TL;DR. AI increases every company’s ability to execute. Everybody gets the same lift, so the speed of getting work done doesn’t separate companies anywhere near as much as it used to. The fight moves upstream to judgment.
Most decisions today get made without knowing the real cost or the likely outcome. That information has always been too hard to surface in time, so we guess.
AI makes that work cheap. You can now sit down to a dashboard with the real-time, accurate numbers in front of you.
The companies that build the simple architecture for this get better at deciding while everyone else stays roughly the same. That gap will increase and compound over time, the more decision you feed into the loop and analyse, the better you get and you nullify luck.
The Red Queen Effect
“Now, here, you see, it takes all the running you can do, to keep in the same place.” — Lewis Carroll, Through the Looking-Glass
The mental model comes from evolutionary biology, where species locked in arms races adapt against each other in perfect step — and neither side gets ahead. The same principle generalises cleanly to markets. When every competitor adopts the same productivity tool, the lift becomes the new floor rather than an edge. Everyone is running harder while relative position barely moves.
This is the trap founders are walking into with AI right now. The instinct is to deploy it on the same workflows everyone else is deploying it on, mostly chasing speed in drafting and analysis, and to mistake the productivity bump for an edge. It isn’t. Your competitors get the same bump. The frontier moves and you’ve simply kept up.
It’s still the same race, and stepping off the track just means losing. You win it by competing on a dimension your competitors haven’t noticed is open yet. The speed of decisions is going to climb across the board; that lift goes to everyone. What won’t is the quality of decisions being made at speed. Hold quality — better still, raise it — while you move faster, and you start to pull away from the pack. That’s where the rest of this thesis lives.
1. AI accelerates work done.
It can draft a doc in seconds. It can rip through an analysis you’d otherwise hand a junior for three days. Most of the execution layer that used to fill your week now fits in an afternoon, and what you get back the other side is a sharp rise in the number of decisions actually being asked of you.
Time-to-decision collapses. The pressure shifts from can we get this done in time? to what do we actually want to do? — which is a different game entirely, and the competitors who figure it out start moving at a pace you can feel as pressure on the market.
2. Judgment is where focus has to shift.
For fifty years, execution was the scarce resource. The good operator did more, faster, with fewer mistakes.
AI breaks that. Every company gets the same lift. Velocity stops being the differentiator.
Less time burned in execution means there’s no excuse to be sloppy upstream. The question becomes am I making good decisions?
The advantage shifts from executing well to steering well. Increased velocity in the wrong direction just gets you to the wrong place sooner.
3. What makes a decision good?
The norm today is a strange lack of transparency around the decisions actually being made. Most founders sit down to make a call without ever pausing on the real risk or true cost behind it. How much do we genuinely expect this project to run us, and how long do we really think it’ll take? How much time, resource and attention does it actually cost to train a new hire to standard? Reference class forecasting is barely used in the wild, even though it’s the most straightforward way to get a grounded predictive read on whether a decision is sound. So most calls get made on instinct, with no real basis for judging whether the instinct is calibrated.
If you had that data sitting next to you when the call landed, every decision would sharpen. And if you went back later and analysed the outcome of each one against the original forecast — a proper feedback loop — your decision-making would improve markedly over time. That’s not abstract. It’s the same reason a poker player beats the field over a long enough sample. Decisions improve from being studied.
This used to be a luxury. Gathering and correlating that kind of information was too expensive to be worth doing for most calls, so it didn’t get done. AI changes that. The cost of producing high-quality decision inputs has collapsed, and a layer of judgment that used to sit out of reach is suddenly within reach.
There’s urgency on top of that. Competitors are moving faster, which means the cost of carrying around blurry, uncalibrated decision-making compounds every quarter. The window in which good instinct alone could carry a company is closing.
4. The company reorganises around managing the work, not doing it.
The question shifts from what makes a decision good to how do you actually make good decisions, fast, every time a call lands on your desk.
It starts with a dashboard. Not a generic BI screen, but a custom-built one organised around the handful of questions a CEO genuinely needs answered in three seconds with full confidence in the data. What’s the cash position this morning, where’s conversion trending this week, which new hires are ramping and which aren’t, where’s the actual bottleneck in the business right now. The questions whose answers, if you had them at a glance, would change how you run the company day to day.
Once that dashboard exists and is trusted, the human job changes shape. People stop managing other people doing work and start managing computers doing it, and their real job becomes keeping the data clean so the dashboard stays honest. That sounds unglamorous and it is, but it’s where the moat actually lives.
The cascade follows from there. A department head sees the live health of their domain at any moment, not a stale Monday report rebuilt from memory. That same picture rolls up to the executive suite as a real-time read of the whole company, and everyone at every layer ends up making calls against current information instead of last quarter’s narrative.
Then the feedback loop closes. Every meaningful decision gets logged alongside its outcome, and the system learns from itself. The fastest way to improve has always been to study the past honestly — most companies have simply never had the architecture to do it at scale.
The ones that build it now lock in a compounding advantage early that changes the game over months.
