Write the roadmap narrative you would present to a board.

A board narrative isn't a features-and-dates slide with AI written on it. Loomweave Textiles runs a defect-detection system called LoomEye on its production lines. Fenn Adisa is the AI PM writing the narrative for the board's next quarterly review, after the first version of that narrative already went wrong once.

How to answer this, stage by stage

Nobody is scoring whether you can format a slide. They're scoring whether the narrative survives contact with the one item that doesn't go as planned.

Let's learn

Here is what happens when a board narrative leads with a single reassuring number instead of the actual size of the problem it's solving.

Before LoomEye, an inspector checked every bolt of fabric by eye at the end of each production line, roughly forty seconds a bolt, catching most real defects but missing the faint ones, the tension lines and dye bleeds that only show under the right light. Loomweave's defect returns cost the company about four hundred and ten thousand dollars a year. With LoomEye flagging likely defects for a final human check on two lines, that cost started to fall.

This is the whole job LoomEye touches. None of it needs a board slide to explain.

Here's the turn: Fenn's first board narrative led with "94 percent detection accuracy achieved," a real number, measured honestly, on the two lines running well-lit standard cotton. The board heard "94 percent" and stopped asking about the AI system in every meeting after that. Nobody in the room realized the number said nothing about the other four lines.

At its worst, a single confident number on a board slide doesn't just overstate progress. It teaches the board to stop asking the one question that would have caught the gap: which part of the problem does this number actually cover.

What I would leave alone: the three-tier structure isn't needed for line-level scheduling decisions Fenn makes with the plant floor. That's an operational detail the board doesn't need broken out; it only needs to see the dollar-cost view.

The lesson: a board narrative isn't dishonest because it hides bad news. It's dishonest when one true number gets asked to stand in for a claim it was never measured against.

Now here is the same thing as a story

The short version above is what you'd say defending this quarter's narrative to Fenn's own manager. Read this one for how a new face in the room asked the question everyone else had stopped asking.

Elin Marchetti has inspected fabric at Loomweave for six years, on a shared tablet three shifts pass between them at the end of each line. She can spot a weave flaw from four feet away, in a light that would make anyone else miss it entirely.

When LoomEye launched on her line, standard cotton, well-lit, easy to photograph clearly, it caught things she sometimes missed at the end of a long shift. She trusted its flags within a month. The board, one floor up, trusted the 94 percent even faster.

Only the line in the upper right ever earned the number that got quoted for all of them.

For two quarters, every board meeting opened the same way: "94 percent, on track." Nobody asked which lines that covered. Nobody asked because the number sounded finished, and a finished number doesn't invite questions.

Then a newly appointed board member, in her first meeting, looked at the slide and asked, simply: "Does that 94 percent apply to every line?" Nobody in the room had a clean answer. Fenn didn't either, not right away, and had to admit afterward that the number had only ever been measured on the two easiest lines.

The next narrative Fenn wrote opened differently: four hundred and ten thousand dollars in exposure, a hundred and twenty-six thousand of it gone, and two lines, dark synthetic and novelty print, named plainly as not there yet. Elin's line kept doing exactly what it had been doing. The narrative just finally said, out loud, what it couldn't say yet about the other four.

The first narrative only had one branch. Every line got treated as if it had already reached the leftmost one.

SPARK, the narrative built to survive its own bad newsNot a features list dressed up for a board. SPARK is what keeps one missed line from sinking the whole story.

The recap, one line per letter: situation is naming honestly that the board stopped asking questions after one confident number, payoff is teaching the board to track dollars removed instead of a percent, anchor is opening every narrative on the same real exposure figure, risk is designing the story so one missed line reads as a stalled sub-item, and keep out is refusing a blended percent or a plant-wide date before the evidence exists.

And if you want to be sure it really works, try it somewhere elseSame five letters, a regional pharmacy chain's prescription-review tool instead of a textile mill. A different old decision breaks the second narrative.

Harrow Vale Pharmacy Group runs a tool that flags likely prescription interactions for a pharmacist's final check. Its first board narrative reported "97 percent flag accuracy," a real figure measured against the chain's busiest, best-documented stores. Mapped onto SPARK: situation is the board treating that number as chain-wide fact. Payoff is getting the board to track dollars of avoided interaction incidents, not a flag-accuracy percent. Anchor is opening the narrative on the actual cost of interaction incidents last year, with each store cluster's contribution named separately. Risk is a rural cluster with thinner records missing its bar, which reads as one named cluster still exploring, not a broken chain-wide promise. Keep out is refusing to publish any store-level rollout date until two clusters independently clear the bar. The old decision here isn't a hidden number, it's a removed affordance: the first narrative dropped the per-cluster breakdown entirely to keep the slide to one page, on the reasoning that a single summary number was cleaner for a busy board to absorb.

The same four parts anchor a textile board slide and a pharmacy board slide alike.

Swap the trigger and it still runs.
Speed: an interviewer caps you at sixty seconds. Say "open on the real dollar cost, tie every item to how much of it it removes, and never let one blended percent stand in for the whole system," and stop.
Cost: there's no time to build a full per-line cost model before the next board meeting. Say so honestly, and start with the exposure number for the two hardest lines alone, since that's where the risk of a false claim is highest.
The model gets better, for real: if dark synthetic genuinely clears its bar next quarter, that's real news to report as its own line moving from piloting to proven, not folded quietly back into one blended number.

Where people run it wrong.
They let one confident percent stand in for a system with very different performance across cases.
They promise a plant-wide date the moment one line looks good, instead of waiting for a second line to confirm the pattern.
They cut the per-line breakdown to keep the slide short, and lose the one thing that would have caught the gap early.

How to use it live. The moment someone asks you to write a board narrative, ask yourself: if I only had one number to defend under hard questions, which one would actually hold up. Build the whole page around that one.

Every point on this line is honest about what it doesn't yet know, which is exactly why none of them can break trust the way the first slide did.

Flashcards (tap any card to flip it)

Check yourself Score: 0 / 0