Open this lesson in your favourite AI. It'll walk you through the why, explain the demo, and quiz you on the try-it list.
You cannot delegate every task to an AI, and you should not try. Some of your work is routine — the answer is deterministic, the stakes are bounded, and a bot that gets it 95% right is a huge win. Some of your work is judgment — the answer depends on context only you carry, and a bot that gets it 95% right is a disaster the 5% of the time it's wrong. And some of your work is learning-critical — it feels routine but is how you actually build the intuition your future judgment will draw on, and delegating it means you gradually stop being able to do it yourself. This module is about looking at your real week and putting every recurring task in one of these three columns before you write a line of bot code — because 80% of the bot's design decisions collapse out of that categorization, and 80% of the failure modes come from getting the categorization wrong.
A minimum-viable delegation matrix: three columns, one row per recurring task. The rubric is deliberately blunt so you can rank a week's worth of work in 30 minutes.
Use these three in order. Each builds on the one before.
In one paragraph, explain the difference between routine, judgment, and learning-critical work like I'm new to thinking about delegation.
Walk me through why delegating a learning-critical task can silently degrade my judgment over months, using a concrete example.
I'm about to delegate weekly report writing to an AI. Given that it's how I currently spot company-wide patterns, what's the right compromise: full delegation, assisted drafting, or something else?
# A 30-minute rubric for placing recurring tasks in the delegation matrix.
# Score each task on three axes; the highest score picks the column.
def classify(task):
routine = 0
judgment = 0
learning = 0
# Routine signals
routine += 2 if task["deterministic_answer"] else 0
routine += 1 if task["bounded_stakes"] else 0
routine += 1 if task["happens_weekly_or_more"] else 0
# Judgment signals
judgment += 2 if task["needs_context_only_i_have"] else 0
judgment += 2 if task["wrong_answer_burns_trust"] else 0
judgment += 1 if task["stakeholder_expects_me_personally"] else 0
# Learning-critical signals
learning += 2 if task["how_i_build_intuition"] else 0
learning += 1 if task["im_still_getting_better_at_it"] else 0
learning += 1 if task["disappears_if_i_never_do_it_manually"] else 0
scores = {"routine": routine, "judgment": judgment, "learning": learning}
return {"column": max(scores, key=scores.get), "scores": scores}
# Try it on a real week
tasks = [
{"name": "Weekly investor update draft",
"deterministic_answer": False, "bounded_stakes": False, "happens_weekly_or_more": True,
"needs_context_only_i_have": True, "wrong_answer_burns_trust": True, "stakeholder_expects_me_personally": True,
"how_i_build_intuition": True, "im_still_getting_better_at_it": True, "disappears_if_i_never_do_it_manually": False},
{"name": "Reply to recruiter cold outreach",
"deterministic_answer": True, "bounded_stakes": True, "happens_weekly_or_more": True,
"needs_context_only_i_have": False, "wrong_answer_burns_trust": False, "stakeholder_expects_me_personally": False,
"how_i_build_intuition": False, "im_still_getting_better_at_it": False, "disappears_if_i_never_do_it_manually": False},
]
for t in tasks:
r = classify(t)
print(f"{t['name']:40s} -> {r['column']:9s} {r['scores']}")
python3 main.py