The two-day PI Planning event is now one day. For a lot of SAFe practitioners that lands as a loss — the most recognizable ritual in the framework, halved.
That reading misses what actually happened. Almost nothing was removed. It was redistributed, and the direction it moved in tells you more about AI-Native SAFe than any slide about agents.
What changed, precisely
Three events shifted at once, and they only make sense together.
PI Planning became PI Outcome Planning. One day instead of two, for the entire AI-Native ART. The focus moved from producing a detailed plan for the coming increment to in-flight realignment around strategy and measurable outcomes.
Inspect and Adapt became Sense and Respond. The half-day event that ran once per PI is now a two-hour event held every iteration. It also absorbs what the System Demo used to do.
Customer Demos became their own discipline. Rather than one internal demo cadence, demo format is chosen deliberately along two axes: is the purpose learning or persuading, and is the reach broadcast or personal. A usability test, a validation demo, an executive briefing and a launch event are different instruments, and the framework now says so.
What did not change is the thing everyone assumed would: the PI itself is still eight to twelve weeks. The nested cadence survived. Only the events hanging off it moved.
Do the arithmetic
Take a five-iteration PI and add up the ART-level ceremony.
Under the old model: roughly sixteen hours of PI Planning, four hours of Inspect and Adapt, plus a System Demo each iteration. Call it twenty-five hours, with about two-thirds of it spent before the increment began.
Under the new model: seven hours of PI Outcome Planning, then two hours of Sense and Respond in every iteration — ten hours across the PI, System Demo included. Call it seventeen.
So total ceremony drops by roughly a third. But look at where the drop lands. Upfront planning falls by more than half. Iteration-level learning and realignment goes from four hours per PI to ten — it more than doubles.
That is not a cut. That is a transfer of time from prediction to correction.
Why the transfer is the right move
The old allocation encoded an assumption: that a room of smart people, given two full days, could produce a plan good enough to guide ten weeks of work. That assumption held when building was slow. The plan aged at roughly the rate the work progressed.
Agentic tooling breaks the symmetry. Build cycles compress; the plan does not age more slowly to match. A detailed ten-week plan produced in August is now stale by September, not because the planning was bad but because reality moved faster than the planning horizon assumed.
When your plan’s shelf life collapses, spending more time perfecting it upfront is throwing good hours after bad. The rational response is to plan to a coarser resolution and correct more often. Which is exactly what the redistribution does.
There is a useful piece of evidence from the Summit floor. Mik Kersten put a number on stage: agile teams spend roughly eight percent of their time on direct value creation. The rest is absorbed by upfront and downstream business process. If that figure is even directionally right, then trimming the largest single block of upfront process is not an attack on rigor — it is the obvious first move.
The structure of the new day
The one-day agenda is more interesting than its length.
The morning opens with business context — vision, ART outcomes, strategy — then alternates: an outcome breakout, a team breakout, another outcome breakout to draft the planning board and surface decisions needed and risks. A draft board presentation, then management review and problem-solving over lunch. The afternoon repeats the alternation at a faster tempo — planning adjustments communicated, outcome breakout, team breakout, outcome breakout to finalize — and closes with a final plan review.
The alternation is the design. Old PI Planning ran essentially one loop: teams break out, plan, present, adjust, present again. The new day runs two nested loops, deliberately interleaved. Outcome groups work on how a shared goal gets achieved across teams. Teams work on capacity and constraints. Each hands off to the other, twice.
That is why it fits in a day without feeling like a compressed version of the old one. It is not the old event with things deleted. It is a different shape.
The confidence vote survives, which matters more than it sounds. Commitment still gets made by humans, out loud, in front of each other.
The artifact that carries it
Underneath all of this sits the outcome tree: ART outcomes at the top, PI outcomes below, team outcomes below that. Purpose communicates downward, commitment communicates upward.
It is worth noticing that Kersten argued for almost exactly this structure from the keynote stage — outcomes cascading up, strategy cascading down, replacing organizational structures built around regions and functions. Two independent lines of argument arriving at the same artifact in the same week is a reasonable signal that the artifact is load-bearing.
The practical consequence is that the planning board stops being a dependency map and becomes an outcome map. Milestones are typed — fixed-date, learning, synchronization, release — which forces a distinction most boards blur. A learning milestone is not a delivery date. Treating them identically is how organizations end up committing to discoveries they have not made yet.
The honest difficulty
The uncomfortable part is not the day that disappeared. It is the ten hours that appeared.
Two hours every iteration, with the whole ART, requires something most organizations do not currently have: the data to answer whether outcomes are actually moving, real signal on how customers responded to what shipped, and the discipline to fix the workflow that matters rather than the complaint that is loudest.
An ART that cannot answer those questions with evidence will run Sense and Respond as a status meeting. And a status meeting every two weeks is worse than the half-day it replaced, because it costs more and teaches less.
So the real prerequisite for one-day planning is not that your teams got faster. It is that your instrumentation got better. Organizations that move to the new event structure without solving for signal will experience it exactly as it first appears — as a cut.
Those that solve for signal first will find they gave up a day of forecasting and bought six hours per PI of finding out whether they were right.
Three questions before you shorten the event
1. Can we answer, with data and within two hours, whether our outcomes moved? If not, Sense and Respond has nothing to sense. Fix that before shortening anything.
2. What in our current two days is genuinely irreplaceable — and is it the planning, or the fact that everyone is finally in one room? Those are different needs, and only one of them requires two days.
3. Are our milestones typed? If learning milestones and delivery dates look the same on your board, the one-day event will surface that immediately and painfully.
The world has changed. The leaders who notice will be the ones the next decade is built around. I write on Thursdays at agile-agilist.com.
