There's a specific moment in most failed products where the path gets set. It happens before the most senior engineers are involved. Someone shows an AI-generated demo. An executive gets excited. Large numbers get multiplied, demonstrating enormous market potential. A commitment is made - to a team, a board, the first clients. By the time the senior engineers hear about it, the answer is already yes.
This is not a communication failure. It's a structural one. The people closest to the technical reality are not the people who carry the narrative into decision-making rooms. Enthusiastic executives are. The engineer who would have said "this breaks at scale" or "that demo hid three critical assumptions" is back at their desk, not in the meeting.
Months later, when the product falls short, the failure gets attributed to implementation. Or the market just wasn't ready. Or both. The original promise is rarely examined. No-one revisits those large numbers that were multiplied. This protects the decision-maker and punishes the team that had to execute against an impossible brief.
The incentive structure runs the same direction every time. Bold claims get rewarded with a "yes". Engineering, Sales or Market caution gets labeled as resistance to change. The organizational system produces exactly the outcome it was designed to produce.
The fix is not more technical or market literacy at the executive level - though that might help. It's changing who is in the room when the yes happens. Dissent has to be structurally invited before the decision is being made. For established software companies with existing customers (100+ employees, 10 Mio+ Revenue) new product ideas should ripe for at least six months before any actual work is allowed to start. Use the time to talk to your existing customers, to your engineers, to your salespeople. Put everything into an AI company intelligence layer to get the full picture. Weed out any potential issues, identify risks and hedge against that risk. Only then start to build.
The best mental model for established companies is to treat new products as "company bets". You only start work on them when you're so sure it will be successful in the market, that you would bet the whole company on it. Stop calling trigger-happiness agile.
Thoughts? Find me on Bluesky.