AI Prompts for ChatGPT for Automation

Help me identify the best automation opportunities in my work. Here is how I spend my time each week: [DESCRIBE YOUR TOP 10-15 RECURRING TASKS WITH APPROXIMATE TIME EACH TAKES]. Analyze each task for: (1) automation suitability (is it rule-based, does it always follow the same pattern, is the data structured), (2) time-to-automate estimate, (3) ROI (hours saved per week vs. hours to build), (4) risk level (what breaks if the automation fails). Rank the top 5 tasks I should automate first and suggest the right tool for each (Python script, Zapier/Make.com, Excel macro, shell script).

Help me document this process so it can be automated: [DESCRIBE THE PROCESS YOU DO MANUALLY, STEP BY STEP]. For each step, identify: (1) the input trigger (what kicks off this step), (2) the specific action performed, (3) where the data comes from and where it goes, (4) any decision points or conditional logic, (5) what could go wrong and how you handle it. Format as a structured process map that an automation tool or developer can use as a blueprint. Highlight the steps that are candidates for automation vs. steps that require human judgment.

I want to automate this workflow: [DESCRIBE THE WORKFLOW, WHAT TOOLS ARE INVOLVED: APPS, DATABASES, EMAIL]. I do not want to write code. Recommend the best no-code or low-code automation tool (Zapier, Make.com, n8n, Microsoft Power Automate) for this specific case. Compare the top 2-3 options for: ease of setup, cost at my volume, reliability, available integrations, and any limitations for my specific workflow. Also describe the trigger and action sequence I would build in the recommended tool, step by step.

Calculate the ROI for automating this task: Task description: [DESCRIBE WHAT YOU DO MANUALLY]. Frequency: [X TIMES PER DAY/WEEK/MONTH]. Time per manual run: [Y MINUTES]. My hourly cost (salary + overhead): $[Z]/hour. Time to build the automation: [ESTIMATE IN HOURS]. Ongoing maintenance estimate: [HOURS PER MONTH]. Calculate: annual time saved, annual cost saved, cost to build, time-to-ROI payback period, and 3-year NPV. Also note any risk costs I should factor in (what happens if it breaks).

Help me design the architecture for an automation system that: [DESCRIBE WHAT YOU WANT TO BUILD END-TO-END]. The system needs to: process [DESCRIBE DATA VOLUME], connect [LIST THE APPS AND SYSTEMS INVOLVED], run [ON SCHEDULE/TRIGGERED BY EVENT/CONTINUOUSLY], handle errors gracefully, and be maintainable by [DESCRIBE: JUST ME / A SMALL TEAM / NON-TECHNICAL OPERATORS]. Recommend: (1) the right architecture (scheduled script vs. event-driven vs. queue-based), (2) the right tools and why, (3) where the data should live, (4) how to handle failures and alerts, (5) how to test before going live.

Write a Python script to automate this email task: [DESCRIBE: SENDING A DAILY REPORT, READING AND FILTERING INCOMING EMAILS, AUTO-RESPONDING TO SPECIFIC SENDERS, EXTRACTING DATA FROM EMAIL ATTACHMENTS]. Use: imaplib/smtplib for basic email or the Gmail API for more advanced workflows. Include: (1) secure credential handling from environment variables, (2) error handling for connection failures, (3) logging of all emails sent or processed, (4) filters and conditions as described, (5) an HTML template for outgoing emails if applicable. Never hardcode passwords.

Write a Python script to automate this spreadsheet task: [DESCRIBE: CONSOLIDATE MULTIPLE EXCEL FILES, APPLY FORMULAS PROGRAMMATICALLY, GENERATE A FORMATTED REPORT FROM DATA, UPDATE A GOOGLE SHEET FROM AN API]. Use openpyxl for Excel or gspread for Google Sheets. Include: (1) reading and writing the specified data, (2) any formatting or formula logic required, (3) error handling for missing files or unexpected data formats, (4) a summary of changes made printed to console, (5) a backup step that copies the original before making changes. Show how to run this script on a schedule.

