16 business prompts
Business automation and AI evaluation
Design repeatable work and evaluate AI.
A question to explore
How can a business operation be repeated and checked?
A practical way to begin
Start with a question you can check.
Define inputs, outputs, permissions and failure conditions before implementing an Action. Evaluate AI against representative labeled examples and retain human approval for consequential steps.
The collection
Find your next question.
Analysis, interactive Apps and repeatable Actions. Choose the outcome first.
Business automation and AI evaluation
- Draft a recurring business review ActionDraft recurring review ActionAction
- Prepare a dependency-aware Model refresh ActionDraft dependency-aware Model refresh ActionAction
- Draft a validated webhook intake ActionDraft webhook intake ActionAction
- Prepare resumable API ingestionDraft resumable ingestion ActionAction
- Evaluate AI classification on labeled casesClassification evaluation with per-class errorsAnalysis
- Score AI extraction against reference fieldsField-level extraction evaluationAnalysis
- Compare prompt and model alternatives fairlyPrompt and model comparison reportAnalysis
- Choose an AI abstention thresholdAbstention threshold trade-off analysisAnalysis
- Investigate AI input and output driftInput, output and error drift investigationAnalysis
- Design Action retry and idempotency checksRetry and idempotency test designAnalysis
- Review an Action release capability contractRevision-specific capability contract reviewAnalysis
- Check composed Action failure propagationParent-child failure propagation assessmentAnalysis
- Evaluate sensitive-data redaction qualityRedaction precision and missed-span evaluationAnalysis
- Build a human review queue for AI outputsAI output review queue AppApp
- Draft a recurring AI quality evaluation ActionDraft recurring AI evaluation ActionAction
- Create a bounded cloud spend alert ActionDraft alert Action and rule-review checklistAction
From a question to useful work
Start with the data.
Build from what you find.
Choose a prompt, bring permitted inputs, and review the result. A useful analysis can become an App for your next review or an Action with configured repeat execution.