Tuesday, May 26, 2026 · ISSUE
Why Your AI Always Agrees With You
A yes-man with a PhD is still a yes-man. One prompt turns it into a red team that attacks your decision before reality does.
Why Your AI Always Agrees With You
A yes-man with a PhD is still a yes-man. One prompt turns it into a red team that attacks your decision before reality does.
AI Super Simplified
May 26, 2026
Prefer to watch? Here's the 3-minute version:
In 1961, a roomful of brilliant people talked themselves into the Bay of Pigs — one of the worst foreign-policy blunders in American history. The Yale psychologist who studied it gave the failure a name that stuck: groupthink. Smart advisors, eager to agree with the most powerful person in the room, quietly killed every objection until only the bad idea was left standing.
A year later, the same president faced a far scarier decision: Soviet nuclear missiles in Cuba. This time he changed the rules. He left the room so no one could read his preference — and, as the case is taught today, assigned his brother to do nothing but attack every proposal on the table. The discipline of appointing someone to argue against the plan even has an ancient name: the Vatican coined the "Devil's Advocate" in the 11th century to debunk candidates for sainthood.
Here's why this matters in 2026: when you ask AI "what should I do?", you've just recreated the Bay of Pigs room. A confident, knowledgeable voice that's been trained to be agreeable hands you a tidy answer — and quietly buries the objections.
Attio is the AI CRM for high-growth teams.
Run deep web research on prospects. Update your pipeline as you work. Find customers and draft outreach emails. Powered by Universal Context, Attio's intelligence layer, Attio searches, updates, and creates across your data to accelerate your workflow.
Ask more from your CRM.
TLDR: The best decision-makers don't ask AI for advice — they make it argue against them. Below: why "what should I do?" is the wrong question, and a prompt that turns any AI into a red team assigned to kill your decision before reality does.
The yes-man problem
Modern AI is extraordinarily good at sounding sure of itself. Ask it whether to take the job, launch the product, or fire the vendor, and it will give you a clean, plausible, well-organized answer. It will also, more often than not, lean toward whatever you seem to want — because agreeable answers are what it was optimized to produce.
That's fine when the stakes are low. It's dangerous when the decision is expensive, because the one thing a confident answer hides is the case against it. You don't walk away smarter. You walk away more certain — which is the opposite of what a hard decision needs.
The fix is an adversary, not a better answer
Intelligence agencies, the military, and serious corporate strategy teams all use the same tool to fight overconfidence: red teaming. You assign people whose only job is to attack the plan — find the flaw, surface the blind spot, argue the opposite — before you commit. The point isn't to be negative. It's to find the failure while it's still cheap to be wrong.
The good news: you don't need a roomful of skeptics. You need a prompt that forces your AI to stop nodding and start swinging.
How to use it
Don't ask AI to decide. Hand it your decision and assign it to assemble a red team — a panel of personas, each tasked with killing the idea from a different angle (the numbers, the people, the customer, the worst case, the long game). Then have it report what survived the attack and where you're most exposed. Bring it a real decision and it won't tell you what to do — it'll tell you the one question you're avoiding.
Same Prompt · Your Decision What the Red Team Found That They'd Have Missed |
Marketing Director, 80-person SaaS Decision: Sign a $90K/yr contract with a new AI content platform. Killer attack (The CFO): The $90K wasn't the cost — the 200 hours of team retraining and the 6-month switching cost if it underdelivers was. Real first-year cost closer to $150K. Blind spot surfaced: She was buying it to look innovative to the CEO, not because the team had a content bottleneck. |
Solo Consultant Decision: Drop her biggest client to go all-in on a new productized offer. Killer attack (The Pre-Mortem): A year out, the productized offer stalled at 40% of replaced income, and the old client had already hired someone else. No path back. What survived: The offer itself was strong — the timing was the flaw. Run it alongside the client for one quarter first. |
Operations Lead, regional logistics firm Decision: Replace three coordinator roles with an AI scheduling system. Killer attack (The Operator): The coordinators weren't doing scheduling — they were absorbing exceptions the system can't handle. Cut them and every edge case lands on him. One question before committing: What percentage of last month's work was the messy exceptions, not the routine? |
Same prompt. YOUR decision. Make it argue against you. |
The smartest people in the room aren't the ones with the best answer. They're the ones who went looking for the reason they were wrong — and got to it before everyone else did.
So the next time you're about to ask AI what you should do, ask it the better question instead: "Why shouldn't I?"
What does that $70k hire really cost?
Salary is only part of the equation. Local taxes, benefits, and employment costs add up quickly.
About This Newsletter
AI Super Simplified is where busy professionals learn to use artificial intelligence without the noise, hype, or tech-speak. Each issue unpacks one powerful idea and turns it into something you can put to work right away.
From smarter marketing to faster workflows, we show real ways to save hours, boost results, and make AI a genuine edge — not another buzzword.
Get every new issue at AISuperSimplified.com — free, fast, and focused on what actually moves the needle.