Why Does Planning Become More Important as AI Gets Faster?
AI increases execution speed. But without deciding what to build, why to build it, and how far to take it, rapid execution can become rapid drift.
AI has become an era where it can make almost anything quickly. Code comes out, screens come out, writing comes out, feature descriptions come out. So there's a temptation to skip planning and just start building. If planning takes a long time, wouldn't it make sense to reduce that stage with fast AI? the thought goes.
But right here, a question arises.
If AI makes things quickly, does planning become less important?
If you tell AI "make this kind of service," screen structures appear. If you say "implement this feature," code comes out. If you say "write the copy for this page," sentences appear.
The speed at which results are produced has certainly increased.
But AI making things quickly doesn't mean moving quickly in the right direction.
AI increases the speed of execution. But it doesn't decide the direction of that execution for you.
What happens when you use AI without a clear direction?
Results come out. Screens come out, features come out, code comes out. But why those results are needed, who needs them, and how much should be built remains unclear.
The faster AI builds, the faster it builds in the wrong direction too.
Using AI without planning isn't fast execution—it can be fast drifting.
Anyone who has experienced screens constantly changing, features continuously being added, and revisions repeating endlessly won't find this unfamiliar. Even if the quickly made results look good, if the initial question is unclear, revisions just repeat quickly.
Without criteria before building, it's difficult to judge AI's output.
"Is this the right direction?" "Do we really need this feature?" "Is this screen what was originally intended?" To answer these questions, you need a direction set from the beginning.
Without direction, what AI produces becomes the criteria. Looking at AI's output, you establish direction. And then you end up building not what you wanted to build, but what's easy for AI to build.
There's one more thing here.
It's not a problem that AI can't make something. It can be a problem that AI makes something too convincingly.
Suppose AI creates output different from the original intention in a state where planning isn't solid. If that output looks poor, it can be discarded immediately. But if it looks good, the sentences are smooth, and the screens look plausible, people hesitate.
"It's different from what I originally thought, but doesn't this look better?"
From this moment on, direction stops being the plan set by people and starts following the output created by AI. The problem isn't changing direction itself. It's changing direction without judgment, drawn along by plausibility.
Planning is not a document for rejecting AI's output. It's a criterion that lets you compare that output against the original purpose, even when the output looks good.
This is the danger of using AI without planning. It's not that results don't come out, but that results in unintended directions accumulate quickly.
When you hear "planning," you might think of long planning documents, detailed requirement documents, tens of pages of project plans. But planning in the AI era is different.
Planning in the AI era is the work of people setting direction and criteria before AI executes.
You can start by establishing at least four things.
- What will you build?
- Why will you build it?
- Who will you build it for?
- How much will you build?
If you tell AI to build with these four things unclear, AI will build. But there's no criterion to judge whether it's right.
Planning isn't a procedure that slows down AI's speed. It's a device that determines where that speed should head.
Before AI, execution was slow. It took time to make screens, time to write code. In that time, direction could be reset.
With AI, execution is fast. The speed of going in the wrong direction is equally fast.
As execution power becomes commonplace, the difference in ability to judge what to build becomes more pronounced. The work people need to do before telling AI to build has not decreased—it has become clearer.
Did I fully determine what to build and why I need to build it before telling AI to build?