Should AI be used?
Start with task suitability, consequences, and verifiability.
Task-first AI decisions
Use AI when it fits. Know when it does not.
Understand the task, choose the right approach, set safe boundaries, and verify the result.
The decision layer
Six questions before choosing an AI tool
Would ordinary software work?
Prefer deterministic rules when they solve the problem reliably.
What pattern fits?
Choose prompts, context, retrieval, tools, or bounded agents by need.
What data is safe?
Classify information before placing it in any AI environment.
How much autonomy?
Limit decisions and actions according to risk and reversibility.
What proves success?
Define evidence, acceptance criteria, review, and stop conditions.