EVERYTHING, WITH THE JARGON TAKEN OUT.

Q1What's the problem?

Your company brought in AI. Somebody is measuring how much time it saves. Nobody is measuring what it costs when it's wrong.

And when it is wrong, the blame doesn't go to the vendor who built it or the executive who bought it. It goes to whoever touched the file last. Usually that's you.

Q2How does AI actually get it wrong?

In 24 documented, repeating ways. Some are loud: it invents facts and sources. The expensive ones are quiet: it drops the word "not" while summarizing a policy, or it agrees with whatever you'd already decided.

Each one is written up the same way: what goes wrong, where it shows up at work, what it could cost, how to catch it, and one thing to do about it this week. The full list is here. It's the best page on the site.

Q3The site uses some odd phrases. Translate them?

TAP A CARD. CORPORATE ON THE FRONT, ENGLISH ON THE BACK.

Q4What's the "155x" thing?

If checking the AI's work costs $1 and prevents $155 of damage, your multiple is 155x. That's a sentence your manager can act on: "here's what my judgment is worth, and here's the math." The full definition and worked examples are here.

Q5How do you know any of this is true?

Every claim on this site carries a tag saying how I know it. Tap each one:

The tag that matters most is the one that admits weakness: MODELED means I did math, not measurement. Sites that never admit that are guessing everywhere. All 36 sources are public and re-checked. The verification record is here.

Q6What's the 90-day thing?

The book ends with a 90-day plan: log the AI errors you actually catch, run one weekly check, do the math on what that's worth, and write the one-page case for your role. About 25 hours total. No permission from anyone required.

And so you don't have to take the plan on faith, the whole protocol is published in the open, free, before you spend a dollar. You can read the plan here. Buyers get every updated edition free.

Q7What's free, and what costs money?