Reducing Token Waste in Documentation Workflows

The problem

A consulting agency used LLMs to generate internal architecture documentation from source code. The prompts were written ad-hoc by engineers and produced long, verbose responses that required follow-up prompts to refine.

This increased API usage and slowed down documentation generation.

The solution

promptctl was used to generate structured prompts automatically.

• Deterministic prompt structuring engine
• Enforced clear sections, constraints, and output formats
• Cost estimation across models before sending requests

The result

"Structured prompts reduced the number of follow-up calls dramatically."

— Lead engineer, software consultancy