At Blueprint 4D in June, Oracle announced a significant strategy shift in AI for PeopleSoft. They would not proceed with developing turnkey AI capabilities built directly into the platform, and they would discontinue development of Picasso, their chatbot product. For organizations that had been waiting for Oracle to deliver ready-made AI features similar to what’s available in Fusion, this news may have felt like a setback.
But here’s the perspective worth considering: this update isn’t a stop sign—it’s a green light.
What Actually Changed
Oracle’s AI investments continue to focus on Fusion, their cloud-native suite. For PeopleSoft, they’re taking a different approach. Rather than building AI features you simply turn on, Oracle is providing AI frameworks that enable you to implement AI in ways that best fit your organization.
An important clarification that often gets lost in the conversation: the Oracle Digital Assistant (ODA) platform that powered Picasso is not deprecated. The infrastructure, the integrations with PeopleSoft, and any custom skills you’ve built all remain fully supported. Only the Picasso skill itself—essentially the pre-built chatbot functionality—is being discontinued.
This distinction matters because it means your existing investments in Oracle’s AI infrastructure remain valid. And it opens the door to building something that actually fits your specific needs rather than adapting your processes to fit a one-size-fits-all tool.
Why This Might Actually Be Good News
The reality of implementing pre-built AI solutions is that they rarely work perfectly out of the box. Organizations typically end up customizing and extending them anyway. A pre-packaged chatbot might handle a dozen common scenarios, but your users have questions that span PeopleSoft and beyond—questions about policies documented in Salesforce, information stored in your learning management system, or knowledge spread across institutional websites.
Good AI requires solid integration across multiple systems, not just PeopleSoft. When someone asks about their schedule, they might also want to know about parking, campus events, or how to contact their advisor. Those answers live in different places. The framework approach Oracle is now providing allows you to build AI-powered experiences that source information from wherever it lives and deliver it through a unified experience.
The technology has also evolved considerably. The effort required to build an AI solution today delivers significantly more value than it would have even a year ago. Large language models have made conversational AI more capable, more natural, and easier to implement. The building has become more efficient, and you get more out of every hour invested.
The Case for Action
The temptation might be to wait—to see how the technology evolves, to wait for something more turnkey to emerge, to push AI initiatives to next year’s budget cycle. But the data suggests that waiting carries its own risks.
Organizations implementing AI effectively are seeing dramatic results. One higher education institution deployed an AI-powered student support system and tracked their busiest month—August, when support volume typically doubles. They saw a 40% reduction in phone calls and a 65% reduction in emails. Students got instant answers instead of waiting on hold or waiting days for email responses.
When asked whether they planned to reduce headcount based on these efficiencies, the answer was telling: “No, we’re going to be able to spend ten times the amount of time with the students that need it most. Before, we just couldn’t do it. We didn’t have enough time.”
Another institution is handling 20,000 automated interactions per month. Consider how many staff members would be required to have 20,000 individual conversations with students in a 30-day period. That’s the scale of impact possible when AI is implemented well.
The Reality of Multi-AI Environments
One shift in thinking that’s becoming increasingly important: the idea of having one AI platform for your entire organization is fading. Every software vendor is building AI into their products. Your service management platform will have AI. Your CRM will have AI. Your learning management system will have AI. PeopleSoft will have AI capabilities.
The question isn’t whether to have multiple AI systems—that’s already happening whether you plan for it or not. The question is how to make these systems work together effectively. Your Oracle AI tools will handle Oracle use cases well. Your service management platform’s AI will handle help desk scenarios well. The strategic focus should be on integration and coordination rather than consolidation onto a single platform.
Build vs. Partner: An Honest Assessment
If you’re considering building AI capabilities in-house, here’s a realistic checklist of what you’ll need: AI engineers who understand prompting, guardrailing, and the differences between language models. Data scientists who can analyze large datasets and identify patterns. Deep expertise in the PeopleSoft data model and integration architecture. Someone focused on PII data flows and privacy compliance. Ongoing budget for maintenance—not just the initial build. And bandwidth to continuously monitor, refine, and improve the solution.
That last point deserves emphasis. AI isn’t a project with a completion date. It’s a journey. Every month brings changes to available models, new capabilities, and shifts in best practices. Organizations that build something, launch it, and consider it “done” consistently struggle. The AI needs ongoing attention—reviewing logs, identifying where users are getting stuck, adjusting responses, and keeping pace with technology evolution.
The statistics on this are sobering. Research indicates that 71% of in-house IT builds fail to meet deadlines or budgets. Internal AI projects show a 33% success rate, compared to 66% for projects involving external partners. Only 12% of IT professionals report having significant AI or machine learning experience.
This doesn’t mean building in-house is wrong—it means going in with realistic expectations about the investment required. If you have strong in-house capabilities and a multi-year timeline, building internally can work well. If you need faster time to value or lack deep AI expertise on staff, partnering makes sense. Many organizations are finding success with a hybrid approach: buying a foundation to get started quickly, then building selectively on top of it.
Common Pitfalls to Avoid
Several issues consistently derail AI implementations. Data privacy concerns top the list—and rightfully so. AI can be a black box, and too often organizations discover late in a project that data is flowing somewhere it shouldn’t. Understanding exactly where PII goes, who has access to model logs, and whether data leaves your control is critical.
Shadow data copies create significant risk. The temptation to copy PeopleSoft data to another system so AI can access it introduces problems: the data isn’t real-time, business rules and security may not be properly represented, and you’ve now created another data store to secure and maintain. Better approaches keep the AI connected to source systems rather than maintaining copies.
Lack of personalization undermines user trust. If someone asks where they can park and your AI directs students to faculty parking lots, you’ve created a worse experience than having no AI at all. Understanding who’s asking—their role, their context, their permissions—and tailoring responses accordingly separates effective AI from frustrating AI.
Finally, auditing often gets overlooked until it’s needed. When questions arise about what the AI said to a specific person at a specific time, you need to be able to answer them. What data did it source? What instructions was it following? When were those instructions last changed, and by whom?
Moving Forward
Oracle’s updated approach to AI for PeopleSoft changes the path but not the destination. Organizations running PeopleSoft can absolutely implement AI that delivers modern, intuitive experiences for their users. The frameworks and infrastructure exist. The technology has matured. The question is simply how you choose to proceed.
Whether you build internally, partner externally, or pursue a hybrid approach, the key is to start. You don’t need to have all the answers. Start making progress. Try things. Learn what works in your environment. AI will continue to get better every year, and organizations that start building experience now will be positioned to capture value as capabilities expand.
The door Oracle closed was a small one. The doors that remain open lead somewhere much more interesting.
This post is based on a recent PeopleSoft webinar presented by Andrew Bediz, Practice Lead for AI and User Experience at Gideon Taylor.
