Chapter 7 Paths in, paths out
You did the exercises in Chapter 6. You felt something.
If it was mostly dread, skip to Section 7.2. No judgment. Better to know now than to spend two years finding out.
If it was mostly energy — even anxious energy — this chapter is for you first. Wanting to be a PM and becoming one are different problems. AI changed both.
7.1 Paths in
There is no canonical path. There never was. But AI broke the apprenticeship model, so the traditional entry routes are worse and some new ones actually work now.
7.1.1 1. APM: The classic, compressed
Before: Associate PM programs at Google, Meta, Microsoft, Uber. Two-year rotation. You wrote specs, shadowed Senior PMs, did the low-risk work that taught you the patterns.
After: AI does the low-risk work. APM programs know this. They’re hiring fewer APMs, they’re hiring them later, and they’re expecting judgment faster. The programs that do exist are explicitly for people who have already shipped something — not something impressive, just something real that real people used (Rachitsky 2023).
How to get in, 2026:
Ship something. The bar is low: a Chrome extension, a Notion template with a hundred users, a Discord bot, a weekend project that solved a problem you had. AI makes building faster, which means “I built this” is table stakes now, not a differentiator. But “I built this, watched users use it, and changed it based on what I learned” is still meaningful. Do that.
Write a teardown. One page. Pick a product with a real flaw, diagnose the problem specifically, propose a solution with tradeoffs named. Send it to someone who works there. Unsolicited. If it’s good, it gets forwarded. If it’s wrong, you learn something.
Get a referral from someone whose referral means something. Not “she’s smart.” “She sent me this doc and it changed how I thought about our retention problem.”
Startup APM: No program. No rotation. You become APM by being the engineer or designer who writes the first PRD and lives to tell about it.
Scale-up APM: Programs exist but are half the size they were in 2021. They hire ex-engineers and ex-designers who “want to go to the product side,” because those people already have the technical and craft context the program used to teach.
Mega-corp APM: Still exists. The bar is: you could have been a founder. Because the tools to be a founder are free, and if you chose not to, they want to know why.
Test: Can you get ten strangers to use something you made? Not your friends. Strangers. If no, fix that before you apply anywhere.
7.1.2 2. The lateral: From another function
Most PMs didn’t start as PMs. They started as something else and got tired of waiting for the PM to decide.
From Engineering
You’re tired of building the wrong thing. You already understand the technical constraints better than your PM. You’ve started writing the spec before anyone asked.
AI impact: You can now prototype the right thing yourself, before the PM conversation happens. Use that. Your interview is not a résumé. It’s the PRD and prototype you built for the problem you believed in and the PM didn’t.
The risk: You’ll miss the flow state of building. PM is not building. It’s deciding what to build, then watching other people build it, then explaining why it shipped the way it did. If you loved being in the code, stay there. PM is not a promotion. It is a different job with worse hours and more blame.
From Design
You’re tired of being handed requirements that are wrong. You want to set them.
AI impact: You can generate ten flows before the PM finishes the brief. That speed is your credential. Your portfolio isn’t mocks. It’s “here is the decision I made, here are the nine options I killed, and here is what the user data said afterward.”
The risk: You’ll fight with your former peers. Designers will tell you you’re “not really a designer anymore.” They’re right. Decide if you care.
From Data
You’re tired of being asked “what happened” when you want to work on “what should we do.”
AI impact: Everyone can pull data now. Your edge is knowing when the data is wrong. “DAU is up because we broke the event tracking on mobile” is a PM answer. “DAU is up 3%” is a dashboard. The ability to tell the difference — and to say so out loud when the number looks clean but smells wrong — is a compounding advantage.
The risk: You’ll miss certainty. PM is not certain. You decide with 60% of the information you want and act with 90% of the conviction that implies. If you need to be right, stay in data. You’ll be more right, more often, with better justification.
From GTM: Sales, CS, Marketing
You’re tired of selling a roadmap you don’t believe in. You’ve spent a year on calls with customers telling you exactly what’s broken. You want to fix it.
AI impact: You can now prototype what Sales has been asking for in an afternoon. That is your interview. Build the thing. If it works, you’re a PM. If it fails, you’ll have a more specific and honest account of why Sales was asking for the wrong solution to the right problem — which is also useful.
The risk: You lose your quota, your commission, your weekly clarity of “did I make my number.” PM has no quota. The feedback loop is quarterly at best. The thing you shipped in March might be what drove the December metric. Learning to work without short-cycle feedback is a real adjustment.
GTM also often underestimates how much of PM work is internal. Selling to customers is easier than selling to engineers who’ve heard this request three times before and know why it’s harder than it looks.
7.1.3 3. The founder path
You started something. It didn’t work. Now you want to be a PM.
AI impact: Every PM hiring manager at a scale-up has a founder in the interview queue now. Tools to start a company are free. Finishing a company, learning from it, and articulating what you’d do differently — that’s still rare.
“Failed founder” is not a credential. “Failed founder who can tell you specifically what was wrong with the market timing, the distribution assumption, and the retention model” is.
Write the post-mortem. One page. What you built, what you assumed, what the users told you, what you’d change with more resources, and — the most important one — what you’d change with fewer resources. If you can write that honestly, send it with your application.
Startup: They’ll hire you. They need the scar tissue. Scale-up: They’ll interview you. They’re worried you’ll leave to try again. Mega-corp: They’ll level you at L4 and tell you it’s because of “calibration.” Take it if the learning is worth the hit, or don’t. That’s a fair choice.
