The Operating ReviewEngagements
A three-week, senior-led read on where your operation is leaking value, with a 90-day plan and a 12-month roadmap.
Free. No signup. Nothing is sent anywhere.
The Survival Test
One question to ask about any AI investment, before you sign anything. It takes about twenty minutes and it costs nothing.
The question
Does this still make sense if the hype disappears completely, and the price of the technology moves ten times, in either direction?
Both directions matter, and they break you differently. If it stops making sense once the technology gets cheap, you never had an advantage, you had a head start. If it stops making sense once the technology gets expensive, you never had a business case, you had an arbitrage on a price you do not control.
A case that only survives at today's price is not a case. It is a bet on the price.
1Start with the business case
One paragraph, ending in a number. If you do not have one written down, that is normal, and it is the first thing this exercise will tell you. Pick a door.
Most people cannot write this paragraph, and that is the real finding. The work has never been measured. So start by measuring it. Paste this into an assistant and let it interview you.
When it hands you the paragraph, bring it back here and we will try to kill it.
2Strike the hype
Click any sentence whose persuasive force depends on the technology being new, exciting, urgent, or something competitors are doing. Be ruthless. It is not an argument if it only works this year.
3Four questions underneath it
4The verdict
5Now let a machine argue with you
You cannot run this test on your own idea, because you are attached to it. A model has no such attachment. But left alone it will flatter you anyway, because agreeing is the path of least resistance and you are the one holding the keyboard. So instruct it not to. Paste this into any assistant.
Run it in both, if you have ten minutes. They flatter differently, and where they disagree is exactly where your case is thin.
WHAT A GOOD ANSWER LOOKS LIKE
It names the sentences it cut, and why. It ends with a number, or it tells you there is not one. And it disagrees with you at least once. If the model likes everything about your plan, you did not get an adversary, you got a mirror. If it comes back warm, reply: “That was too kind. Argue the other side as if you were being paid to kill this project.” Then compare the two answers.
The part worth keeping
If you used a model to help write this case, notice what just happened. The same machine argued for your project and then argued against it, with equal conviction, and it never once told you which answer it believed.
Because it does not believe anything. It has no opinion about your business. It will argue whichever side you set it up to argue, as fluently as you like, and confidence is not a signal of correctness. It is a formatting choice.
So the truth is not in either answer. It is in what survived both. That is the whole discipline, and it is the same discipline whether you are testing a business case or running a system in production.
ONE HONEST CAVEAT
This test is easy to state and genuinely hard to run, because almost everything looks defensible when you are the one who fell in love with it. I have talked myself into things this way. It only works if you let someone who does not care about your feelings run it.
No signup. No email. Nothing you type here leaves your browser.
