Guide
A practical AI workflow for research and everyday work
Use AI to organize material and expose questions, then return to the source before you publish or decide.
By Cernalo · · 7 minute read
Begin with extraction
Ask for the main claim, supporting points, dates and named assumptions in the material you provide. This creates a compact map without asking the model to judge everything at once.
Check that each extracted point is present in the source.
Move to comparison
Give the assistant explicit criteria such as cost, evidence quality, implementation effort or risk. Ask it to show where the material is silent instead of filling every table cell with a guess.
Challenge the conclusion
Ask which missing observation would most change the answer and what alternative explanation still fits the available facts. This turns the model into a prompt for further investigation rather than an authority.
Draft from verified notes
Once the facts are checked, use the assistant to adapt them for an email, memo or presentation. Specify the audience, tone and length, and forbid invented metrics, quotations or commitments.
Preserve the source trail
Keep links or document references beside your notes. If the result matters later, another person should be able to reproduce the reasoning without relying on the AI conversation alone.