Iterate to the Destination

Problem

No destination, or a fake one

Pattern 01 says to define what "done" looks like before you start. Most of the time that's right. But some work can't be defined up front: a new strategy, an early product idea, a problem nobody has framed yet. You don't know what good looks like until you've seen something that isn't it.

People handle this in one of three ways, and all three fail. Some skip the destination and just start prompting. The AI is happy to go wherever they point it, so they go around in circles - twenty drafts, each different, none better, and no way to tell when to stop.

Others write a destination anyway and treat it as fixed. Three rounds in, they've learned the real question is different, but the original goal is still steering the work. They end up with a polished answer to the wrong question.

And some do iterate - but with no limit. The destination changes every week and never locks. The work feels busy and never finishes, and the sponsor stops believing it ever will.

The failure isn't starting without a clear destination. It's never writing down where you think you're going, so you can't tell whether you're getting closer.

Value

Exploration that converges

Each round gets closer - because each draft is judged against a destination, even a rough one.

What you learned is written down - the changes to the destination are the record of what the exploration taught you.

You know when to stop - when the destination stops changing and the draft meets it.

Sponsors see progress, not churn - the sponsor gets one page: the current destination and its version history, with what changed and why. That's something they can approve in two minutes; twenty unrelated drafts are not.

Measure it: how many rounds until the destination locks, and how often a locked destination had to be reopened later.

Context

When this pattern applies

Use it when you can't fill in Pattern 01's destination honestly before you start: new strategy or positioning, an early-stage product or service idea, research into a question nobody has framed, a first attempt at something your organization has never done.

It's also the right pattern when a sponsor says "I'll know it when I see it." Turn that into a written guess you can both react to.

Organizations already know this discipline from funding. They fund a discovery or research phase before they fund the final product, because nobody can write the final destination yet - and the research phase ends with a sharper destination and a decision. This pattern is the same discipline at the scale of one piece of AI work.

When not to use it: when the destination is knowable and you're just skipping the work of writing it down. If you could fill in Pattern 01's format, audience, constraints and success criteria in five minutes, do that instead.

Solution

Write the destination as a guess, then rewrite it every round

Step 1

Write your best-guess destination.

Use Pattern 01's four parts - format, audience, constraints, success criteria - and fill in what you can. Label it "Destination, version 1 - a guess." Blanks are fine; name them as questions. Name the Business Owner - the person who has to agree with the destination - and show them version 1. A destination the Business Owner hasn't seen is a guess the team made alone.

Step 2

Ask for options, not an answer.

In early rounds, ask the AI for two or three different takes, not one polished draft. Make them differ on one thing that matters - the audience, the format, or the approach - so the comparison teaches you something. Three versions of the same idea teach you nothing.

Step 3

React, then rewrite the destination.

After each round, say what you liked, what you didn't, and why. Then update the destination - not just the draft - with what changed and why. "Version 2: it's for the operations leaders, not the board - because the board only approves; operations has to use it" is the real progress.

Step 4

Keep the versions.

Keep each version of the destination, with one line on what changed and why. That history is what you learned, and it's what you show a sponsor.

Step 5

Time-box it, lock it, then switch to Pattern 01.

Set the limit up front - a number of rounds or a date. When the destination stops changing for a round, lock it and say so. If it isn't locked when the time box runs out, the Business Owner decides. From there, the work is ordinary Pattern 01: drafts against a fixed destination until it's met.

Notes

A companion to Pattern 01

Use it with Pattern 01. Begin with the End in Mind says to define the destination first. When you honestly can't, this pattern is how you get there - and once the destination locks, you switch back to Pattern 01.

An example. The first plan for our ten-minute sessions was eight slides and a linear run sheet. It didn't tell the story the way it needed to. The destination was rewritten - a five-act arc, about ten minutes, half of it live - and every draft after that moved toward it. Version 1 was a guess; it did its job by being wrong in a useful way.