level to run very specific apps, rather than at a board process level, adding: “Technically, this is a much easier problem to solve and much easier to have agent constraint.”

While a large language model (LLM) needs to handle any type of user input query, he said the Oracle AI agents work in a very specific domain. Oracle also adds the ability to put a human in the loop to check if the outcomes produced by the AI make sense. “If you’re using it to scan invoices, [the AI agent] can get it right or wrong. It’s very measurable,” said Miranda.

According to Oracle, Fusion Agentic Applications autonomously progress routine actions within guardrails and surface only exceptions, trade-offs and decisions where human judgement materially changes the outcome.

Processing invoices is one of the areas agentic AI can be deployed. For instance, NHS Shared Business Services (SBS) has used Oracle Fusion Applications to standardise and automate financial operations to process 7.1 million invoices, recover £7.4bn in debt and handle up to £355bn in NHS transactions per year.  

Erika Bannerman, managing director, of NHS SBS, said: “We’ve partnered with Oracle to provide a scalable, AI-powered platform that will help to transform how the NHS works so the focus stays on providing the highest standard of care for patients.”

The cost of AI is another concern both for IT leaders and technology providers. Miranda said Oracle’s approach is to continue charging for Oracle applications on a subscription basis and charge additionally for the AI functionality that is used to to extend the applications: “But there will come a day where our pricing model for our base subscription changes from user based to some sort of transaction based [pricing], or [includes pricing regarding the] size of company.”

Miranda admits he is “not terribly concerned about that day” because if the agentic AI is able to deliver an improvement, it’s a “win-win” both for the customer and Oracle.

Acknowledging that agentic AI in enterprise applications could eliminate certain manual tasks, he said: “Nobody is in business to run ERP. The more we can save them on the ERP side, the more they invest in what they do.”