AI Without Illusions

Most AI projects fail. Not for the reasons you've been told.
About 95% of enterprise AI pilots deliver no measurable impact on the bottom line - despite tens of billions spent (MIT, 2025). The cause is almost never the technology. It's a series of comfortable management illusions that sound like good judgment right up until the budget arrives.
AI Without Illusions is the field guide for the manager who has to make these decisions and answer for them - not the technologist, and not the futurist. It refuses both the hype ("AI changes everything") and the cynicism ("AI is a bubble"), and does something more useful: it gives you the arithmetic, the criteria, and the working tools to implement AI without setting money on fire.
Every chapter names an illusion - and hands you the discipline that replaces it:
- ****Most AI projects fail. Not for the reasons you've been told.**
About 95% of enterprise AI pilots deliver no measurable impact on the bottom line - despite tens of billions spent (MIT, 2025). The cause is almost never the technology. It's a series of comfortable management illusions that sound like good judgment right up until the budget arrives.
*AI Without Illusions* is the field guide for the manager who has to make these decisions and answer for them - not the technologist, and not the futurist. It refuses both the hype ("AI changes everything") and the cynicism ("AI is a bubble"), and does something more useful: it gives you the arithmetic, the criteria, and the working tools to implement AI without setting money on fire.
Every chapter names an illusion - and hands you the discipline that replaces it:

• "AI is cheaper than people." Meet the Loaded-Cost Test: would the work still beat the human once you price the reviewer who checks the output, the rework when it's wrong, and a vendor price hike?
• "The pilot worked, so we're ready to scale." A working demo is roughly one-ninth of a shippable product. Here's the other eight-ninths - and why teams keep mistaking the demo for the business.
• "More AI usage means more value." Measure outcomes, never inputs - the same trap that turned agile "velocity" into a number everybody gamed.
• "We should build our own model." The validated arithmetic starts near $192 million. This makes "should we build our own?" a five-minute conversation instead of a six-month one.