A maintained library of practical guides, official sources and Indexify workflows, organised around the work you need to finish. Every record shows what was checked, what it helps you do and where its limits begin.
01 · Find
Search by the job you need to complete.
02 · Check
See the source, review date and caveats.
03 · Test
Use the resource on representative work.
04 · Keep
Save a short list in this browser.
Catalogue
Search the outcome, or narrow the list by practical pathway and resource type.
16 resources shown
An original worksheet for mapping inputs, outputs, human checks and realistic ROI assumptions.
Useful outcome
Frame one marketing workflow and define how you will test it.
A short, model-agnostic introduction to finding work where AI can make a useful and measurable difference.
Useful outcome
Identify one worthwhile workflow and describe the result before choosing a tool.
A structured route from prompting and model choice to applications, marketing workflows, video and agents.
Useful outcome
Choose a sensible starting course and build a practical outcome at each stage.
The Find, Frame, Test and Embed method for moving from an attractive AI idea to a controlled working practice.
Useful outcome
Frame an AI opportunity, define controls and decide what evidence would justify adoption.
A browser-based calculator for testing the time, cost and break-even assumptions behind a proposed automation.
Useful outcome
Produce an initial annual saving, first-year ROI and break-even estimate.
OpenAI guidance on writing clear instructions, adding context and improving prompts iteratively.
Useful outcome
Turn a vague request into a clearer prompt with context, constraints and a defined output.
Anthropic documentation covering clear instructions, examples, context, structure and role prompting.
Useful outcome
Build a reusable prompt structure for a repeated task and test it against varied inputs.
OpenAI's guide to deciding when an agent is appropriate and combining models, tools, instructions and guardrails.
Useful outcome
Decide whether a workflow needs an agent and sketch its tools, controls and escalation points.
Anthropic's engineering explanation of workflows, agents and the value of starting with the simplest workable pattern.
Useful outcome
Separate deterministic workflow steps from genuinely autonomous decisions.
UK Information Commissioner's Office guidance on data protection risks across the AI lifecycle.
Useful outcome
Identify privacy questions that need an owner before personal data enters an AI workflow.
A voluntary framework for governing, mapping, measuring and managing risks in the design and use of AI systems.
Useful outcome
Create a first risk checklist and allocate ownership for evidence, controls and monitoring.
First-party product demonstrations, research presentations and event recordings from OpenAI.
Useful outcome
Follow product changes from the maker while keeping demonstrations separate from independent evaluation.
First-party talks, product walkthroughs and technical material about Claude and Anthropic's research.
Useful outcome
Find current Claude demonstrations and compare the claims with your own task tests.
A one-page working template for defining the job, current friction, desired outcome and evidence before selecting AI technology.
Useful outcome
Turn a loose AI idea into a small, comparable opportunity statement.
A reusable template for agreeing inputs, output, rules, human decisions and completion criteria for an AI-assisted task.
Useful outcome
Give a person or AI system a testable brief rather than an open-ended instruction.
A compact set of test cases, expected results, failure notes and acceptance questions for evaluating an AI workflow.
Useful outcome
Compare workflow behaviour across normal, difficult and unsafe examples before adoption.
Proposed founding test · not yet for sale
The public library remains the discovery layer. A small paid membership would add original working templates, maintained workflow packs, progress and an ad-free experience. We will only build accounts and payments after people show they want it.
Founding price to test
Two months effectively free
No long commitment