Every company wants to automate. Almost none are willing to start by deleting.
01The Order Matters More Than the Method
The “Musk algorithm” is a five-step engineering discipline used at SpaceX and Tesla, documented by Walter Isaacson and former executives such as Karim Bousta. It reads like a checklist: question, delete, simplify, accelerate, automate.
The value is not the list. It is the sequence. Invert two steps, or skip the first two because they make people uncomfortable, and the project fails quietly — with a signed vendor contract and a deadline met to prove it.
Automation is the last step. In most companies, it is the first decision made.
02The Five Steps
Question every requirement
Every requirement is either improvable or obsolete — regardless of who wrote it. The rule: a requirement never belongs to a function — “legal wants this” — but to a named person. That is the only way it can be challenged. Anonymous requirements live forever.
Delete everything you can
Steps, approvals, mandatory fields, meetings, reports. The 10% rule: if you are not forced to restore at least a tenth of what you removed, you did not cut hard enough. The classic trap is keeping a step “just in case” — which means forever.
Simplify and optimize
Only after the first two steps. Optimizing a process that should not exist means making waste elegant.
Accelerate cycle time
Shorten the time to execute what remains. One warning: never speed up a broken process. If you are heading off a cliff, do not floor it.
Automate — last
CRM, scripts, ERP, AI agents: the tool only comes in once the process is stripped, clear, and fast. Automate too early and you encode the inefficiency — making it harder and more expensive to fix later.
03Why Step 3 Kills Most Projects
The most common mistake is not poor optimization. It is optimizing a process that should not exist. The committee approving a 3% discount, the weekly report no one reads, the three layers of proposal approval — all of it can be made smoother, faster, more “digital.” None of it asks whether it should exist at all.
Optimizing is comfortable: it offends no one. Deleting is not. Deleting forces you to name the owner of a requirement and ask why it still stands. That is why steps one and two are skipped — and why automation programs so often deliver a faster version of the same mess.
“AI does not improve a process. It scales it — flaws and all.”
04Applied to a Sales Organization
In a sales organization, four tests close the gap between what teams say and what actually happens.
The CRM
Count the mandatory fields in your opportunity process. For each, name the person who required it and the decision it supports. Fields that support no decision go. Automating useless data entry is still useless work, only faster.
The pipeline ritual
A pipeline review with no agenda and no single owner is a meeting without a decision. Kill it for a month. If no one asks for it back, you have your answer.
Proposal approvals
Three approval layers do not reduce risk. They dilute accountability and lengthen the sales cycle. Move to one named decision-maker with a clear threshold. Then measure cycle time before and after.
Sales AI tooling
Before you buy, run the process manually in its leanest possible form. If you cannot describe it in five steps, no tool will save you. If you can, the right tool becomes obvious — and its ROI is measured, not promised.
05What It Demands From Managers
This method has a leadership cost. It requires managers who understand the operational detail: the real process, the data, the product. A manager who does not understand how a step actually works cannot challenge a constraint the team has declared non-negotiable. They can only sign off on it.
It also requires a tolerance for being wrong. Deleting aggressively means occasionally deleting too much. That is acceptable — as long as you can reverse it quickly. Iteration speed is the only safety net that matters.
Nothing in this sequence is technical. Question, delete, simplify: these are organizational choices. That is exactly why most companies prefer to start at step five.
“If you cannot explain your process in five steps, you do not have a tooling problem. You have a process problem.”