Oracle has put AI agents inside Primavera Unifier, and it is worth being precise about what that means, because the precise version is less exciting than the headline and more useful to know. The new AI doesn't write a schedule, judge its logic, or say whether a critical path is real. It summarises workflows, moves data between systems automatically, and links routine site photos to risk models. That's a genuine improvement to how a capital programme gets administered. It has almost nothing to do with whether the programme underneath is any good.
What Oracle Actually Announced
Oracle's April 2026 announcement for Primavera Unifier describes three things. Workflow and business process summarisation is built into daily task execution, so a review cycle produces its own record of participants, timestamps, decisions and attachments without someone assembling it by hand. Oracle Integration automates data movement between Unifier and other enterprise systems, ERP, asset management, scheduling and collaboration platforms, through pre-built adapters and event-driven triggers. And routine progress photos and field observations get fed into Oracle's Construction and Engineering Advisor for Safety, sharpening how risk gets flagged from what is happening on site.
All three are administrative wins. None of them touch the schedule's logic.
The Feature That Would Matter Is a Different, Older One
Unifier sits in Oracle's own product line as capital program management and project controls, the budgets, workflow and document layer around a programme, not the scheduling engine itself. The tool that actually tests whether a schedule's logic holds up lives in a different product: P6 Professional's built-in Check Schedule Report, available from version 22.12 onward, and Oracle Primavera Cloud's Schedule Health Score, its own name for the equivalent capability. Both existed before this AI wave arrived, and this announcement doesn't extend either one. The part of Oracle's suite that got an AI upgrade is the part that moves information faster. The part that judges whether the plan is sound was not on the list.
Faster Administration Around a Bad Plan Is Still a Bad Plan
That gap matters more, not less, once the surrounding workflow speeds up. If routing, sign-off and summarisation all move quicker, a programme carrying negative float propped up by constraints, or progress statused against forced dates rather than genuine forecasts, passes through review with fewer people lingering over any one document, not more scrutiny at the point where scrutiny is actually needed. A faster process around a plan isn't the same thing as a better plan. It just means a weak one reaches the next stage sooner, with a cleaner audit trail sitting behind it.
We've made a version of this argument before, about AI more generally in schedule work: who checks the schedule the AI wrote. This is the sharper, specific case of it. The AI isn't inside the plan. It's inside the paperwork around the plan, and paperwork moving faster isn't a signal that the thing the paperwork describes has improved.
The Honest Counter-Argument
Dismissing this as marketing gloss would be wrong, and it's worth saying so plainly. Reducing the hours a project controls team spends chasing sign-offs, reconciling change history across disconnected systems, and manually tagging field photos to risk categories is real time given back to people who could be doing more useful work. It also strengthens exactly the kind of record-keeping that matters if a dispute ever needs reconstructing months later. A clean, automatically-generated workflow trail, who decided what and when, is a genuine asset if a programme's history is ever tested. Firms running large capital portfolios through Unifier will feel that benefit immediately, and it's not a small one.
None of that is the same claim as "the schedule is sound," though, and it shouldn't be read as one. Better paperwork about a programme and a trustworthy programme are two different things that happen to look similar from the outside, and the whole risk of this generation of tools is mistaking the first for the second because it arrived wrapped in the word AI.
What Actually Needs Checking, Regardless of the Tool
Whatever administrative layer sits on top, and whichever vendor's AI happens to be inside it this year, the question a Project Manager, funder or board actually needs answered stays the same and stays narrow: does this programme's logic hold up, is the critical path genuine, and does the float tell the truth. That is a DCMA-style schedule quality check, and it is worth running on its own terms, independent of whichever software produced the schedule or how smoothly it moved through review to get there.
Run your own .xer or .xml through Planned's free schedule health check to see where the logic actually stands, whatever built it. Where more is riding on the answer, a resubmission after a rejection, or a programme underpinning a live claim, Planned's Independent Schedule Assurance Review tests the plan properly before anyone downstream relies on it.
If you want that assurance done independently, see our AI Schedule Assurance service: a human review that an AI-generated or AI-assisted schedule holds up against the contract and the site.