The operating idea, in sixty seconds
Shortform cuts from the journal and the channel. One idea, one minute, one surface.
When a task gets stuck, I give it fifteen minutes, then I change the plan. After fifteen minutes, stubborn is not a strategy, so I pick one: ask someone, hand it to AI, or drop it for today. Most stuck tasks are not hard; they are just stuck.
Watch →An email here, an invoice there, a quick approval. Each one breaks your focus. So I batch: all approvals at one time, all replies at one time, all the small stuff in one block. The rest of the day stays clean for deep work. Synthetic demonstration data, not client results.
Watch →Do the work in front of you, with full care, and let go of the result. When I stop gripping the result, I see the work more clearly: fewer panic moves, better calls. Care about the craft, and the scoreboard follows.
Watch →The yearly plan looks cheaper, but a discount on a tool you stop using is still money gone. Every new tool starts monthly and has to earn its place before I switch to yearly. Synthetic demonstration data, not client results.
Watch →A limit is not a problem, it is the design brief. If work cannot spill into the night, it has to fit inside the day: fewer meetings, small stuff batched, AI doing the typing. Pick your hard line first, then build the work around it.
Watch →One long video goes in, and out come clips, a post, a thread, an email, a quote card, a blog draft, a podcast cut, captions, titles and a thumbnail idea. AI does the cutting and the first drafts, and I check every line before it goes out. Synthetic demonstration data, not client results.
Watch →Every project I run gets one of three labels. Grow gets my best hours and a clear goal, maintain runs on a checklist with no new ideas and no new spend, and admit it means it is done and we say so out loud. The danger is the fourth label nobody writes down: hope.
Watch →Satya is Sanskrit for truth. At work it means real numbers in every room, bad news reaching me first, and no dashboard theater, because a dashboard that flatters is a dashboard that lies. Teams need one person whose numbers nobody has to double check.
Watch →Before any big meeting I make three pages: their world, my evidence, my questions. AI pulls the background, I check anything I will say out loud, and every likely question gets matched to a real result. Synthetic demonstration data, not client results.
Watch →How I take on any new operating problem. Days 1 to 30: listen and map, and change almost nothing. Days 31 to 60: ship one fix a skeptic can see working, and by day 90 the scoreboard goes live with numbers the team agrees to.
Watch →AI builds a one-page Monday brief overnight from my calendar, my inbox and last week's scoreboard: three priorities, a watchlist of deadlines, late payments and quiet deals, and the one thing that breaks the week if I ignore it. Then I am operating, not orienting. Synthetic demonstration data, not client results.
Watch →Five things: systems that survived, hours given back to the team, clean calls on the decisions you cannot undo, people who now run things without them, and truth under pressure. Not on the list: hours worked, tools known, buzzwords. Judge outputs, not effort.
Watch →Five stages: idea, owner, in progress, shipped, reviewed. Every item gets one owner, one next step and one date. No date means no project. Every other week I review it like revenue: what moved, what is stuck, what dies. A project with no next step is a wish.
Watch →The first question I ask in any company: are we on track, and how long does it take you to answer that? Then four fast checks: where decisions wait, which number nobody trusts, which process lives in one head, and what gets redone because of versions. Five answers, and the real roadmap writes itself.
Watch →Growth by subtraction. Adding feels productive, but cutting is where the leverage is. My review asks one question: what stops? Cut until the important thing is loud, then feed only that.
Watch →A process that lives in one head is a risk. Record the task once, talking out loud, and AI turns the transcript into a numbered checklist with edge cases, an "if it breaks" section and an owner. Synthetic demonstration data, not client results.
Watch →Every plan looks great in January. Then it is a Tuesday in March: the shipment is late, a key person is out, the launch hits audit week. Companies die in the gap between the deck and the Tuesday.
Watch →Not inbox zero. Inbox three. Every morning AI sorts my email into needs me, drafted for a one-tap yes, and handled. Email is a conveyor belt, not a room you live in. Synthetic demonstration data, not client results.
Watch →The old reversible-door test, the way I run it. Two-way doors get minutes. One-way doors get the hours, a second opinion and a night of sleep. Most teams do it backwards.
Watch →My meetings end with every decision in writing: owner, date, done. One prompt turns the transcript into decisions, open questions, named risks, and a list called "decided by nobody". Synthetic demonstration data, not client results.
Watch →AI does the typing in my work: drafts, data pulls, sorting, summaries, reminders. It decides nothing. Prices, people and promises stay with me.
Watch →I built myself an exec team. A CFO, a CMO, merch and ops. All AI. Every number has to point to a cell in my own files, and when it cannot prove something it says not established, instead of guessing. Synthetic demonstration data, not client results.
Watch →I do the work of ten people now. Not because I work harder. Because I hired AI, and I treat it like a team. AI does not make you 10x. It makes your judgment 10x louder.
Watch →Eleven tools. No single view of a customer. I connected them into one warehouse. The demo uses synthetic data, not client results.
Watch →I used to read the numbers monthly. By then the month was already gone. Weekly is how you catch a miss while you can still fix it.
Watch →I made several CEO mistakes. I still got through them, because I stopped repeating them. You do not have to be brilliant. You have to stop doing the same fumble twice.
Watch →We got bigger. We did not get richer. I fell in love with revenue before profit, and growth without profit is just a more expensive job.
Watch →I discounted when I got scared. Fear pricing feels like a win in the week, and it trains the customer to wait for the next cut.
Watch →I spent money where the customer never looks. Pretty work that does not change the buy is a tax on the P&L.
Watch →Everybody posts the highlight reel. Here are my fumbles instead. That is where the lesson actually is.
Watch →There is no talent shortage. There is an operator shortage. I use one test: who actually owns the number when it misses.
Watch →Private equity is sitting on about $1.3 trillion it cannot put to work. The money is not the constraint. The operators are.
Watch →PE boards do not start on page one. They go straight to three numbers. If any of them is a surprise, the rest of your story does not matter.
Watch →Before I owned a P&L, I thought I understood customers. Then the question changed from conversion events to unit economics.
Watch →Early at Nike I had no formal authority, and making the case louder did not move anyone. Influence without a title is still the skill I use most.
Watch →Your numbers can look great while nobody knows how to work with you. Every company has two org charts: the one on paper, and the one that actually moves.
Watch →300,000 five-star reviews look like a marketing number. I read them as a floor, not a ceiling. The most useful brand signal is not what your best customers say.
Watch →I thought a PE-backed CEO job was: run the company, hit the targets, build the team. They hire you to optimize for an exit. That changes every decision.
Watch →