Write a Python script to automate this web browser task: [DESCRIBE: FILL OUT A FORM, SCRAPE DYNAMIC DATA, TAKE SCREENSHOTS, AUTOMATE A LOGIN AND DOWNLOAD FLOW]. Use Playwright (preferred) or Selenium. Include: (1) browser launch in headless mode, (2) stable selectors (prefer data-testid, ARIA roles, or text-based selectors over CSS class chains), (3) proper waits for dynamic content (wait for element, not arbitrary sleep), (4) error handling with screenshots on failure, (5) extraction and storage of any collected data, (6) cleanup and browser close. Note any login credentials should come from environment variables.

Write a Python Flask or FastAPI application to receive and process webhooks from [DESCRIBE THE SOURCE: STRIPE, GITHUB, SHOPIFY, SLACK, CUSTOM APP]. Include: (1) the POST endpoint at /webhook, (2) webhook signature verification using the source's secret key, (3) payload parsing and event type routing (different handling for different event types), (4) async processing using a background task or queue to avoid timing out the webhook response, (5) idempotency handling to avoid processing the same event twice, (6) logging every received event and its processing status. Show a sample test using curl.

Write a bash shell script to automate this system task: [DESCRIBE: ROTATING LOG FILES, BACKING UP A DIRECTORY, MONITORING DISK USAGE, BATCH PROCESSING FILES, ENVIRONMENT SETUP]. Include: (1) proper shebang line and set -euo pipefail for safety, (2) variable declarations at the top, (3) a usage/help function, (4) error handling that exits with a non-zero code on failure, (5) logging to a file with timestamps, (6) an email or webhook alert on failure, (7) idempotency where appropriate. Make the script safe to run multiple times and safe to interrupt mid-run.

Write a Python script to integrate [SOURCE TOOL/API] with [DESTINATION TOOL/API]. What I want to accomplish: [DESCRIBE THE SYNC OR TRANSFORMATION: E.G., PULL NEW LEADS FROM HUBSPOT AND CREATE TICKETS IN JIRA, SYNC STRIPE PAYMENTS TO A GOOGLE SHEET, PUSH DATA FROM POSTGRESQL TO SALESFORCE]. Include: (1) authentication for both APIs from environment variables, (2) incremental sync using updated_since timestamps or pagination, (3) field mapping from source to destination format, (4) duplicate detection before creating new records, (5) error handling with retry for transient failures, (6) a log of records synced.

Design a Zapier or Make.com workflow for this automation: [DESCRIBE THE TRIGGER, DATA TRANSFORMATIONS, AND DESIRED ACTIONS]. For each step, specify: (1) the trigger (what app and event starts the workflow), (2) any filter conditions, (3) any data transformation steps needed (formatter, code step, or data aggregation), (4) the actions in order with the exact app and action type, (5) any error handling steps. Also note: which steps are multi-step (requiring premium Zapier plan) and any limitations of this platform that might require a workaround.

Design a data synchronization architecture for keeping [SYSTEM A] and [SYSTEM B] in sync. These systems need to stay in sync because: [DESCRIBE BUSINESS REASON]. Key challenges: [DESCRIBE: DIFFERENT SCHEMAS, REAL-TIME VS. BATCH, CONFLICT RESOLUTION, VOLUME]. Recommend: (1) sync direction (one-way or bidirectional), (2) sync mechanism (polling, webhooks, CDC), (3) conflict resolution strategy, (4) how to handle records that fail to sync, (5) how to bootstrap the initial sync, (6) monitoring and alerting design. Include a data flow diagram description and a recommended technology stack.

