Aim AI at the constraint, not the convenient.
The easiest thing to automate is almost never the thing limiting growth. Don't let easy pick for you.
An executive operating model
Every organization is limited by a constraint.
Most leaders optimize everything else. Great leaders build systems that find it, exploit it, elevate it, and when it breaks, begin again.
Scroll00 The take
The team is almost never the issue. The system around them is.
In all my years running operations, I've watched companies lose sales, bleed profit, and in the worst cases go under. Almost never because of their people, and almost never because there was no market for the product. It's because they were running on a broken system: unclear priorities, decisions no one owned, work no one had redesigned.
Fix the system and everything changes. Same team, same product, suddenly winning. The difference is focus: instead of improving everything at once, I find the one constraint limiting the whole company and aim everything at it, AI included, treated as headcount, not magic. When it's solved, I move on to the next one.
Theory of Constraints calls it a process of ongoing improvement.
01 The Problem
Companies attack symptoms.
In 1984, Eli Goldratt showed that every system has at least one constraint, and that improving anything else is an illusion of progress.
The factory floor has changed. The law hasn't. Today's constraints look like unclear priorities. Unowned decisions. A slow operating cadence. Data no one trusts. AI bolted onto work that was never redesigned.
The modern constraint is usually the operating system itself.
02 The Model
The Constraint Loop.
Six moves. Then again. Leadership is not optimizing everything. It is running the loop.
The constraint moves. The loop follows.
A constraint elevated is not a project completed. It is the next constraint revealed. And the loop's oldest warning still holds: never let inertia become the constraint.
In the lineage of Goldratt's Theory of Constraints, extended from the factory floor to the executive floor, for organizations that run on software, people, and AI.
03 AI
AI doesn't fix broken systems.
It accelerates them.
Most AI programs automate whatever is easiest to automate. The constraint is untouched. The waste just moves faster. In the Constraint Loop, AI has one job: elevate the current constraint.
Redesign the work, then add the machine.
Ask what the function would look like if built today. Start from the answer.
An agent is a hire.
A role, a scope, a manager, a performance standard. Treat AI as headcount and the system designs itself.
04 Evidence
The loop, run in the field.
Fifteen years. Different industries, different constraints, same loop.
No. 1 · Enterprise Software
−30% annual software license costs
An application portfolio, put on trial.
The constraint
Software spend growing faster than the organization. An application portfolio no one owned, full of overlapping tools kept alive by habit.
The loop
An application portfolio rationalization process: every application evaluated on three criteria (utilization, cost, and strategic value). Keep, consolidate, or retire.
The result
Software license costs down 30%, measured by annual software expenditures.
I expected to find a few redundant tools. The real constraint was that no one owned the portfolio at all. Habit was the line item.
No. 2 · Global Technology
140% operational revenue growth
A cost center that became a profit engine.
The constraint
The operating system itself: services running on memory and goodwill, with no pricing discipline and no queue. The team wasn't the bottleneck; the system starving it was.
The loop
ERP-driven operations, tiered contracts, standardized rates. The same team, re-pointed at throughput instead of firefighting.
The result
Operational revenue up 140%. Same people. New system. Different company.
Everyone blamed the team. The team was fine. The system starving it was the constraint, so I rebuilt the system, not the roster.
No. 3 · Medical Devices
$800K → $10M revenue in 18 months
The operation couldn't keep up with sales.
The constraint
Everything ran manually, and the bottlenecks sat exactly where money changed hands: lead-to-customer conversion and the warehouse. Sales grew; the operation didn't.
The loop
ERP implementation and automation replaced the manual work. The bottlenecks were exploited first. Leads that once waited days for quotes, approvals, and financing answers got them in hours, so fewer deals died in the queue, and warehouse flow was streamlined to ship what was sold. Focus narrowed from five products to two, and payment plans turned expensive one-time purchases into deals more customers could say yes to.
The result
Revenue grew from $800K to $10M within 18 months.
Sales looked like the win. The real constraint was the quote sitting in an inbox for three days. Money died in the queue, not in the market.
05 About
I build executive operating systems.
Fifteen years running the loop inside high-growth companies: three organizations built from zero, a 1,000-person software company operated on a $150M budget, hospital deployments, AI-native platforms.
I help CEOs find the constraint limiting growth, then build the operating system that exploits it, subordinates everything else to it, elevates it, and begins again.
MBA, Baruch College · Executive Education, MIT Sloan
06 Letters
Letters from the loop.
Occasional letters on finding the constraint: what I'm learning in the field, and where AI actually earns its place. Read them here, or get each one by email.
-
An agent is a hire, not a feature.
Why I give every AI agent a role, a scope, a manager, and a performance review, and what happens to the org chart when you do.
-
The cadence is the strategy.
Most strategies don't fail on the page. They fail in the calendar. On building an operating rhythm the constraint can actually move through.
-
Your constraint is not where you're looking.
The first letter: how to find the one thing limiting the whole company, and why it's almost never the team you're tempted to blame.