Four-category automation triage: automate now / automate later / assist / keep manual — classify by judgment requirement and frequency
Classify workflow steps as 'automate now' (well-defined, no judgment, high frequency), 'automate later' (well-defined, no judgment, low frequency), 'assist' (requires judgment but tools handle mechanical portions), or 'keep manual' (requires judgment, taste, or values only you provide).
Why This Is a Rule
The binary question "should I automate this?" is too crude. Some steps are fully automatable, some need human judgment for a portion, and some should never be automated. The four-category classification provides the right granularity for automation triage by crossing two dimensions: judgment requirement (does the step require human judgment, taste, or values?) and frequency (how often is this step executed?).
Automate now (well-defined + no judgment + high frequency): File renaming, data backups, email filtering, report generation from templates. These are the highest-ROI automation targets — the investment pays back quickly because frequency is high and the risk is low because no judgment is needed. Automate later (well-defined + no judgment + low frequency): Annual report formatting, quarterly data migration, one-time setup procedures. Same characteristics as "automate now" but lower ROI due to frequency — automate when you have spare capacity. Assist (requires judgment but has mechanical portions): Writing (judgment in content, mechanical in formatting/distribution), data analysis (judgment in interpretation, mechanical in data cleaning), decision-making (judgment in evaluation, mechanical in option-set generation). These need human-machine partnership. Keep manual (requires judgment, taste, or values): Relationship decisions, creative direction, ethical evaluations, strategic choices. Automating these produces outputs that miss the point even when technically correct.
When This Fires
- When reviewing a workflow and deciding which steps to automate
- When evaluating a new automation tool and identifying where it applies
- When automation has produced poor results — likely a "keep manual" step was automated
- When prioritizing automation investments for maximum return
Common Failure Mode
Binary thinking: either fully automate or keep fully manual. This misses the "assist" category, which is where most knowledge work lives. A writing workflow isn't "automate" or "don't automate" — it's "automate the formatting and distribution, assist the drafting with templates and outlines, keep the voice and argument manual."
The Protocol
(1) List all steps in your workflow. (2) For each step, answer two questions: "Does this step require human judgment, taste, or values?" (Yes/No) and "How frequently is this step executed?" (High/Low). (3) Classify: No judgment + High frequency → Automate now. No judgment + Low frequency → Automate later. Judgment required but has mechanical sub-steps → Assist (automate the mechanical, keep the judgment). Judgment is the entire step → Keep manual. (4) Prioritize automation in order: automate-now first (highest ROI), then assist (highest complexity-reduction), then automate-later (when convenient). (5) Apply Sovereignty check before automation — if the automation produces wrong output, will you notice? Automate only when the answer is yes's sovereignty check to every automation decision before implementing.