Never automate a process that shouldn't exist in the first place.
Most people think automation is the first step to efficiency. It's the last. Someone sees a slow process and immediately reaches for tools. n8n, Claude Code, custom scripts, whatever gets the steps moving faster. But if you automate a mess, you just get a faster mess.
Before you touch a single tool, you need to do what I call Workflow Archeology: digging into the foundations before building on top of them.
The Dig
Every process has layers. The visible layer is what people do today. Beneath that are the assumptions that shaped it, decisions made months or years ago by people who may have left the organisation. Beneath that are the original requirements, which may no longer be valid.
The Dig means excavating all of it. Who owns this process? Why does each step exist? What was the original requirement? Is that requirement still real?
This is uncomfortable. People don't like being asked "why do you do this?" because it feels like a challenge to their competence. It's not. It's a challenge to the system they inherited.
Requirements from smart people are the most dangerous. You question them less. When a senior leader implemented a process two years ago, nobody wants to ask whether it still makes sense. But smart people make contextual decisions, and context changes. What was the right call in 2023 might be dead weight in 2025.
The Purge
Once you understand why everything exists, you can start deleting.
Most teams skip this entirely. They go from understanding the process to automating it. But the highest-return move is removing steps, not speeding them up.
If a step doesn't serve a current, validated requirement, it goes. If a form captures data nobody uses, it goes. If an approval exists because "we've always done it that way," it goes. Not simplify. Delete.
Elon Musk talks about this ordering explicitly, and I think he's right about the sequence: challenge the requirements first, then delete what shouldn't exist, then simplify what remains, then make it faster, and only then automate.
Each step depends on the one before it. Simplifying a process that contains unnecessary steps just gives you a cleaner version of something that shouldn't exist. Automating before you've done the preceding work means you're locking in every bad assumption, every dead requirement, every vestigial approval, and running it all at machine speed.
Most organisations start at the last step and wonder why their automations create more problems than they solve.
The Speed
Only after The Dig and The Purge do you earn the right to accelerate.
Now every remaining step exists for a reason you can articulate. Every handoff serves a purpose. Every input has a consumer. You're not encoding dysfunction into software. You're taking a validated, minimal process and making it fast.
This is where automation becomes powerful instead of dangerous.
But the discipline is doing this repeatedly. Processes accumulate cruft. Every quick fix adds a step. Every edge case adds a branch. Every new stakeholder adds a review. Six months after you clean a workflow, it'll have grown new layers unless someone is actively watching.
Simplify first. Automate last. The most powerful automation move you'll ever make is deleting the process that shouldn't exist, before anyone wastes time making it faster.
Views expressed here are my own and don't represent any other parties.