The most dangerous moment is when everyone in the room agrees. A field-notes case for why agreement isn’t evidence – and what to do when you can’t change who’s in the room.
Years ago I built a law firm that billed like a phone company.
Lawyers have billed by the hour since roughly the invention of the hour. I charged a flat monthly subscription instead, the way your carrier sells you a plan. It worked absurdly well – clients loved the predictability, I loved each client paying $6,000 – $25,000 per month, and the competition couldn’t work out why the firm charging a flat monthly fee kept on landing clients.
The part that matters: I didn’t invent anything. I lifted a pricing model out of telecom and dropped it into law. The idea was powerful because it was foreign – it came from an industry that shared no playbook, no assumptions, and none of the same blind spots as the one I was disrupting.
An idea formed somewhere else brings answers your field can’t see, because your field never had to look.
Let’s set aside the smug, cringy, platitudes of “diversity of thought” and talk about how we can engineer smart decisions:
Agreement is only information when the minds are independent.
The Agreement Trap
The hard question is this: when everyone in a room agrees, you can’t tell from inside that room whether each person reached that answer because it’s the right answer, or because they were never going to disagree in the first place. The two feel identical, and let’s face it, it feels good to agree.
The cascade of “yes” leaves out whether the deck was stacked towards agreement in that decision. Everyone thought the same way about the same problem, so agreement isn’t signal, it’s noise.
In other words agreement is ambiguous.
Another way to think about it: you could be in a room full of unwitting yes-men. They’re agreeing because they were selected to be agreeable, not because they actually brought relevant independence to the table.
Same Training, Same Blind Spot
A decade ago, my dad had back pain and was told by two spinal surgeons that he needed spinal surgery. He didn’t, as the third opinion said – bad knees threw off his walk and hurt his back. Two knee replacements fixed the biomechanics that were causing back pain.
Two surgeons said back surgery, because they were back surgeons. It took different training to see the actual cause, and the actual fix.
I use AI. A lot. And one of the things that it’s really good at is giving feedback. But similarly, that’s one perspective. Helpful that it’s not mine, but can’t take it as gospel. Now I ask different AIs, each with different lineages, the same question and then synthesize – and it finds all kinds of things that a 1:1 never surfaced. Different minds, same thing.
Here’s what I found: whether it’s a management team, doctor’s advice, or AIs, the same principles hold:
- A single reviewer has blind spots.
- Multiple reviewers with the same background are more likely to converge. Two opinions doesn’t mean it’s more likely to be right, if both opinions are from the same knowledge base.
- Sometimes the most valuable opinion is the outlier that the consensus never even spotted.
That last one I implemented thanks to a movie that scarred my daughter for life. In World War Z, I got introduced to the idea of “The 10th Man” – when everyone agrees, it’s the one person designated to take the opposite stance and steelman the decision with the best argument against consensus. If it’s good enough for spotting zombie threats, good enough for me.
Who Holds The Pen Matters
But wait, there’s more. And it’s the last, important link.
If you’re the decider and you share the blind spot of the others in the room … You could be part of the problem. In my CEO-days, this was me discounting the VP Sales saying it wasn’t possible to close that many deals in the quarter. I shared the blind optimism of product and marketing, customers would be crazy not to want this…
I was grading my own paper. I weighted the decision to what I wanted to hear, because I was a marketer at heart. Worse, when we missed the quarter it didn’t even register that the lone dissent from last quarter was the signal that really mattered, because it felt pretty much like consensus.
Productive Disagreement
So brass tacks, how do we make this real?
- Hiring: if you feel like the new hire “gets you” right away, it should be a yellow flag – not green. We usually hire people we like. That’s nice and all, except that good decisions are not made by agreeableness.
- Rotate the Dissenter: if you’ve got an existing team, appoint a rotating role of dissenter. When everyone agrees, the dissenter’s job is to disagree and provide the strongest arguments why. Not as good as a truly independent mind, but if you’re not hiring new folks it’s a good patch.
- Beware the Anchor: often the loudest voice in the room means others whose personalities are less forceful stay silent. Consider adopting a playbook from Amazon: when making a decision, make people write down what they recommend, why, and the pros/cons. Take it out of live exchange and you’ll see the wallflowers blossom.
If you take nothing else from this, walk away with a nagging feeling that whenever people agree, that’s a sign that you’re not getting the information you need to make the best decision.
For the technically inclined: the AI decision tool above is real – it takes a hard question, fans it out to several frontier AI models from different labs on maximum reasoning, then has a separate model synthesize their critiques into one map, precisely because models from the same family share the same blind spots. If that’s useful to you, the code is totally free here: https://github.com/mindheadllc/adversarial-review
