When should I automate a task vs use AI to augment my own work?
TL;DR
Automate repetitive, well-defined tasks with clear success criteria. Augment high-judgment tasks where your expertise drives the decision.
This is the most useful framework question in the current AI moment, and most people get it backwards by automating judgment calls and manually doing things that should run on autopilot.
The decision matrix
| Task type | Best approach |
|---|---|
| Repetitive, rule-based, success is clear | Automate — build a pipeline, run it unattended |
| High-judgment, expertise-dependent | Augment — AI as co-pilot, you make final calls |
| Creative, exploratory, taste-dependent | Augment — AI generates options, you curate |
| Novel, first-time-doing-this | Augment first, automate later — understand before automating |
What "automate" actually means here
A fully automated workflow runs without your review in the loop. Examples:
- Social media scheduling from an approved queue
- Newsletter RSS digests with templated formatting
- File organisation and naming conventions
- Monitoring alerts and basic triage
These have clear success criteria. You can verify they work by spot-checking output.
What "augment" means
AI assists you in a workflow where you remain the decision-maker. Examples:
- Writing: AI draft → your edit → publish
- Research: AI aggregates → you assess relevance → you synthesise
- Code: AI suggests → you review → you approve
- Strategy: AI generates options → you score and pick
The failure modes to avoid
Over-automation: running AI unattended on anything that goes to customers or represents your brand. Voice gets lost, errors compound, you find out late.
Under-automation: spending 3 hours a week on tasks a $0.02 API call could handle. Opportunity cost is real.
Premature automation: automating before you understand the task well enough to write good success criteria. Always do it manually first.