Why Can't You Just Prompt Better?

The question we hear most often: “Can’t you just give agents better prompts?”

Short answer: No.

Longer answer: Prompting is necessary but not sufficient for coordination.

Here’s why:

The prompt fallacy: You can tell an agent “consider other stakeholders” but it doesn’t know what those stakeholders care about unless you spell it out. Every time. For every stakeholder. Updated constantly as things change.

The context window limit: Even if you could spell out every stakeholder’s perspective in every prompt, you’d blow through context limits on any non-trivial organization.

The stale information problem: Prompts are written at a point in time. Stakeholder concerns change. Relationships evolve. A prompt written last week might be dangerously outdated today.

The coordination problem: Prompting helps individual agents make better decisions. It doesn’t help multiple agents coordinate their decisions with each other.

Standpoint isn’t a replacement for good prompting. It’s infrastructure that makes good prompting possible at scale.

The question isn’t “prompting vs infrastructure.” It’s “how do we give prompts the context they need to actually work?”