Low-Cost Mistake Autonomy Rule
Increase autonomy when mistakes are cheap and direction is easy to change.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 92%
The Low-Cost Mistake Autonomy Rule determines how much supervision a decision needs by comparing the likely cost of error with the speed of correction. When an AI-native team can prototype, ship, observe, and change direction quickly, a wrong choice may be cheaper than the coordination system designed to prevent it. In those cases, leaders set bounded exposure, delegate the decision, and monitor results rather than requiring advance approval. If evidence shows the direction is wrong, the owner changes course immediately. The rule does not celebrate careless mistakes or eliminate risk controls. It concentrates oversight on decisions whose consequences are large, irreversible, regulated, or slow to detect while freeing low-cost experiments from unnecessary management.
Origin
Elena Verna used the falling cost of mistakes to answer objections about larger autonomous teams and reduced managerial supervision at Lovable.
Core principles
- 01Match oversight to the cost and reversibility of mistakes.
- 02Do not impose expensive coordination on inexpensive decisions.
- 03Use rapid correction as a substitute for excessive prediction.
- 04Let independent owners learn through bounded errors.
- 05Reserve supervision for consequential or irreversible risks.
How to run it
- 1
Estimate error cost
Identify the plausible customer, financial, operational, legal, and reputational downside if the decision is wrong.
Pro tip Use a realistic worst case rather than an undefined feeling of risk.
Watch out Hidden security or compliance consequences can make an apparently cheap mistake expensive.
- 2
Test reversibility
Determine how quickly and completely the team can undo the decision after receiving feedback.
Pro tip Prefer staged releases when reversibility is uncertain.
Watch out A technically reversible action may still leave lasting customer or reputational damage.
- 3
Bound the experiment
Limit audience, duration, budget, permissions, or data exposure so the maximum loss remains acceptable.
Pro tip Write the stop condition before beginning.
Watch out An unbounded experiment invalidates the low-cost assumption.
- 4
Delegate without preapproval
Let the owner choose and execute the direction within the agreed boundary.
Pro tip Provide direct access to evidence so the owner can adjust independently.
Watch out Do not delegate accountability while retaining hidden vetoes.
- 5
Correct rapidly
Monitor outcome signals and change direction as soon as the evidence invalidates the original choice.
Pro tip Treat correction speed as part of the operating capability.
Watch out Autonomy becomes negligence when teams ignore clear negative evidence.
In the wild
An owner exposes a new onboarding flow to five percent of new users with a predefined rollback trigger. The flow underperforms, so the owner reverts it the same day and tests a different direction without waiting for a management meeting.
→ The company gains evidence quickly at lower cost than a lengthy approval process.
Common mistakes
Assuming every mistake is cheap
Some decisions involve lasting safety, legal, privacy, financial, or trust consequences and require stronger controls.
Skipping exposure limits
A reversible idea can still cause excessive damage if initially released without boundaries.
Failing to change direction
The rule depends on rapid correction; persisting after contrary evidence removes its advantage.
Is it for you?
Best for
It is best for product experiments, prototypes, internal tools, and other reversible decisions with rapid feedback.
Not ideal for
It is not ideal for safety, legal, security, financial, or reputational decisions where a mistake is difficult to reverse.
From the transcript
“Like the cost of mistakes is not so big anymore. So you don't need that coordination matrix and like that supervision as much.”
“And I'm like, who cares? Because we can really quickly go into a different direction.”
From the episode
Inside the $130M AI Startup Growing Faster Than ChatGPT