tech

March 25, 2026

55% of employers regret AI-driven layoffs. Agents are good at tasks and terrible at jobs. Here's what that means for your team and the 3 prompts that close the gap.

The average software job in America lasts somewhere between 18 months and two years. The average AI agent run lasts about two hours. Even if we lament how short job tenure has gotten, those two numbers are not comparable. And the people who really hold institutional context, who keep a business going, often stay four, five, six, seven years. Maybe longer.

55% of employers regret AI-driven layoffs. Agents are good at tasks and terrible at jobs. Here's what that means for your team and the 3 prompts that close the gap.

TL;DR

  • The average AI agent run is two hours, compared to 18-24 months for a software job, highlighting a significant "memory gap".
  • AI agents, while capable of performing tasks, lack institutional context and common sense, leading to errors like wiping production databases.
  • Powerful AI tools that fail silently are more dangerous than mediocre tools that fail obviously.
  • The gap between AI tasks and human jobs is a major problem in tech, leading to over-optimism.
  • Human judgment and contextual stewardship are critical for managing AI risk and ensuring effective integration into organizations.
  • The best tools for managing agent risk are human brains and human-crafted evaluations, not just better prompts or larger context windows.

Continue reading the original article

Made withNostr