Choosing how to frame intent

Reading this as an agent: you are the coach. Run it as a conversation, a question or two at a time. The deliverable is not a filled-in canvas — it is a decision about which frame fits this work, and a local skill configured around it. Resist filling anything in until the choice is made.

Why this exists

There is no correct template for stating what an initiative is for. There are several good ones, each shaped by the kind of work its author was doing, and the cost of picking the wrong one is real: a frame built for product features makes an infrastructure programme look empty, and a frame built for portfolio bets makes a two-week experiment look ceremonial.

This is also the honest position of the wider Portfolio PDLC material. It is not a methodology and it does not issue templates. It is a way to navigate the journey — and part of navigating is picking instruments that suit the terrain you are actually on.

So this skill does not hand over a canvas. It explains what the options are good and bad at, helps someone choose, and then writes them a local skill around that choice.

That last part is deliberate. For a thinking artifact, AI is the coach, not the generator — someone who has never worked a frame themselves should run one a couple of times before an agent fills it in for them, and the version they argue their way to is the one that survives a sponsor. So the output is a skill that will ask them questions later, not a document with the answers already in it.

The curated set

Four options, in the order worth considering them. None is reproduced here — each links to its source, which is where you should send anyone who wants the real artifact.

Lean Product Canvas (and Lean Strategy Canvas)

Jeff Gothelf and Josh Seiden, published through Sense & Respond. Two linked canvases: a strategy one that runs goal → obstacle → where to play → how to win → OKRs, and a product one that runs problem → users → outcomes → solutions → hypotheses → what to learn first and the least work needed to learn it.

Strongest when there is genuine uncertainty about whether users will behave differently, and someone is willing to be wrong in public. The hypothesis box is the load-bearing part: it forces a falsifiable claim linking a business outcome to a user benefit through a specific solution.

Weakest when the work has no user-facing behaviour change to test — a compliance deadline, a platform migration, a contractual obligation. You can fill the boxes, but the hypotheses come out hollow, and hollow hypotheses train people to treat the whole exercise as paperwork.

This is Yuval's default. It carries the strategy altitude and the product altitude in one artifact and refuses to let an initiative be its own goal.

Both canvases, and the free PDF download: https://www.senseandrespond.co/the-lean-product-canvas
Gothelf on its lineage from the Lean UX Canvas: https://jeffgothelf.com/blog/the-lean-product-canvas/
Licensed CC BY-NC-SA on the artifact itself. Download it from the source rather than copying it out of anywhere else, and note the NonCommercial term before you build a paid deliverable on it.

SAFe epic hypothesis statement

Scaled Agile's one-page frame for a portfolio epic: the epic description, a business outcome hypothesis, leading indicators, and non-functional requirements, with an MVP defined before the rest is funded.

Strongest when the organisation already runs SAFe. The frame is the one the portfolio Kanban expects, the vocabulary matches the governance around it, and using something else means translating at every gate.

Weakest when it becomes the funding artifact rather than a thinking one. The leading-indicator field is the tell — filled honestly it is the best part; filled to satisfy a template it is where watermelons are born.

The epic: https://framework.scaledagile.com/epic/
The template itself, as a .docx: https://framework.scaledagile.com/wp-content/uploads/2025/03/Epic-Hypothesis-Statement.docx
© Scaled Agile, Inc.; use their material under their permissions, not a copy of it.

A feature template

Mark Richards' feature template, refined with customers over years, pulls naming, problem statement, hypothesis and prioritisation into one artifact sized for a feature rather than an epic or a product.

Strongest when the unit of work is a feature heading into PI planning and the team keeps arriving with solutions that have no stated problem.

Weakest when used above its altitude. It is deliberately feature-sized; stretching it to cover a portfolio bet loses the strategy context that the Lean Strategy Canvas carries.

