All writing

Two Weeks to Three Days: What Happens When AI-Native Teams Stop Waiting on the Calendar

“Manage by intervention, not by calendar.”

That is a section heading in McKinsey’s October article on the AI operating system, and it’s the line most executives will remember from it. The instruction underneath is narrower and more useful: “CEOs should instead redesign the management calendar around decision points.” McKinsey goes on to recommend eliminating standing reviews and defining “the handful of signals, thresholds, and exceptions that should trigger leadership intervention.”

I agree with most of that. Where I part ways is on how much of the calendar should survive, and the disagreement matters most in the sectors where I spend my time: aerospace, defense, financial services, energy, and transportation. In those industries, certification gates, regulatory filings, and board-approved budgets set dates that no signal can move. That is the problem PI Outcome Planning, the planning event in AI-Native SAFe® (the Scaled Agile Framework), was redesigned to solve.

The Operating Model Is on the 2027 Agenda

The 2027 Gartner CEO Survey, conducted in April and May with 203 CEOs and senior executives, found that 63% see radically redesigning the operating model as a top move to accelerate financial returns from AI through 2027, and 52% cited aggressively deploying AI agents as employees. Gartner’s David Furlonger summarized it in one line: “The AI value problem will not be solved by simply adding more AI to existing ways of working.”

McKinsey’s article supplies the numbers behind that impatience. In its 2026 State of AI survey, 40% of large-company respondents reported scaling AI agents, while only 6% said AI generates at least 5% of their earnings before interest and taxes (EBIT). When adoption runs that far ahead of earnings, a proposal to change how the leadership team meets is a reasonable next step, and many of you will see one in the next planning cycle.

Status Meetings and Decision Forums Do Different Work

McKinsey explains that calendar-driven rhythms “emerged in part because gathering and analyzing information and aligning on decisions took time.” That sentence describes two separate jobs.

The first is moving information. Think of the monthly performance review, the deck that takes analysts a week to assemble, and the steering committee that spends most of its hour on status. When performance data is continuously available, McKinsey is right that these meetings “become less necessary,” and I’d retire them as soon as the data feeding them is trustworthy.

The second is forcing a choice by a known date. A fixed planning event tells several hundred people that funding, scope, and cross-team commitments will be settled on a specific day, so they can plan their own work around it. AI doesn’t do this job. It adds load to it, because continuous data multiplies the number of things that could be escalated at any moment. Forrester’s 2027 predictions describe leaders racing to gain control over AI and slowing their organizations in the process, and an intervention model with dozens of triggers is one way that happens.

So the useful question for a leadership team is what each recurring meeting is for. Information can flow continuously. Commitments need a date that everyone can plan around.

PI Outcome Planning Is Built Around Decision Points

Scaled Agile released AI-Native SAFe® in June, and its guidance says “the increased variability and speed introduced by AI have driven significant changes to the events.” The change most relevant to McKinsey’s argument is to the planning event for an Agile Release Train, the team of teams, usually 50 to 125 people, that delivers a single value stream. That event happens once per planning interval (PI), every 8 to 12 weeks.

Core SAFe®‘s PI Planning runs two days and centers on features, business-value scores from 1 to 10, and a planning board of cross-team dependencies. PI Outcome Planning runs one day. Scaled Agile describes its purpose as realigning the train while work is in flight, and its outputs are commitments to measurable PI and team outcomes, agreed milestones, and decisions and managed risks. Teams write outcomes as objectives with measurable key results, and Business Owners approve them instead of scoring them. Dependencies appear as synchronization milestones, which record the date when two teams’ work has to converge.

During the day, teams post items to a management review board under four headings: Decisions Needed, Risks, Decisions Confirmed, and Actions. The first heading corrects a habit common in Core SAFe® events, where an unmade decision gets logged as a risk and can sit in the register for a full quarter.

Between planning events, Sense and Respond replaces the half-day Inspect and Adapt that closed each planning interval in Core SAFe®. It’s a two-hour session held at least once every iteration, it absorbs the system demo, and Scaled Agile designed it for what it calls “continuous adaptation.”

