April 24, 2026
Insurance that burns your house
Familiarity is the enemy: On why Enterprise systems have failed for 60 years
Big logos feel safe; bold ideas still get a hard no
TLDR: A founder says enterprise knowledge tools keep failing because buyers choose safe brands over working products. The crowd split: some mocked “big-name insurance,” others doubted AI can auto-organize messy info, and many warned that startup risk is real—showing why this tug-of-war still stalls progress.
An AI founder says enterprise knowledge tools have flopped for decades because buyers pick what feels familiar, not what actually works—and the comments section turned it into a roast. A senior exec reportedly loved the startup’s demo but wouldn’t buy it because they needed the “big-name insurance.” Commenters cheered the call-out, with one zinger saying it’s like buying fire insurance from the company that torches your house. Others memed the latest trend as “just add AI to your wiki” and slapped a big “LOL” on it for good measure.
But it wasn’t all high-fives. One camp pushed back on the tech claims, arguing that large language models (LLMs—AI that predicts words) don’t care about your favorite programming language the way humans do. Another camp said the real blocker is human: nobody wants to do the boring organizing work, and magic “auto-structure” promises sound like wishful thinking. The HP–Autonomy fiasco got dragged back on stage as Exhibit A in why big brands aren’t actually safer. Yet risk hawks countered with a hard truth: if a tiny vendor disappears, who maintains the system for the next 10 years? Verdict from the crowd: enterprise buyers crave cover, innovators crave courage, and the gap is where good software goes to die.
Key Points
- •The author argues enterprise buyers prioritize familiar, large vendors as perceived “insurance,” overriding product quality and cost.
- •A recent demo to a senior enterprise AI lead was praised but not purchased due to risk concerns about buying from a small vendor.
- •Large consulting firms quoted high-priced projects with accuracy claims (up to 99.5%), selling their learning curve rather than a finished product.
- •HP’s 2011 $11.1B Autonomy acquisition and 2012 $8.8B write-down are cited as a major example; Mike Lynch was acquitted by a U.S. jury in June 2024 and died two months later.
- •The essay estimates enterprise knowledge management has cost over $250B in losses and criticizes “just add AI to your wiki” as the weakest current iteration.