A delegation principle traced to a 2021 quote from Evan Veneter: "if you can hire someone to do a task 80% as good, do it." The claim is that accepting a short-term drop in quality or speed while training a delegate — human or AI — pays off over the long run, because the delegate's ceiling rises with training while your own bandwidth stays capped.
Applied to AI agents specifically, this is the argument for pushing through the friction of Step 3 – Supervised Autonomy (AI Agency CEO Model) rather than reverting to doing the task yourself the moment an agent's first attempt is slower or worse than your own.
Apply: When handing a task to AI feels initially slower or lower-quality than doing it yourself, treat that as the expected cost of training the delegate and push through rather than reverting to solo work.
A decision rule for deciding what to hand fully to AI: if a task's output is fine at roughly 80% quality, it's a good candidate for full automation; if quality on that task is critical, don't automate it end-to-end.
"If you can hire someone to do a task 80% as good, do it." "Where is good good enough? Automate that. Where is good not good enough? And don't automate that."
Apply: sort your task list by whether 'good enough' output is acceptable — automate the ones where it is, and keep quality-critical tasks under closer human control rather than fully delegating them.