The practice: Identify the 5-10 tasks you use AI for most frequently and build a saved, reusable prompt template for each. Never write a new prompt from scratch for a recurring task.
Why this matters: Most reps start each AI session with a blank page and a vague request. The result is inconsistent output quality. Reps who build and save templates produce consistently better outputs because the engineering work is done once, not every time.
What to include in a reusable template:
• Role assignment
• Task description
• Your standard context (solution, ICP, tone preferences)
• Required output format
• Negative examples (what to avoid)
• Placeholder fields in [BRACKETS] for variables that change each time
Where to store them: Claude Projects (as the system prompt), a Notion database, a pinned Google Doc, or the Promptifi library.
The compounding benefit: Each time you improve a template based on a better output, that improvement applies to every future use. Reps who write prompts from scratch every time never build this advantage.