I will not invent a loss figure to make this essay dramatic. What I will do is name the categories of mistakes that nearly broke CI Web Group — the slow, expensive kind that do not show up in a single bad quarter. They show up in attrition, margin, and the year you realize you became the bottleneck.
What makes a lesson "expensive"?
Expensive lessons share a shape: you knew something was off, you had a story for why waiting was prudent, and the bill compounded in silence. Not one invoice — a drift. Culture, architecture, customer trust, your own calendar.
The restructure from 320 to 38 was not one mistake. It was the invoice for many tolerated drifts finally coming due. I wrote that chapter in From 320 to 38 and Hands Up, chapter seven.
Category one: keeping the wrong leaders too long
Loyalty is a virtue until it becomes negligence. Leaders who were perfect for the last chapter — task managers, hero worshippers, politics-as-strategy — become anchors in the next. I paid in stalled initiatives, A players leaving, and customers feeling inconsistency.
The fix is not cruelty. It is clarity about the job the company needs now — see The CEO's Job Changes Every Year.
Category two: confusing activity with architecture
We were busy. We were not compound. Departments bought tools. Nothing connected. That is how you rent intelligence forever instead of building a revenue engine. The agency model felt safe because it was familiar. Familiar was the expense.
Category three: soul without systems
I cared. Customers still hit voicemail. Teams still burned out heroes. Caring without encoding is private virtue and public failure — the whole argument in Systems and Soul. The lesson cost me credibility I had to earn back with work, not words.
Category four: waiting for consensus on the obvious
AI was not ChatGPT for us. It was architecture. Waiting for every voice to feel ready meant the ready competitors moved first. I should have named the rebuild earlier, with dates, and carried people through instead of letting fear masquerade as prudence — bringing the team with you is the people side; this is the CEO delay tax.
Category five: measuring the wrong scoreboard
Headcount felt like progress. Revenue without margin discipline felt like winning. I write how I measure now in How I Measure Business Success — the old scoreboard hid problems until they required surgery.
What repair looks like
Repair is public and dated. Repair is replacing a leader when the job changed. Repair is killing a tool that only served internal theater. Repair is writing Hands Up so the industry hears the human cost alongside the architecture win — not because guilt is strategy, because trust is.
The most expensive mistakes are the ones you narrate as patience.
Category six: under-investing in bench
I waited too long to build leaders who could run domains without me — then wondered why I was exhausted. Bench-building is not "maybe someday." It is leadership teams that scale, hired and trained before you need them desperately.
What changed after the bill came due
Public accountability. Architecture ownership with Chris. Hiring for stewards — A players who build systems, not drama. Client-first as enforcement, not poster — The Client-First Inversion.
I still get it wrong. The difference is speed to name it — on stage, in the book, in essays like this. Operators respect repair more than perfection.
Action
Pick one category above. Write the drift you are tolerating today. Set a date to fix or stop. If the drift is company-sized, read Hands Up — it is the letter I wrote so fewer owners pay the same tuition. If you want a partner on the fix, work with me.