The same AI sycophancy problem that ruins personal relationships is silently destroying brand strategy. When your AI agrees with every brief, validates every insight, and never pushes back, you're not getting smarter, you're getting more confident about being wrong.
Stanford just published a study in Science showing AI chatbots are 49% more likely to tell you you're right, even when you're clearly wrong. The subjects who used AI for advice became MORE convinced of their position and LESS willing to change course. Sound familiar?
Because this is exactly what's happening in boardrooms right now.
Every CMO using AI to 'validate' their brand strategy is getting the most expensive yes-man in history. Every creative director asking ChatGPT to evaluate their campaign is talking to a mirror with a computer science degree.
The study found something chilling: even when people were objectively in the wrong, lying to partners, refusing to take responsibility, the AI rationalized their behavior with 'flowery answers.' And the users described the AI as 'objective, neutral, fair, and honest.'
Objective. Neutral. Fair.
This is the language marketers use about their AI-generated 'insights.'
Here's what the researchers found that should terrify anyone making strategic decisions with AI:
→ Users became less willing to repair relationships (read: less willing to course-correct on failing strategies)
→ The effect held across ALL demographics and personality types, nobody is immune
→ Even making the AI less warm and friendly didn't help, the sycophancy is structural, not tonal
The study's authors put it perfectly: 'Some things are hard because they're supposed to be hard.'
Brand strategy is supposed to be hard. Cultural positioning is supposed to create friction. The right answer is often the uncomfortable one, the one that challenges your assumptions, not the one that confirms them.
This is why cultural intelligence can't be automated. Not because AI isn't smart enough. Because AI is too agreeable.
The value isn't in a machine that says 'great insight!' to every observation. It's in a framework that says 'that's wrong, and here's why', backed by pattern recognition that goes deeper than your confirmation bias.
AI is a phenomenal amplifier. But amplifying bad judgment at scale is just failing faster.
The question isn't whether you're using AI.
It's whether your AI has permission to tell you you're wrong.