Step Decision Tree Card
A one-page companion to AI Helped Me Build It. AI Does Not Have to Run It.. Print it and keep it at the desk.
Ask these questions of each step in a process, not once for the whole process.
Before you start
Is this step still necessary? Automation does not cure a process that exists only because “we’ve always done it this way.” Automating a process nobody understands is just an efficient way to produce confusion on schedule.
Choose the starting approach
| # |
Question |
If yes |
If no |
| 1 |
Is the step necessary and understood? |
Go to 2 |
Simplify, document, or redesign first |
| 2 |
Can we write the rule, and should the same input always give the same result? |
Start with deterministic automation (Excel, Power Query, SQL, Python) |
Go to 3 |
| 3 |
Does it require interpreting language or context? |
Consider an LLM, subject to data restrictions, validation, and review |
Go to 4 |
| 4 |
Do reliable historical examples hold a pattern we can’t write as rules? |
Consider machine learning |
Keep it human-led, or redesign |
Then, for every option you chose
| # |
Control question |
What it means |
| 5 |
Does the step involve judgment, approval, restricted data, or contractual limitations? |
If yes, a qualified person stays responsible, whatever technology prepared the work. |
| 6 |
What evidence will show that it worked as intended? |
Define validation, logging, exception handling, and sign-off before you build. |
The four kinds of work
| Kind |
The test |
Typical home |
| Rules |
Can we state exactly how the result is calculated or classified? |
Formula, Power Query, SQL, Python, RPA |
| Patterns |
Do past examples hold a relationship we can’t write down? |
Machine-learning model trained on your history |
| Language |
Does it require reading or writing unstructured text? |
LLM, with validation and review |
| Judgment |
Does it need accountability, authorization, or a facts-and-circumstances conclusion? |
A qualified person |
The rule of thumb
- If the rule is known, encode the rule.
- If language must be interpreted, consider an LLM, and validate it.
- If historical examples hold a pattern you can’t write down, machine learning may fit.
- If the decision needs professional accountability, keep a person responsible.
| *PythonMuse LLC |
www.pythonmuse.com. Educational purposes only; not professional advice. Content licensed under CC BY-NC-SA 4.0.* |