A PeopleSoft consultant I know ran a test last month.
He pasted a nasty piece of custom PeopleCode into an AI and asked it to explain what the code did. Back came a clean, confident, well-organized answer. Paragraphs of it. Correct syntax, correct logic, correct reading of every line on the screen.
It was also wrong.
Not wrong about the code. Wrong about the system. The customization encoded a union rule that capped a certain overtime calculation, and the AI had no way to know that. It read the lines. It missed the reason. The explanation would have sailed through a code review, right up until it broke a paycheck.
That gap is the whole story. So let me tell it properly.
The question everyone is asking is the wrong one
“Will AI take my job” is the question in every PeopleSoft Slack and every LinkedIn comment right now.
Wrong question.
Code Assist, the AI chat window planned for Application Designer in 8.63 tools, is going to be good. It will explain PeopleCode. It will suggest fixes. It will draft new code faster than you can type it. Fighting that is like fighting a calculator.
Here is the better question. What can AI get confidently wrong about your system, and who is in the room to catch it.
Confidence is the dangerous part. A junior developer who is unsure asks someone. An AI that is unsure sounds exactly like an AI that is right. It does not hedge. It does not know what it does not know about your shop. And your shop, like every PeopleSoft shop, is held together by things that were never written down.
That is not a knock on AI. It is a description of where the value moved.