PMaker home
Search
中EN
System
Color theme
空格的键盘
空格的键盘空格的键盘产品经理 · AI Builder
Sixteen steps, each phase with a clear deliverableThink it throughUsers, problem, value, competitorsThe phase most often skippedDesign itFlows, screens, prototype, planSkipping review and heading to devBuild itDevelopment, integration, test, acceptanceRequirements keep moving until the last minuteRun itLaunch, rollout, metrics, retroLaunch is treated as the finish lineEvery step you skip, you pay for somewhere downstream

The full workflow runs in four phases. Each phase's deliverable is the input the next phase starts with.

The Product Workflow

From idea to launch—which steps you go through, and where each role comes in.

Getting an idea to launch is not one step. It's sixteen, and each one has a defined deliverable: if the previous step produces nothing, the next step can't start.

After reading, you should be able to answer:

  • What actually happens between an idea and a launch
  • What each step has to hand off so the next one can begin
  • Which steps AI truly compresses, and which ones it doesn't touch

This workflow was forged by large-team coordination, not ceremony. Its value is that it pins down the deliverables: no handoff, no start. When you build alone, the headcount drops but the deliverables don't—they just move from documents into judgment calls in your head.

The four phases

The full workflow runs in four phases: think it through, design it, build it, and run it. The first phase is the easiest to skip, because its output is words, not screens—invisible to bosses and unexciting to teammates. But it's the most expensive to skip: if the direction is wrong, the faster you move through the next three phases, the more you lose. Its deliverable is a one-sentence value proposition and a requirements note with clear boundaries—especially what's explicitly out of scope, which settles half the arguments later. The third phase hands off a tested, launchable build, with development, integration, testing, and acceptance each doing its part. The fourth hands off metrics and a retrospective—launch, gradual rollout, watching the numbers, and deciding what's next. Each phase's deliverable is the input the next one starts with: no requirements, no plan; no plan, no estimate; nothing accepted, nothing ships.

Which steps AI compresses

Re-rank the sixteen steps by how much AI can do for you, and you'll see where your time should go. The "production" steps—drawing, writing docs, writing code, writing tests—AI can almost fully take over. The "figure out what's right" steps haven't shrunk at all. This explains a common feeling: the faster you build, the more you throw away. Output speed went up tenfold; judgment speed didn't. You can now produce in a day what used to take a week, but you can't think more clearly in a day than you used to. What AI also compresses is output speed itself: prototypes, docs, code, and tests now generate an order of magnitude faster, so the real queue on a team is review and decision—who decides what's right. Judgment that can't keep up with production is why the harder you work, the more tired and the more anxious you feel. So the most valuable skill in the AI era isn't building fast—it's deciding what's worth building and knowing when to stop.

How one person runs it

Compress sixteen steps into five: a one-sentence statement of the problem, a diagram of the solution, a build that runs, a handful of real users, and one retrospective. Keep the deliverables; swap the form from documents into something you can feed straight to AI. Say you want to build a budgeting tool: first write "for young people who hate bookkeeping—log an expense in three seconds, auto-categorized." Then sketch a prototype with only three screens, ask AI to build the first version, hand it to five friends who actually overspend, and run them for a week. Finally, review which part nobody used. Five steps done beats a perfect three-month plan.