The sycophancy tax: how agreeable AI makes us worse at decisions
Agreeable AI feels helpful, but a Science study shows it leaves us surer we're right and less willing to listen. Here's how to design dissent back in.
176 essays
Agreeable AI feels helpful, but a Science study shows it leaves us surer we're right and less willing to listen. Here's how to design dissent back in.
Anthropic found its AI agents spent 80% of session time waiting on a human. Your attention is now the bottleneck. Here's how to make agents need less.
IBM lost 13% in a day when AI learned COBOL. If clients pay you because the work is hard, AI is coming for that premium. Here are the moats that hold.
AI multiplies the skill you bring to it, so your power users can keep pulling ahead. Why counting logins misses that gap, and what to measure instead.
Beyond a certain point, more intelligence stops helping at work. Judgement, persistence and follow-through decide who gets results, so hire for those.
Before you buy an 'AI-powered' feature, ask to see it with the AI switched off. Three vendor questions that separate real AI tools from a new sticker.
Small firms can't hire and won't fire, so the team you have is the team you'll grow with. That calls for designing the work, not running more job ads.
Executives say AI saves them hours a week; staff say it saves nothing. That gap measures how well bad news travels upward in a business, not the tool.
T-shaped hiring assumed humans did the work. Now that AI does the first pass, hire people who can frame problems, direct systems and judge the output.
OpenAI just signed McKinsey, BCG, Accenture and Capgemini to sell its agents. Most AI consulting is training in disguise. Here is what to buy instead.