You're redesigning how customer escalations flow across your company's customer service, billing, and engineering teams. The redesign will touch about 40 standard operating procedure documents, change how three different ticketing tools integrate, and requires choosing between two competing approaches: a hub-and-spoke model where customer service routes everything, or a direct-handoff model where teams hand off to each other based on tags.
You're working in Claude Code with the team's process documentation.
Which approach should you take?
Why did you pick that answer? Two or three sentences. The act of articulating it is what builds the judgment — not the click that follows.
Plan mode is built for exactly this kind of work — large scope, multiple valid architectural approaches, cross-cutting changes that affect many documents. Diving into direct execution without exploring the current state and committing to an approach almost guarantees you'll discover something halfway through that requires re-doing earlier work. Plan first; execute once the plan is solid.
Detailed instructions presume you already know which approach is right — but the whole task involves choosing between two approaches, which requires exploration first.
"Letting the approach emerge" produces inconsistent procedures and forces later documents to work around early decisions made without full information. For cross-cutting changes, this is especially dangerous.
Skipping the design phase on a 40-document, multi-team change is how you end up redoing the work.