Read the Room
We expect others to respond to information they don't have. Turns out, so does AI. The problem is older than both of us.
Several months ago, I received a piece of feedback that totally caught me off guard.
I was told I needed to do a better job of "reading the room."
As I asked for clarification, I quickly learned the feedback was about meetings I hadn't attended and conversations I hadn't been part of. The concerns being referenced had never been shared with me, including by the person delivering the feedback.
I sat with that for a while.
My initial reaction was frustration. How can anyone reasonably be expected to respond to information they don't have? But the longer I turned it over, the more I realized this wasn't really a story about one awkward feedback conversation. It was a story about something much more universal and much more human.
How often have you gotten frustrated with someone for not doing something you expected of them, yet never actually communicated that expectation?
"Of course they should know." "Anybody would know." "It's just common sense." "I shouldn't have to tell them."
We've all thought some version of this. If we're honest, we've all said it too.
We get frustrated at our partners for not living up to expectations we never shared. We expect our children to follow norms we never explained. We expect coworkers to respond to information they've never heard, bosses to intuit personal situations we've never disclosed, and direct reports to deliver work to standards we never clearly defined.
The ability to read minds remains remarkably rare, despite how often we seem to expect it from one another.
And here's where it gets interesting: we do the exact same thing to our AI tools.
We open a chat window, type a half-formed question, and then feel vaguely disappointed when the output misses the mark. We didn't share the context. We didn't explain the constraints. We didn't define what success actually looks like. And then we conclude the technology isn't quite there yet — when really, we just set it up the same way we set each other up. We built a perfect little trap and walked straight into it ourselves.
The problem isn't the intelligence. It's the missing information.
In the absence of information, people don't stop making sense of the world. They simply start making things up.
And so does AI.
What we call hallucination—when a model confidently produces something plausible but false—isn't really a malfunction. It's the system doing exactly what we do when the context runs out: constructing a coherent narrative from insufficient input. Filling the gap with something that feels like it fits.
We've been doing this forever.
Think about the last time rumors of a re-org circulated at your company. You probably knew someone… maybe you were that someone… who immediately assumed it meant their job was on the line. Humans are extraordinary storytellers and even better story believers. In many ways it's our superpower. It's also our greatest vulnerability.
When we suspect that someone knows something we don't, we assume there's a reason. We start writing the story ourselves. Why else would so many people be losing faith in institutions? Why else do flat earth theories find audiences? Why else does every generation produce its own constellation of conspiracy thinking?
You don't need many experiences of a leader withholding information before you start assuming it must be intentional. And it's a short, surprisingly well-paved road from "my boss didn't tell me what they needed" to "my boss is out to get me."
From classified government programs that turned out to be real to moon landing theories that almost certainly aren't, the distance between reasonable suspicion and full-blown conspiracy is shorter than most of us would like to admit.
We have always created stories to explain the gaps in our reality. Nearly every ancient civilization independently invented gods who controlled weather, seasons, oceans, and fate. The ship didn't just hit a storm. The harvest didn't just fail. Something, someone, must be responsible. Something must explain this.
The mechanism hasn't changed. Only the gaps have.
This is why I believe so firmly in transparency. In making the invisible visible, whenever and wherever it's reasonable to do so.
Yes, there are times when information genuinely can't be shared. Legal constraints are real. Timing matters. Some things aren't mine to disclose. I understand that.
But when we minimize those occasions, and when, once information can finally be shared, the people who were kept in the dark can understand why, we stop setting each other up to reach for explanations that aren't there.
The same principle applies to how we work with AI. The more context you provide, the less it has to invent. Clear constraints, explicit goals, relevant background… these aren't just prompting best practices. They're the same things that make human collaboration work. Shared context is the foundation of trust, whether you're working with a person or a model.
When we all know, and we all know that we all know, we communicate more clearly. More honestly. More effectively. When we operate from a place of transparency and shared understanding, we're more comfortable, more confident, and more capable of building something that actually matters together.
Because when people can't see the room, they don't stop trying to orient themselves. They simply start looking for windows.
The room doesn't read itself.
But it gets a lot easier to navigate when everyone in it has the same map.
Ryan Yepsen
Ryan writes about enterprise AI, customer reality, systems thinking, and the operating conditions that make meaningful execution either possible or painful.