You added an approval step last quarter. Nobody remembers who asked for it. You added a dashboard to track how long approvals take. Now someone wants to "fix" the slowdown by automating the whole chain, faster.

That is the exact moment most businesses make things worse.

Automating a broken process does not fix it. It makes the broken version run at full speed, with software wrapped around it so tightly that nobody can ever pull it back out. You have built a machine to defend a mistake, and it hums along smelling exactly like the stale process it replaced.

2026 is pouring gasoline on this. Every founder is racing to wire up an AI agent for something this month. The pressure is real: automate or fall behind. But an agent bolted onto a bloated process just produces bloated output faster. Speed is not the fix. Speed is the amplifier.

There is an older, better order of operations, and it comes from a source you would not expect for a sales and marketing firm: Elon Musk's engineering teams at SpaceX, Tesla, and Boring Company, documented in Walter Isaacson's 2023 biography. He calls it The Algorithm. Five steps, always in this order.

1. Question every requirement. Every rule, every approval, every recurring meeting needs a named human owner: a person, never "legal" or "the department" or "best practice." No name attached? Delete it. This is the step almost everyone skips, because questioning a requirement from someone smart feels like a fight not worth having. That instinct is exactly how bad requirements survive for years.

2. Delete the part or process. Before you optimize anything, try removing it completely. Musk's own test: if you are not adding roughly 10% of what you deleted back in, you were not aggressive enough. Most people are too gentle here. The bias should run the other way.

3. Simplify or optimize. Only now, after 1 and 2. Musk's line for this one is the sharpest: "the most common error of a smart engineer is to optimize a thing that should not exist." Say it out loud in your next planning meeting and watch who flinches. It's a bitter one to chew on if you spent the quarter polishing the wrong step, but it's the sentence in this whole framework worth remembering.

4. Accelerate cycle time. Speed up whatever is still standing. Doing this earlier just gets you to a wrong answer faster.

5. Automate. Dead last, always. Automate before steps 1 through 4 and you are building expensive, rigid infrastructure around something that should not exist in the first place.

Here is the same failure in miniature. Picture the request as it actually lands in your inbox. A team asks: "can we add an approval step before this goes out, and build a dashboard to track how long approvals take?" Run it through The Algorithm instead. Question: who owns the requirement that this needs approval at all? No name, no reason, it is a leftover from when the team was three people. Delete: remove the approval step, replace it with a standard the sender checks against themselves. Simplify: one checklist, three items, checked before sending. Accelerate: sends now go out same day instead of waiting on a reviewer's calendar. Automate: nothing yet, and nothing until the manual version has run clean for a month. No dashboard was ever needed. The thing it was measuring should not have existed.

We ran this on our own sales process first. The 8 Steps of the Repeatable Sale, the method we teach every client, looks simple because it used to be twice as long. We questioned every stage, deleted the ones that existed to make the seller feel busy rather than move the buyer forward, simplified what survived into a fixed sequence, and only then built tools to move faster through it. It is the same reason the Jetpack Workshop is not a finished system we hand you. We build it live, in the room: question the current process out loud, delete what nobody can defend, and only then talk about what to automate. Most workshops start with tools. This one starts with a knife.

Before Strategy Sprints, I spent years personally coaching sales and negotiation inside teams at Google, BMW, and Airbus. The pattern was identical at every size of company: the process that needed fixing almost always had too much in it, not too little. Nobody in twenty-plus years has ever asked me to help them delete less.

The founders who already run this filter aren't the ones announcing a new automation project this month. They're quietly running fewer, better things, and they run the same five questions again next quarter, not just once.

We turned this into a free, open-source Claude Code skill so you can run your own process, team structure, or half-built automation through the same five questions before you spend another dollar building on top of it.

Grab it here: https://github.com/SimonTheSalesBooster/the-algorithm

Paste in whatever is bloated. Question it. Delete what has no owner. Then, and only then, simplify, speed up, and automate what is left.

(If you'd rather have someone push back on your answers out loud, that's what a coffee is for: calendly.com/simonseverino/coffee-with-simon)

Simon & The Sprinters