Document A
The Spec Sheet
> for the machine · speed · precision
Wire your data sources straight into an AI drafting step.
Required
- Zapier / Make / n8n (free tiers ok)
- OpenAI / Anthropic API key
- Your data source (Sheets, Jira, CRM export…)
- 1
Pick one report, define inputs
Write down exactly what goes in: 3 metrics from Sheets, closed tickets from Jira, notes field. One source per automation trigger.
- 2
Build the flow
Trigger: schedule Fri 7 AM. Step 1: pull data. Step 2: send to LLM with a prompt like 'Draft a weekly update for my manager from this data. Confident tone, bullet points, flag anything below target.' Step 3: email/Slack it to you.
- 3
Add guardrails
Instruct the model to never invent numbers — only use provided data, mark gaps as [NEEDS INPUT]. Review the first 3 outputs closely.
- 4
Ship it
Once trust is built, deliver straight to stakeholders and keep yourself as CC'd reviewer.
EFFICIENCY NOTE: ~90 minutes setup, then ~100 minutes saved every week. Pays for itself in week one.
Document B
The Worksheet
— for your hands · craft · patience
No-code-free version: a template + a ritual that halves the effort today.
Materials
- One-page report template (structure never changes)
- Saved filters/views in each data source
- A recurring 15-min calendar block
- 1
Freeze the structure
Sections: Headline → Wins → Numbers table → Risks → Asks. Write section headers once, forever.
- 2
Create saved views
In each tool, save the exact filter ('my tickets closed this week'). Monday-you should never rebuild queries Friday-you needs.
- 3
Batch the collection
Friday 9:00–9:15, collect all numbers into the template in one sitting. No tab-switching during the week.
- 4
Keep a running wins log
One line per win, noted when it happens. Friday becomes assembly, not archaeology.
OLD-HAND TIP: The template IS the automation. Structure decided once removes hundreds of micro-decisions later.