How to Run SAFe PI Planning Using AI Tools

PI Planning is the single most important event in a SAFe Release Train. When it works, the whole ART leaves the two days aligned, energised, and with a shared understanding of what the next 10 weeks are actually about. When it fails, teams leave with a board full of stories nobody owns and risks that were politely ignored.

I have been in both kinds. The difference is almost always preparation — how much thinking was done before the event rather than during it.

AI tools have cut my PI Planning preparation time roughly in half. Not because they do the thinking for me, but because they remove the blank-page friction from every component of the preparation.

Before PI Planning: The Four Preparation Jobs

There are four things a Product Owner needs to walk into PI Planning with. Most POs start preparing two or three days before the event and scramble. With AI assistance, I now start four days out and arrive with everything done.

1. PI Objectives

PI objectives need to be SMART — specific, measurable, achievable, relevant, and time-bound. They also need a business value score so the ART can prioritise during the event if capacity forces trade-offs.

I am a Product Owner preparing PI objectives for the upcoming Program Increment. Our team focus this PI is: [describe your PI theme]. Write 4 SMART PI objectives. For each objective include: the objective statement, a business value score from 1 to 10, and a stretch goal variant. Format them ready to present in a SAFe PI Planning session.

2. Feature Breakdown

Features that arrive from Product Management are often too large to plan directly. Breaking a feature into sprint-ready stories is work that usually happens in refinement, but for PI Planning you need the breakdown ready before the event starts.

Product Management has given our team this feature to deliver in the next PI: [paste the feature description]. Our PI consists of 5 sprints of 2 weeks each. Break this feature into 6 to 10 user stories that can be planned across those sprints. For each story include: the story title, a one-sentence description, a rough story point estimate using Fibonacci sizing, and which sprint it should ideally land in.

3. Dependency Identification

Dependencies are the thing that derail ARTs. Most teams discover them during the PI Planning event itself, which wastes half a day.

My team is delivering these features in the next PI: [list your planned features]. Identify the likely dependencies by category: technical dependencies on other teams or systems, feature dependencies where another team needs to deliver something before we can start, and operational dependencies on infrastructure. For each dependency suggest which team most likely owns it and what agreement needs to be in place before PI Planning.

I send this output to the relevant team POs three days before PI Planning. You resolve maybe 40% of the dependencies before you walk in the door.

4. Risk Identification for ROAM

We are planning a Program Increment with the goal of [describe PI goal]. Our team of [X] people will deliver [summarise planned features]. Identify the top 5 to 7 risks for this PI. For each risk: write a one-sentence risk statement, classify it as Resolved, Owned, Accepted, or Mitigated, and if it is Owned or Mitigated suggest a specific mitigation action and who should own it. Format the output ready to present in a SAFe PI Planning ROAM session.

During PI Planning: What AI Cannot Replace

None of the above replaces what happens in the room. PI Planning is a social event as much as a planning event. The conversations between teams, the informal agreements made in the breaks, the moment a Systems Architect spots a dependency that nobody had noticed — these are things that only happen when people are in the same space talking to each other.

AI preparation makes those conversations better because the teams arrive with more context. They spend less time building the picture from scratch and more time refining it together.

The Bottom Line

PI Planning preparation used to take me the better part of a week. Now the first drafts of all components take a single afternoon, and I spend the rest of the preparation time on the things that require actual human judgement.

The prompts in this article are part of the 30 AI Prompts for Product Managers — a pack built specifically for POs in SAFe environments. Six of the thirty prompts cover PI Planning alone.