Method

Every aiMinds experiment follows the same minimum method.

1. Define the real task

Describe the work in operational terms. State the intended user, input, output, and success condition.

2. Establish a baseline

Record how the task is performed without the tested AI approach. If no baseline exists, state that clearly.

3. Define the division of work

Separate:

• human decisions and judgement;
• AI-generated or AI-executed work;
• deterministic automation;
• external sources and tools.

4. Define measures

Use measures appropriate to the task, such as:

• elapsed time and active human time;
• completeness or error rate;
• number of revisions;
• cost;
• user acceptance;
• repeatability.

Do not imply causal improvement when only a single observation exists.

5. Run and preserve evidence

Keep the prompts, configuration, source references, code, test output, and relevant intermediate artefacts that can be published safely.

6. Validate

Use deterministic checks where possible. A human remains accountable for the final conclusion. Record who or what validated the output.

7. Report honestly

Distinguish between:

• Measured: produced by a defined measurement.
• Observed: directly seen during the experiment.
• Reported: provided by another source or participant.
• Hypothesis: plausible but not demonstrated.
• Unknown: not established.

Include failures, uncertainty, and limitations.

8. Decide

Conclude with one of these actions:

• repeat;
• revise;
• scale;
• use a standard tool;
• keep the work human-led;
• stop.