Workflows
Use AI as a process, not a one-off prompt.
Each workflow shows where AI helps, where human verification belongs, and which prompt or tool fits each step.
QUICK START
Pick the outcome. Follow the path.
No workflow terminology required.
Popular
Start with the outcome
Research → sourced briefDiscover sources, verify claims, synthesize disagreements and write a decision-ready brief.Best fit: Perplexity + NotebookLM
Meeting notes → actionsSeparate decisions, actions, owners and unresolved questions without guessing.Best fit: Otter + ChatGPT / Claude
Raw material → presentationBuild the argument, evidence and visual plan before formatting slides.Best fit: ChatGPT / Claude + Gamma
Idea → stronger AI imageDefine purpose, composition, lighting and constraints, then iterate deliberately.Best fit: ChatGPT Images / Midjourney
Bug → verified fixObserve, form hypotheses, diagnose, change code and prove the result.Best fit: Cursor / Claude Code
Requirement → test coverageClarify ambiguity, map risk, design cases and automate stable behavior.Best fit: ChatGPT / Claude + coding assistant
AI-generated PR → production-risk auditReview silent failures, concurrency traps, rollback gaps, query explosions and scale assumptions.Best fit: Cursor / Claude Code + runtime evidence
Scattered project context → durable AI rulesTurn PRDs, architecture notes and decisions into portable, versioned context.Best fit: repository context + tool-specific instruction files
Need a workflow for your own job?Generate a step-by-step AI process.
Open Workflow GeneratorDescribe the outcome, inputs and risk level. We’ll map AI steps, tool roles and human verification points.
Pattern
A good AI workflow has four parts
01Human intent
Define the outcome, risk and evidence before asking AI to act.
02AI leverage
Use AI for expansion, synthesis, drafts or repetitive transformation.
03Verification
Check consequential facts, assumptions, changes and final outputs.