Oracle has been clear that generative AI will not be embedded directly into PeopleSoft applications. At first, that can read as reluctance, as though PeopleSoft were standing apart from a shift happening everywhere else. A closer look suggests something far more considered.
Rather than building its own AI into the application, Oracle has focused on the infrastructure that connects PeopleSoft data and operations to the agentic platforms and large language models each customer chooses, while keeping PeopleSoft security firmly in place. With PeopleTools 8.63, now available on OCI and on-premises, that infrastructure has arrived.
Why it helps
Oracle's approach rests on three principles. Together, they place the customer, rather than the vendor, in control of how AI enters and is involved in the organization.
No vendor lock-in
Customers select the AI services, models and use cases that correspond to their own environment, rather than adopting whatever AI a single vendor has chosen to build in.
This principle runs through the 8.63 building blocks. MCP server support allows the AI agents an organization chooses to find and use PeopleSoft APIs. The Natural Language Search Service answers plain-language questions through a REST API, so it can sit behind whichever assistant an organization places in front of its users. Even Application Designer AI Assist works with the customer's own choice of LLM.
PeopleSoft, in effect, takes its place within an organization's broader AI landscape rather than operating as a self-contained island.
PeopleSoft stays the secure system of record
Any serious conversation about connecting AI to an ERP arrives at two questions: what the AI can see, and what it can do. Oracle's answer to both is that PeopleSoft security continues to govern.
Search returns only what the signed-in user is permitted to see, so an assistant built on the Natural Language Search Service never widens anyone's view of the data. AI agents work through PeopleSoft APIs that administrators choose to expose, and PeopleSoft security and business rules still apply. PeopleTools 8.63 also adds native SAML 2.0 single sign-on and built-in two-factor authentication (TOTP), part of the groundwork responsible AI adoption requires.
AI becomes a new means of reaching PeopleSoft data, governed by the same rules as before, rather than a way around them.
Configuration over customization
AI is introduced into existing business processes with less custom code, allowing organizations to adopt it without adding to the custom code a team must carry forward.
The Magic Button illustrates the idea well: a button next to long text fields that runs an AI action, set up entirely through configuration with no page customization. Oracle has described using it to summarize peer-review feedback, draft performance-review comments, and summarize or improve long text entries.
What it asks of you
None of this delivers value on its own. The benefit depends on the AI an organization connects and the use cases it configures, which leaves three decisions in its hands: which AI services and models to connect, which PeopleSoft APIs to expose to AI agents, and which use cases to pursue first. On that last question, Oracle's guidance is to begin with repetitive tasks, high-volume questions, search pain points, development work, and processes suited to natural-language or agentic experiences.
Oracle has kept PeopleSoft as the secure core and entrusted the AI choices to its customers. The organizations that approach those choices deliberately will be the ones best positioned to benefit.