The template, from Mark Richards: https://www.shapingagility.com/blog/feature-template
His reasoning on the SAFe Fellow blog, Crafting Clarity: Using Feature Templates to Shape and Communicate Intent: https://framework.scaledagile.com/blog/crafting-clarity-using-feature-templates-to-shape-and-communicate-intent/

Shape your own

Sometimes none of the above fits — an AI use case where the user is an internal team and the benefit is time rather than delight, a research programme, a partnership. Then the honest move is to name the four or five questions this particular kind of work keeps getting wrong, and make those the frame.

Strongest when you have run one of the above two or three times and can say specifically which box kept coming out empty and why.

Weakest when it is the first move. Rolling your own before you have felt a real frame push back on you usually reproduces your existing blind spots with better formatting.

How to run this

Establish the work before discussing frames. Most of the choice falls out of three answers:

What is the unit — a portfolio bet, an epic, a feature, a use case, a programme? This mostly settles the altitude.
What is genuinely uncertain: whether anyone wants it, whether it can be built, whether it will be adopted, or only when it lands? A frame with a hypothesis box is wasted on work whose only uncertainty is scheduling.
What does the surrounding governance already expect? If a portfolio Kanban wants an epic hypothesis statement, a beautiful canvas nobody reads is not a win.

Then say which you would pick and why, in one or two sentences, and name the runner-up and what would make it the better call. Do not present a matrix.

If they cannot choose, invert it. Ask what happens if they use no frame at all — just write a paragraph and go. Then name which of the failure modes below they are choosing to own. Most people pick a frame quickly once the question is which symptom they are accepting rather than which template is best.

Watch for the two failure modes:

The initiative as its own goal. "Roll out the platform" is a solution wearing a goal's clothes. Whichever frame gets chosen, the goal slot holds a business outcome the sponsor would still own if this initiative were cancelled.
Frame shopping. Someone who has rejected three frames usually does not have a template problem — they have an initiative nobody can state a purpose for, and changing the form will not fix that. Say so.

Then configure a local skill

Once the choice is made, write them a skill in their own workspace, shaped around the frame they picked. This is the deliverable.

Put it where their agent will find it — skills/<name>/SKILL.md for Claude Code and Codex, or wherever their harness resolves skills.
Build the coaching sequence from the frame's own logic, in their words and their domain's vocabulary. Do not paste a copy of someone else's canvas into their repo — link to the source and coach against it, which is both the licence-safe path and the one that survives the source being updated.
Carry the source attribution into the skill they end up with, including the licence where there is one. A skill that quietly encodes someone's framework with no trail is a problem for whoever inherits it.
Include the two failure modes above. They recur regardless of frame.
Keep it a conversation: one or two questions at a time, repeat back what was heard, mark what is confirmed versus inferred.

If they would rather not have a local skill, a one-page written frame is a fine outcome. The point is that they chose deliberately, not that a file exists.

Output

End with:

The frame chosen, and the one or two sentences of why.
The runner-up, and what would change the call.
What is genuinely uncertain here — the thing the frame has to earn its place by exposing.
A local skill configured around the choice, or an explicit decision not to build one.

Source

The frames above are other people's work. This skill teaches the differences between them and points at the originals; it does not reproduce any of them, and you should not either.

The Lean Product Canvas and Lean Strategy Canvas are Jeff Gothelf and Josh Seiden's, published through Sense & Respond and licensed CC BY-NC-SA on the artifact. Get them from https://www.senseandrespond.co/the-lean-product-canvas.

If you want to go deeper than a canvas read, Sense & Respond Learning run a Lean Product Discovery class, delivered by their Certified Training Partners. Yuval Yeret is one of those Certified Training Partners — so treat this as a recommendation with an interest attached, and discount it accordingly. The material is Gothelf and Seiden's either way, and their books (Lean UX, Who Does What By How Much?) cover the thinking without a class.

The epic hypothesis statement is Scaled Agile's (© Scaled Agile, Inc.) — https://framework.scaledagile.com/epic/ — and the feature template is Mark Richards' at Shaping Agility.

These are Yuval's questions about other people's frameworks, not those authors' endorsement of anything here.

intent-framing-chooser