Set against McKinsey’s prescription, PI Outcome Planning is the calendar of decisions and Sense and Respond is the intervention layer between them. The framework adds one element the article leaves open: a fixed date for commitment. The train senses every iteration and recommits to outcomes on a day everyone knows in advance, which is what lets a certification milestone or a regulatory filing sit on the same board as a learning experiment.

What This Looks Like in Practice

A Fortune 500 technology company I’m working with is piloting AI-Native SAFe® with three AI-native teams of three to five people each. They’re about to start their second planning interval, so this is early evidence from a small group.

The pilot teams dropped the iterative delivery cadence entirely. They have no fixed iterations and no iteration-boundary ceremonies. In their place, they run Sense and Respond every two to four days, called when a team member says the team needs to align or when the product owner brings back new customer feedback. The published guidance anchors Sense and Respond to the iteration and leaves additional sessions to leadership. These teams have no iteration to anchor to, and the people closest to the work decide when to meet. That was a deliberate choice, and it shortened the teams’ time to feedback from two weeks to three days.

What they kept is the commitment rhythm. The teams still commit to outcomes at quarterly planning events, and those team outcomes roll up to quarterly outcomes and then to annual ones. The delivery calendar went away. The decision calendar stayed.

The team’s product owner described the change as “unbelievable.” He said the team is more focused than ever on delivering meaningful outcomes, and that team members feel they’re getting more done. I can’t share the outcomes themselves, and after one planning interval, everything beyond the feedback time is a report of focus and perceived throughput, not a measured result. What it does show is that the pattern McKinsey describes can run at team level today: sensing called by people rather than by a meeting schedule, anchored to commitments made on known dates.

The second planning interval will test the part I care about most, which is whether those quarterly outcomes hold up when the teams are adjusting every few days.

Where the Model Can Fail

Sense and Respond can drift into a status meeting under a new name. Two hours every iteration is a generous slot, and a train under delivery pressure will fill it with progress updates because updates feel safe. Once that happens, the organization pays for the meeting and gets none of the intervention it was meant to provide. A cadence of every two to four days, like the pilot’s, raises the stakes, because frequent sessions earn their time only when each one ends with a decision or a changed plan.

PI Outcome Planning has the opposite risk. The management review runs inside a single day with little slack, so decisions that weren’t prepared in advance get made late in the afternoon by tired people, or leave the room without an owner. A one-day event that closes with unowned risks has saved a day of meetings and added a quarter of unmanaged risk.

I’d also be careful with McKinsey’s headline benefit. The article says early AI operating system deployments “accelerated planning cycles by up to ten times,” and it doesn’t publish the baseline or the sample size. I expect the direction is right. I’d plan for a smaller gain and measure your own.

Five Changes to Make Before Your 2027 Plan Is Final

01

Sort every recurring leadership meeting into two groups.

One group moves information. The other settles a decision. Retire the first group as soon as the data behind each meeting is reliable somewhere else, and give each meeting in the second group a named decision and a date.

02

Write down the signals that justify breaking the rhythm.

McKinsey's phrase is "the handful of signals, thresholds, and exceptions." If your list runs to dozens, your leadership team will spend its week responding to triggers.

03

Have Business Owners state their risk tolerance in writing.

In AI-Native SAFe®, Business Owners own the risk tolerance that teams work inside. Without a written version, every delegated decision near the edge comes back up the chain.

04

Prepare decisions before PI Outcome Planning opens.

Assemble the decision packet for the management review in the week before the event, and close the day with an owner and a date on every open risk.

05

Pair every throughput key result with a quality key result.

AI increases output faster than it increases review capacity, so an outcome set that rewards speed alone will look healthy until the defects arrive.

McKinsey’s heading will appear in a lot of 2027 planning decks. When it shows up in yours, read the sentence underneath it. A calendar built around decision points, with continuous sensing between them, is a design several hundred people can plan their work around, and AI-Native SAFe® already gives you the events to run it.

Sources

Work With Atlas

If the stakes are high, let's talk.

Atlas Revolutions partners with a small number of Fortune 100 and government leaders at a time. If you're navigating a transformation where getting it wrong isn't an option, start with a conversation.

Start the Conversation