WICKED.TODAY
03 / PUT IT TO WORK

Less tool collecting.
More finished work.

Start with a task, a deliverable and a quality check. These playbooks give you a small, repeatable way to try AI in real work.

01 / RESEARCH

Turn a question into a source-backed brief.

Objective: make a decision with evidence you can trace. Use AI to organize a small set of sources and surface disagreements.

Get the source packet and brief template Find research tools

DELIVERABLE

A one-page brief with a recommendation, three supporting facts, linked sources and open questions.

  1. Define the decision.Write the question, intended reader and what would change your recommendation.
  2. Collect the evidence.Choose three to five primary sources. Record each URL, publication date and the relevant passage.
  3. Ask for a structured draft.Provide the source text. Ask AI to separate supported facts, conflicting evidence and unknowns, with a source for each claim.
  4. Check and compress.Open every cited link, verify the important claims and edit the brief to one page.

QUALITY CHECK

Can a reader trace every decision-critical fact to a real source? Remove unsupported claims and label uncertainty.

02 / BUILD

Ship one small, reviewable feature.

Objective: use an AI coding assistant to produce a focused change you can understand, run and review.

Get the mini feature and runnable tests Find coding tools

DELIVERABLE

A small diff, a working demo and a record of the checks you actually ran.

  1. Write the acceptance check.Describe one user action and its expected result. Include an empty state and a failure case.
  2. Give the assistant context.Share the relevant files and project conventions. Ask it to explain the approach before editing.
  3. Implement in a small slice.Ask for the minimum change that satisfies the acceptance check. Review dependencies, data handling and the diff.
  4. Run it yourself.Exercise the user flow and its failure case. Run the relevant project checks and fix any regression before merging.

QUALITY CHECK

Can you explain the diff and reproduce the result? A successful AI response does not replace a passing check.

03 / AUTOMATE

Take one repetitive task off your plate.

Objective: build a narrow automation that is easy to inspect, pause and recover when its input changes.

Get the duplicate-safe worked example Find automation tools

DELIVERABLE

A repeatable flow with a sample run, an error log and a documented manual fallback.

  1. Map the manual task.Write its trigger, input, transformation and output. Pick a task you repeat often with a clear success condition.
  2. Build with sample data.Use a copy or sandbox. Start with a dry run that shows the proposed output before making changes.
  3. Test the awkward cases.Try missing fields, duplicates and a failed service request. Add a limit, error message and safe stop.
  4. Roll out with a review gate.Review the first outputs. Keep human approval for external sends or destructive actions and document how to disable the flow.

QUALITY CHECK

Does a retry create a duplicate or cause an unintended action? Fix that behavior before scheduling repeated runs.

These original playbooks are planning templates, not verified product tutorials. Follow each full guide for a reproducible worked example and acceptance checklist. Tool links lead to source briefs; check current documentation and share only data you are allowed to provide.