The Operating Stack
A five-layer diagnostic for teams that have the tools and still don't have a system. It is the same method I run inside a paid audit, written out in full. You do not need me to use it.
Every team I meet is working the stack upside down.
They start at the top. Pick the tool, build the workspace, then try to talk everyone into using it. That last part is where it dies, every time.
It is an understandable order to work in, because the tool is the only layer that is visible. You can see a workspace. You can show it to someone. You cannot see a handoff, or a trigger, or the fact that one person quietly redoes half of what another person sends them. So the tool gets all the attention and all the blame.
Underneath it sit four layers nobody has looked at. When those are undiagnosed, the tool inherits every one of their problems and then gets held responsible for them. That is why teams end up on their third project manager in two years, each time convinced the last one was the issue.
The debt is not technical. It is descriptive. Nobody ever wrote down how the work actually happens, so every decision on top of it is a guess. And a guess automated is still a guess, running faster.
Never automate a workflow that should not exist.
Most of what a team wants to speed up should have been deleted, merged, or handed to one person a year ago. Automating it makes the wrong thing permanent, and gives it a maintenance cost on top.
So before we build anything, we work out what is actually worth your time and what is not. The five layers below are how you do that in order.
Read it from the bottom, work it from the bottom.
Each layer answers one question. You cannot answer a higher one honestly until the one below it is done, which is the whole reason the order matters.
Start with the thing that eats your Tuesday. The step everyone has quietly built a workaround for. The task that leaves the whole team flat. Not what is trending, not what you watched someone automate online. What is costly, in your business, this week.
The word that matters here is costly, not busy. Busy work is obvious and everyone complains about it. Costly work is the block you brace for on Sunday night, the one that takes ninety minutes on the calendar and three hours of your attention. Teams consistently underestimate it, because the preparation and the recovery never get counted.
Most operations fail at this layer for a boring reason: nobody has a number. Without one, priorities get argued rather than calculated, and the loudest problem wins instead of the most expensive one.
Take the costly work and break it apart until every step is visible. Three things to name, in this order: what sets it off, what happens while it runs, and what it leaves behind.
The trigger is the part almost nobody examines, and it is usually where the delay lives. Work that starts when someone notices behaves completely differently from work that starts on a defined signal. If your onboarding begins whenever the account manager remembers, you do not have a slow onboarding process. You have an unreliable trigger with a process attached.
Most work that feels enormous turns out to be five ordinary steps and one bad one in the middle. You cannot see which one is bad until the list exists. This is also the layer that decides whether anything can be delegated, taught, or automated later, because all three require a description, and a description is exactly what most teams have never written.
One step's output is the next step's input. A person's habit is part of the machine as surely as any software is. Until you can see the whole circuit, including where things change hands and who is quietly waiting on whom, you are tuning parts of something you have never actually looked at.
Delay almost never happens inside a step. It happens in the gaps between them, which is precisely where no tool is watching and no report is counting. A task that takes twenty minutes of work and four days of elapsed time has a connection problem, not an effort problem, and no amount of optimizing the twenty minutes will touch it.
Copy and paste belongs here too. It is a workflow. It is just an undocumented one performed by a person, and it is almost always the cheapest thing on the entire list to remove.
This is the turn, and it is the layer worth slowing down for.
Your team has things only it can do. The judgment. The relationships. The taste. The particular way your best person reads a room. Name those precisely, in writing, and something useful happens immediately: everything you did not name is now fair game.
That is why this layer belongs before the tools rather than after them. Teams who skip it either automate nothing, because every candidate feels risky, or automate the wrong things first, because nothing was ever ruled out. The list is what makes the decision calm.
It is worth being plain about the point of this. It is not about needing fewer people. It is about giving the ones you have their week back, so it goes to the work that actually required them. The real cost of a weak system is rarely the hours. It is that your most capable person spends them on work almost anyone could have done.
The question to sit with
Write down the three things your team does that nobody else and no software could do. Then look at last week honestly and work out what share of the time actually went there. The gap between those two answers is your agenda for the next quarter.
Notion, Claude, the automations between them. This is the last decision, not the first.
By the time you arrive here the choice is usually obvious, and usually smaller than what you would have bought on day one. You are no longer shopping for a system. You are looking for something to carry a specific, described, priced piece of work that you have already decided should not be done by a person.
That is a much easier thing to choose well. It is also a much easier thing to get a team to adopt, because you are not asking them to change how they work. You already changed how the work runs, and the tool is just what it runs on.
The first hour of every audit I do, given away.
Nothing here is withheld. This exercise is the opening move of a paid engagement and there is no reason you cannot do it tonight with a calendar and a blank page.
Open your calendar to the last two weeks
Your real one, not a typical week. If your work is not calendar-driven, use your task list or your sent folder instead. Recent and real beats representative.
Mark every block that felt costly
Costly, not busy. The ones that took preparation, or attention, or recovery afterwards. You will usually find between four and seven.
Take one and write down what happened before, during, and after
Before is the trigger and the preparation. During is the sequence. After is the follow-up, the tidying, the thing you had to chase. Most of the hidden cost lives in the before and the after, which is exactly why an hour on the calendar is never an hour.
Put an honest number on it, per week
Across everyone who touches it, not just you. Estimates are fine. Being roughly right beats being precisely unsure.
Ask how much of it is identical every time
This is your proxy for how cheap it is to change. Work that is the same every time is the cheap work. Work that is different every time usually needs a person, and that is a finding rather than a failure.
Repeat for three or four, then rank them
Hours consumed against how repeatable it is. Start with whatever scores highest on both. Momentum is worth more than optimality here, so take the fast felt win first even if a bigger prize sits further down.
If you would rather have this done with you, and scored, the Self‑Audit runs the same logic in about eight minutes and gives you a report at the end.
It came out of teaching, not software.
I spent ten years building Notion workspaces for venture capital and private equity firms, marketing and talent agencies, consulting practices, and early-stage startups. More than a hundred of them. The ones that lasted were never the best-designed ones. They were the ones where the team understood why it was built that way.
That pattern is what produced these five layers. When I built a system for someone, it decayed the moment their business changed, because the reasoning behind it lived with me. When I walked a team through the diagnosis first and then taught them to build it, the thing evolved without me, which is the only outcome worth selling.
So the method is deliberately boring at the top and specific at the bottom. Look before prescribing. Describe before deciding. Decide what only humans should do before deciding what a machine will carry. It is not clever, and that is rather the point. Clever is what people buy when they have not looked yet.
Score your five layers, or have me run it properly.
The Self‑Audit is free and takes about eight minutes. The Operating Stack Audit is the full version: a deep dive into your workflows, a written report on where the gaps and the opportunities are, proposed solutions, and a sequenced roadmap of what to start with.