Workflow Automation

Prompts That Build Automations.

5 prompts for designing automations, writing agent system prompts, building Make.com flows, and specifying n8n blueprints. For founders, ops managers, and no-code builders.

5 Copy-Ready Prompts

Click copy · Paste into any AI tool
Prompt 01
Full Automation Blueprint
Design a complete end-to-end automation for [TASK] that runs without human input once triggered. Include: (1) Trigger — what starts the automation and from which tool. (2) Steps in exact sequence — each step, the specific tool handling it, and the data passed. (3) Decision branches — every conditional and its conditions. (4) Error handling — what happens at each potential failure point. (5) Output — what is produced and where it goes. Format as a numbered flowchart in text I can share with a developer or build in Make.com.

Most automation attempts fail because they are not designed fully before building. This prompt forces every edge case and failure mode to be considered upfront.

Prompt 02
AI Agent System Prompt
Write a system prompt for an autonomous AI agent that specialises in [DOMAIN]. The agent's goal: [SPECIFIC MEASURABLE OUTCOME]. Instructions for the agent: work methodically and explain your reasoning before each action; if you hit a blocker, generate 3 alternative approaches and select the best one with justification; never make irreversible actions without confirming with the user; report progress at each milestone; format all outputs as [SPECIFY FORMAT]. Constraints: [LIST ANY CONSTRAINTS].

A poorly written system prompt produces an agent that hallucinates, gets stuck, or makes irreversible mistakes. This template enforces methodical reasoning, escalation paths, and output standards.

Prompt 03
Convert Manual Process to Automation Spec
I have this manual process I do [DAILY/WEEKLY]: [DESCRIBE EACH STEP IN DETAIL]. Redesign it as an 80%+ automated workflow. For each original step: (1) Can AI fully replace it — if yes, which tool exactly. (2) Can AI assist a human — how. (3) Must stay human — why. Then give me: the full automated flow in sequence, the Make.com or n8n modules I would need, estimated time savings per [week/month], and implementation order (what to automate first).

The best automation starts with a deep map of the current process before touching any tool. This prompt structures that discovery before any building begins.

Prompt 04
WhatsApp/Email AI Triage Agent
Design an AI agent that monitors [WhatsApp Business / Gmail] and processes every incoming message as follows: (1) Classify: query type (support/sales/complaint/spam) and urgency (high/medium/low). (2) For routine queries: draft a reply using this tone and these templates: [DESCRIBE]. (3) For anything requiring judgment: flag with a recommended response for human review. (4) Log everything to Google Sheets: sender, timestamp, classification, response sent, escalated (yes/no). Give me the complete Make.com scenario structure with every module.

An AI triage agent can handle 70-80% of incoming messages automatically while surfacing the 20% that genuinely need human attention — with a recommended response already drafted.

Prompt 05
Data Pipeline Specification
Write a technical specification for a data pipeline that: (1) Pulls [DATA TYPE] from [SOURCE TOOL/API] on [SCHEDULE]. (2) Transforms it by [DESCRIBE TRANSFORMATIONS: filter, enrich, normalise]. (3) Validates it — flag records that fail [DESCRIBE VALIDATION RULES]. (4) Loads clean data to [DESTINATION: Google Sheets / Airtable / database]. (5) Alerts me on [SLACK/EMAIL/WHATSAPP] if [ERROR CONDITIONS]. Include: the specific Make.com modules or n8n nodes, data schema at each stage, and error handling at each step.

Most data pipeline failures happen at the transformation or validation stage — steps that are routinely underspecified. This prompt forces those details before you build.