Write a script to monitor [DESCRIBE WHAT TO MONITOR: A WEBSITE, A DATABASE METRIC, AN API STATUS, FILE CHANGES] and send alerts to [SLACK, EMAIL, TEAMS, PAGERDUTY] when conditions are met: [DESCRIBE THE ALERT CONDITIONS]. Include: (1) the check logic that runs on schedule, (2) threshold management (alert only if condition persists for N consecutive checks, not on every fluctuation), (3) alert deduplication (do not spam repeat alerts for the same ongoing issue), (4) resolution notification when the condition clears, (5) structured alert message format with severity, what triggered it, and a link to investigate.

Write a Python ETL script to extract data from [SOURCE DATABASE/API], transform it, and load it into [DESTINATION: GOOGLE SHEETS, A BI TOOL, ANOTHER DATABASE, A FLAT FILE]. Schedule this to run [DAILY/HOURLY]. Include: (1) extraction query or API call with incremental loading strategy, (2) transformation logic: [DESCRIBE ANY CLEANING, AGGREGATION, JOINS, OR CALCULATIONS NEEDED], (3) loading with upsert logic to avoid duplicates, (4) a simple data quality check that validates row counts and key metrics before loading, (5) performance optimization for large datasets, (6) run time and row count logging.

Write a Python script that uses the OpenAI or Anthropic API to automatically classify incoming documents or emails into categories: [LIST YOUR CATEGORIES AND WHAT MAKES EACH DISTINCT]. The system should: (1) process each document through an LLM with a structured classification prompt, (2) return a category and a confidence reasoning, (3) route low-confidence results to a human review queue, (4) log all classifications for accuracy tracking, (5) be cost-efficient (truncate very long documents, cache repeated classifications). Include the prompt template for classification and show how to evaluate accuracy against a labeled test set.

Write a Python script to extract structured data from unstructured text using an LLM. The text sources are: [DESCRIBE: EMAILS, PDFS, WEB PAGES, DOCUMENTS]. I need to extract these fields: [LIST FIELDS]. The script should: (1) send each document to the LLM with a JSON extraction prompt, (2) parse and validate the returned JSON against an expected schema, (3) handle extraction failures and missing fields gracefully, (4) output results to CSV or a database, (5) track extraction accuracy and flag records that need manual review. Include the system prompt and show the JSON schema I should request.

Design and implement an AI-powered email response system for [DESCRIBE THE TYPE OF EMAILS: CUSTOMER SUPPORT, SALES INQUIRIES, HR QUESTIONS]. The system should: (1) receive incoming emails via IMAP or webhook, (2) classify the email type and urgency, (3) generate a draft response using an LLM with appropriate context and tone guidelines, (4) for routine queries: auto-send if confidence is high enough, (5) for complex queries: route to human with the AI draft pre-populated, (6) log all auto-sent responses for review. Include the prompt template and the decision logic for auto-send vs. human review.

Write a Python automation that processes meeting transcripts or notes and produces structured outputs. Input: [DESCRIBE INPUT FORMAT: AUDIO FILE, TXT FILE, PASTE FROM ZOOM/TEAMS TRANSCRIPT]. The automation should: (1) if audio: transcribe using OpenAI Whisper or another service, (2) send transcript to an LLM to extract: a 3-5 bullet summary, action items with owner and due date, decisions made, and open questions, (3) format the output as a clean document, (4) optionally create Notion/Confluence pages or send the summary to Slack/email. Include the extraction prompt and show sample output for a mock meeting.

Build an AI-powered quality control checker for [DESCRIBE WHAT YOU ARE CHECKING: CUSTOMER SERVICE RESPONSES BEFORE SENDING, CONTENT BEFORE PUBLISHING, DATA BEFORE LOADING, CODE BEFORE REVIEW]. The checker should: (1) evaluate each item against a rubric you define: [DESCRIBE YOUR QUALITY CRITERIA], (2) return a pass/fail decision with specific feedback on what failed, (3) route failed items to a revision queue, (4) track pass rates over time to identify systematic quality issues, (5) be calibrated so human reviewers validate a sample of AI decisions weekly. Include the evaluation prompt and the rubric structure.