7.1.4 4. The AI-native PM path
This role didn’t exist in 2022. It exists now and it’s real.
You’re not applying to “PM” jobs. You’re applying to “PM, AI” — where the product is an agent, an orchestration layer, a model-powered workflow. You’re not adding AI to a product. The product is the AI behavior.
How to get in: Ship an agent that a hundred people use. Not a demo. A tool. “This saves me two hours a week” from a stranger who found it without your help is your résumé.
Example: A PM got hired at an AI company because he built a browser extension that summarized Reddit threads using GPT-3. Twenty thousand users, no marketing, no funding. He didn’t have a PM title. He had users who came back.
The risk: The path is crowded now. It will be more crowded in six months. What matters is not having the title but having the shipped thing with users who cared.
7.2 Paths out
You did the exercises. You felt dread in most of them. Good. That’s the most useful thing this chapter can give you.
The exercises in Chapter 6 are not PM-specific tests. They’re tests of the underlying disposition. The inbox exercise tests tolerance for reactive, multi-priority work. The no exercise tests tolerance for holding positions under pressure. The room exercise tests comfort with named conflict. Each of those points toward specific adjacent roles where that disposition is less load-bearing.
7.2.1 If you liked the story exercise but hated the room
You like narrative. You find the communication of a clear idea energizing. You do not like conflict navigation — or it drains you faster than it should.
Product Marketing: You write the launch narrative after the PM sets the strategy. You own the story from the announcement forward. Less blame, less internal conflict, more external performance. The gap is that the story you’re telling was shaped by someone else — the tradeoff is real.
Content and editorial: You find the signal and make it legible at scale. Increasingly powerful as AI floods the zone with low-quality content and the premium on genuine clarity rises.
7.2.2 If you liked the call exercise but hated the inbox
You like deciding. You don’t like the reactive, always-on nature of the PM role.
BizOps and Strategy: You build the analysis that informs the decision. You’re often right. Someone else absorbs the execution risk. The gap is distance from impact — the decision you made lives in a deck someone else presented.
Investment (VC, PE, growth equity): You make decisions all day with incomplete information and someone else executes. The feedback loop is years, not weeks. If you liked “the call” exercise because you’re comfortable with long-horizon uncertainty, this is worth exploring.
7.2.3 If you liked the inbox exercise but hated the no
You like action and triage. You struggle to hold a position when someone pushes back.
Chief of Staff: You run the operating system. You translate between functions, you own the process, you make the wheels turn. The decisions still get made by someone else. That can be the right tradeoff.
Product Operations: You own the system, not the strategy. As AI makes product teams faster, the gap between what they ship and what works operationally is widening. Someone has to own that gap. It’s often not a traditional PM role.
7.2.4 If you liked the prototype and data exercises but hated the stakeholder ones
You like building and analyzing. You don’t like the politics.
Design Engineer: You code, you design, you ship product without being accountable for the strategy. At companies like Linear, Vercel, and Figma, these are the people who actually make the product feel right. AI makes you more powerful here, not less.
Founding Engineer: You do the early startup version of PM work — you decide what to build because you’re building it — without the meeting overhead that accumulates as the company grows. Get in early, shape the thing, and leave before you’re three layers away from the code.
7.2.5 If you hated most of them
Leave tech. Seriously.
Tech is not a higher calling. It is a domain. If the domain’s work — the product decisions, the stakeholder management, the metrics, the constant uncertainty — doesn’t pull you, there is no shame in that and no version of persistence that fixes it.
The Stanford MBA who did three PM internships, hated all of them, kept applying, got the Meta job, and quit after eight months to run a bookstore — he says it’s the first job he liked. The PM job wasn’t the prize. He was wrong about what he wanted. Eight months at Meta was the cost of finding out.
Some people find out in six exercises. That’s the cheaper option.
7.3 The half-life problem
Whatever path you pick, it decays.
2015: You could be a PM for a decade. The job evolved slowly. The skills you built in year one were relevant in year ten.
2026: The job changes meaningfully every eighteen to twenty-four months. The agent PM role that exists now will look different in two years. The skills being tested in APM interviews this fall were not the skills being tested two years ago.
This doesn’t mean the path is wrong. It means the path requires maintenance.
Ship something every six months. Doesn’t have to be your job. GitHub, a Substack, a Discord community tool. If you haven’t shipped in six months, you’re accumulating distance from the thing that PM judgment is built on.
Have opinions in public. “I think this product is broken because X” is worth more than “AI is interesting.” Be wrong sometimes. Update when you are.
Build a small network of PMs you can call when you’re uncertain whether your instinct is good or broken. AI cannot calibrate you. Another PM who shipped something similar last quarter can.
7.4 Try this
Exercise: Based on your results from Chapter 6, pick one path in or one path out.
Write three specific actions you will take in the next thirty days to start. Not “learn more about PM.” Specific. “Ship a Chrome extension by the fifteenth.” “Send a product teardown to someone at the company by Friday.” “Have a thirty-minute call with a PM who transitioned from engineering.”
If you can’t write three, you don’t want it yet. If you wrote ten, pick three and do them. The others are procrastination with better branding.
Chapter 8 is for both groups. If you’re in: how to stay relevant. If you’re out: how to leave cleanly and use what you learned.