What does a hiring manager actually open first in an AI PM portfolio?

Tunde Bakare spent nine years as a maintenance planner at a sheet-metal stamping plant before he started learning to build AI products. His sample project is modeled on Kestrel Sheet Metal, the kind of plant he used to keep running: a tool that triages predictive-maintenance alerts so a small crew isn't drowning in false alarms. Helena Marsh is the Director of Product doing the hiring, and she gets through about forty portfolios a week.

How to answer this, stage by stage

You're not being graded on whether you know every rule of portfolio design. You're being graded on whether you can name the one decision that decides if anyone reads past your homepage.

Let's learn

Say a hiring manager sits down with a stack of AI PM portfolios and forty minutes before her next meeting.

Before Tunde changed anything, his portfolio homepage opened with three paragraphs about his nine years in manufacturing, then a "select a project" menu with five roughly equal options underneath. Helena spent about ninety seconds on the page and moved on, same as she does with most of them.

After the change, the homepage opens on one line about the alert-triage tool, then a single button: try it. Helena clicked it, ran three sample alerts through it, and read the short write-up underneath only once the tool itself had already earned her trust.

At its worst, before the fix: Helena closes the tab having read three paragraphs of biography and touched exactly zero of Tunde's actual work, and moves to the next portfolio in the stack thinking, fairly, that there was nothing to look at.

What I would leave alone: the full write-up, with the real numbers and the failure cases, still belongs on the page. It just doesn't belong first. Someone who does want the depth still needs it sitting one click deeper, not cut.

The lesson: a portfolio's order isn't a style choice you make at the end. It's the decision that decides whether the good stuff you spent three weeks on ever gets seen at all.

Now here is the same thing as a story

The short version above is what you'd say defending this in an interview. Read this one for how Tunde actually found the fix.

The Kestrel Sheet Metal floor runs a night shift that starts at 11pm, and for nine years Tunde was the person a foreman called when a stamping press threw an alert nobody trusted yet.

When he switched into product work, he spent three good months building his first portfolio: a homepage, an about-me section written carefully over several drafts, and five case studies, each one polished, each one about the same length.

Five options at the top of the page didn't give Helena five chances. It gave her one reason to open none of them.

The habit thinned out fast, without him noticing: first he assumed reviewers would read the about-me section fully, then he hoped they'd at least skim it, and by the third week he'd stopped checking whether anyone got past it at all.

Then a mentor, watching over his shoulder, opened the page cold for the first time and gave up after ninety seconds without clicking a single project.

Ninety seconds isn't a lot of time. It's also the entire budget most portfolios get.

Tunde spent the whole next weekend rebuilding, and it cost him something real: he cut two of his five case studies entirely, including one he personally liked best, because it didn't fit behind the new flagship link.

He'd been designing for step four. Helena's actual reading order runs one, two, three, four, in that order, every time.

He rebuilt the homepage around a single line: "An AI tool that tells a maintenance crew which alerts to trust." One button underneath: try it. The five case studies became one flagship, with the rest moved to a second page for anyone who wanted more.

Everything a reviewer needs from a first pass fits on one page now, four labeled parts, nothing else competing for the click.

The old homepage asked a stranger to read her way to the proof. The new one hands her the proof first and lets the reading be optional.

I wrote that about-me paragraph first because it felt honest, and because I genuinely believed the story mattered more than the tool. Watching my mentor close the tab at ninety seconds is what taught me that the story only gets read by people who already believed the tool was worth their time.

ORDER, ranked for a portfolio's front pageNot by what's biggest. By what unlocks the next ninety seconds.

The recap, one line per letter: outcome is getting read past ninety seconds, reversibility is the first-click that rarely gets a second try, dependency is headline unlocking project unlocking write-up, evidence is one cold read from a real person, and rank is one flagship project, first, defended.

And if you want to be sure it really works, try it somewhere elseSame five letters, a transit-maintenance dispatcher instead of a factory floor. A different reviewer, and this time the fix isn't about clicking at all.

Wren Castillo is building a portfolio project around dispatch triage for a regional transit agency's bus fleet, flagging which reported mechanical issues need a bus pulled before its next route. Applied to ORDER: outcome is getting a technical hiring panel to trust the model's judgment calls, not just its accuracy number. Reversibility is that a panel who catches one unexamined false negative in the first five minutes rarely trusts the rest of the write-up, even if it's sound. Dependency is that the panel needs to see the failure cases before the success rate means anything to them. Evidence is running the write-up past one skeptical engineer before the interview, not after. Rank: Wren leads with the three times the model was wrong and what happened next, ahead of the overall accuracy number.

For Wren's technical panel, the click leads somewhere different: straight to the failure cases, not the demo button.

The flagship project is the only piece that's both fast to open and heavy with proof. Everything else trades one for the other.

Swap the trigger and it still runs.
Speed: an interviewer gives you thirty seconds to describe your portfolio's structure. Say "one project first, everything else behind it," and stop there.
Cost: you only have time to finish one project before applying season starts. Ship that one well rather than three half-finished ones sitting behind it.
The market gets better, for real: even if AI PM roles get less competitive and reviewers get more patient, the ordering still holds, a stronger job market doesn't reward gambling your best work on being paragraph three.

Where people run it wrong.
They lead with credentials or a career-change story, assuming the reader wants context before evidence.
They spread five average projects across the front page instead of committing to one strong one.
They bury the one clickable thing behind a written introduction long enough that most readers never reach it.

How to use it live. When you're stuck describing your own portfolio's structure, ask yourself one question: if someone gave this page ninety seconds, what's the one thing you'd want them to touch. Build the order around that answer, not around what felt honest to write first.

Flashcards (tap any card to flip it)

Check yourself Score: 0 / 0