AI Can Build Steps. It Doesn't Model Momentum.

AI is very good at generating steps.

It can scaffold onboarding flows, multi-step setups, and configuration sequences instantly.

Inputs render.
Validation works.
Data saves.

Technically correct.

But flows aren't about correctness.

They're about momentum.

The Dependency Problem

AI often misorders steps.

It asks for advanced configuration before basic identity.

It surfaces downstream options before upstream choices are made.

Order matters.

If a decision requires context, that context must come first.

When it doesn't, users hesitate.

Hesitation breaks momentum.

Step Chunking

AI tends to overload a single step with too many inputs
or create too many thin steps with no psychological progress.

Great onboarding balances cognitive weight.

Each step should reduce uncertainty and reinforce progress.

Momentum lives in pacing.

The Modal Trap

AI frequently defaults to small modals for structured tasks.

Technically functional.

Spatially wrong.

Some flows deserve full-screen focus.

Users need spatial cues:

Now I'm configuring.
Now I'm reviewing.
Now I'm committing.

Modal misuse flattens those cues.

AI can build the structure.

It does not instinctively design the experience of movement.

Momentum is designed.

It is not assembled.