As of 13 August 2026, AI can only partly write Python code for your business.

What to actually do

1. Hand it to a person

   The route this page recommends

   A person who owns the outcome does this end to end, worth it when the failure is dear.

2. Use a tool built for this

   Second choice

3. Do it yourself

   The distant third

   A chat interface, power-user skill, and roughly 1 hour until you can act on the result.

   How to actually do it

   1. Open a plain document and write the business outcome, users, inputs, outputs, rules, edge cases and what must never happen.
   2. Create harmless sample files or records that represent normal, empty, invalid, duplicate and boundary cases, with real customer data removed.
   3. Collect the relevant system documentation, Python environment details and deployment constraints, but keep passwords, API keys and personal data out of the chatbot.
   4. Paste the prompt with the completed task details and sample data into a chatbot, then save the proposed file tree, code, tests, assumptions and unresolved questions.
   5. Create the files in a separate development folder, install only the stated dependencies, run the generated tests and compare each output with your expected examples.
   6. Ask a technically experienced colleague to inspect authentication, permissions, input validation, data handling, logging, dependency risks, error recovery and deployment steps.
   7. Run the corrected code against copied or synthetic data, compare its results with a manually checked result, and only then arrange a controlled deployment with backups and a rollback plan.

   Prompt

   Act as a careful Python developer. Build a small, maintainable Python application for this business task:
   
   Task: [describe the workflow and the desired result]
   Users: [who will run it]
   Environment: [operating system, Python setup, hosting or deployment method if known]
   Inputs: [file formats, API details, example records or other safe sample data]
   Outputs: [files, reports, database changes or messages required]
   Rules: [business rules, edge cases and calculations]
   Existing systems: [systems it must connect to, using documentation or mock interfaces where details are missing]
   Constraints: [security, privacy, performance, access and maintenance requirements]
   
   Do not invent API endpoints, credentials, database fields or business rules. If information is missing, state the assumption and isolate it in one clearly marked configuration section. Never include real secrets or ask me to paste them. Use standard-library Python where practical and explain any dependency.
   
   Return:
   1. A short design and a file tree.
   2. Complete code for every file, with clear filenames.
   3. Automated tests covering normal cases, invalid input, empty input, duplicate input and the important edge cases.
   4. A requirements file or exact installation commands where needed.
   5. Safe configuration instructions using environment variables or placeholders.
   6. A runbook for local testing, backup, logging, failure recovery and deployment.
   7. A list of assumptions, unresolved questions and risks.
   
   Before writing the final code, restate the acceptance criteria as observable checks. Make errors explicit rather than silently ignoring them. Do not claim the code is production-ready. After the code, explain how I can test it with harmless sample data and what a technically experienced reviewer must check before it touches live data.

   Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.