The engines · Techniques

Speed is a default you engineer.

Your team isn’t slow because people are timid. It’s slow because the standing rule is ask first. Change the rule, not the people.

The idea, sharpened

Being quick to act is not a personality trait or a value on a wall. Every slow-deciding team is full of people perfectly capable of deciding. They’re slow because the standing rule is wrong.

The standing rule is ask first. To act, you need permission; to get it, you join a queue; and the queue forms at the most senior desk. The leader becomes the bottleneck — not by deciding badly, but by being the place decisions go to wait.

A guardrail is not a rule about who approves a decision. It’s something you build ahead of time so that if a decision goes wrong, only a little breaks — small enough that nobody needs to sign off at all.
The Techniques engine

The two questions that sort any decision

  1. 1 Can you undo it cheaply and quickly?
  2. 2 How much breaks if you’re wrong?
Little breaks
A lot breaks
Can undo
Just do it
No approval. Maybe write it down.
Guardrail it
Go ahead, but behind something that limits the damage.
Can’t undo
Decide fast, write it down
Whoever’s closest decides; record the reasoning.
Gate it
This one stops and comes to the leader. Slow down on purpose.

Move decisions out of the bottom-right — add a guardrail and turn a gate into a go.

The runbook

1

Flip the default, in writing.

If a decision can be undone, make it at the lowest level with the context, and don’t bump it up. Put it somewhere durable so people can point to it when tempted to queue.

2

Build the guardrails.

Staged rollouts. One-click undo with a speed promise. Pre-set spending limits per team. A pre-approved tool list. Written blast-radius limits. A quick note of each decision and why. This is the investment.

3

Keep the gate list short — and defend it.

Gates are for the genuinely can’t-undo-and-a-lot-breaks calls: security, data wiped for good, pricing promises, one-way technical paths, senior hires. Almost nothing else.

4

Settle disagreement without re-opening the decision.

Disagree and commit: note the disagreement once, then go ahead. It only re-opens if the results say so — not because someone’s still unhappy.

5

Measure the decision system itself.

Time from raised to resolved (the headline); how often decisions get bumped to you (should be low); how often fast decisions had to be undone — more than zero, but not high.

The numbers worth watching

Time from raised to resolved the headline — falling
Decisions bumped up to you low and steady
Fast decisions later undone low, but not zero
Make decisions safe by making them small and easy to undo — then let them go fast. A team that trusts its guardrails moves at a speed that looks reckless from outside and is in fact the opposite.