This one has bugged me for a while:
How is it that we can all leave the same meeting (someone even took notes!) and, two days later, we’re all operating on completely different memories of what happened?
In some cases, I’ve seen it happen just hours after people leave the room.
I’ve seen it happen even when meetings are led by a skilled facilitator and a scribe. No matter what, memory gets in the way. Nobody seems to be immune. Why does this happen? And perhaps more importantly, what can we do about it?
The first question, fortunately, has some scientific answers. Infuriatingly, the question of what can be done is decidedly unresolved.
“And I think AI, as promising as it is, could actually make this all worse.”
We hear what we fear
Memory isn’t a passive process. It’s active, shaped by where we place our attention—and where we don’t. And our attention is rarely neutral.
By the time a meeting begins, we’ve already made a dozen micro-calculations. What am I walking into? Who’s in the room? What’s at stake for me here? Am I expected to defend something, decide something, deliver something?
Those subconscious scripts influence what we tune into. When someone says, “We may need to revisit the timeline,” one person hears opportunity, another hears warning bells. One person hears flexibility, another hears failure.
Power complicates it further. The higher the stakes, the more likely people are to filter for what keeps them safe. Safe from blame, from being wrong, from being associated with risk.
So what gets remembered isn’t necessarily what was said. It’s what we felt. What we feared. What we believed mattered to our own standing. And that becomes our “truth” about the meeting.
The problem is, no two people walk in with the same filters. So no two people walk out with the same story.
The forgetting curve doesn’t help
It’s not just that our filters distort what we hear in real time—it’s that even the parts we do hear often fade quickly.
A large-scale meta-study (“A New Look at Memory Retention and Forgetting”, Journal of Experimental Psychology) found that:
- When we take in new information, much of it is lost in the first few seconds to minutes, as it fades from working memory.
- Over the next several hours (what the authors call early long-term memory), the rate of forgetting slows but still continues.
- After about a day, remaining memories begin to stabilize and consolidate, becoming easier to recall over the following days and weeks.
This helps explain why we can feel so sure we remember a meeting accurately—while often forgetting that what we remember is incomplete, shaped by what we paid attention to in the first place.
Importantly, most memory research is done in controlled lab settings. In real life—with back-to-back meetings, multitasking, and constant distractions—our memory likely degrades even faster.
The failure of the meeting notes
Meeting notes are supposed to be our safety net. Our shared record. The thing we can point to when memories diverge.
And yet, more often than not, they’re part of the problem.
We tend to document meetings like they’re courtroom proceedings:
- What was said.
- What was decided.
- Who’s doing what.
Clean, linear, and devoid of the ambiguity that defined the actual conversation.
What’s usually missing? The things people didn’t say out loud, but communicated in body language, in silence, in tone. We don’t write down, “There was tension around the budget conversation,” or “Lisa seemed disengaged when the marketing plan came up.”
And let’s not forget: the person taking notes is also a participant. They’re subject to the same selective attention, the same cognitive filters, the same forgetting curve. Even if they’re trying to be objective, they can’t help but emphasize what they heard—or thought was important. They, too, are constructing a version of the meeting, not capturing a transcript.
That means meeting notes don’t just reflect the meeting—they shape the team’s memory of it. They codify one participant’s interpretation as the record. That’s a lot of power, and it’s rarely acknowledged.
So when misalignment surfaces later, we turn to the notes like a referee. But the notes are just as filtered as our memories. They document the surface of the meeting, not its depth. And by treating them as gospel, we risk papering over the real issues we failed to resolve.
AI could help…maybe.
AI is quickly becoming the new scribe in the room—listening, transcribing, summarizing, organizing. And on the surface, that seems like a huge improvement. No more relying on scattered notes or fuzzy recollections. Everything is recorded. Everything is searchable.
Not to mention, we’re less likely to miss action items, because the models are trained to listen for them. And, if there are any disagreements, we can even pull a transcript of the meeting and search for a common version of the truth.
Or can we?
AI is great at processing speech. But meetings aren’t just about speech. They’re about subtext, timing, discomfort, intuition. They’re about how long we stayed with an idea (or didn’t). A 20 minute conversation could get reduced to the two action items that stemmed from it, while what everyone actually remembers and takes action on is how they FELT during the conversation.
AI is being trained to recognize “sentiment,” which could help.
Here’s what AI can’t do:
- Feel, in a body, the personal stakes of a conversation.
- Put a meeting in the context of the “hallway conversations” that led up to it
- Force people to read the notes later
So yes, AI can help us capture the “what.” But I worry that it encourages us to gloss over the “how” and the “why.” And that’s why…
AI could make all of this a lot worse.
The danger isn’t just in what AI misses. It’s in what we come to believe because of what it captures.
There’s a seductiveness to clean transcripts and tidy summaries. It feels definitive. It feels factual. But that sense of certainty can be dangerous when it’s built on incomplete foundations.
What happens when someone reads an AI-generated summary and says, “That’s not what I experienced”? Do we assume they’re wrong? Misremembering? Being difficult? Or do we ask what their version reveals that the AI missed?
I worry we’ll stop asking. That we’ll start deferring to the AI’s record as the one true version. That we’ll dismiss disagreement as noise, instead of signal. That we’ll build strategies, launch projects, and make decisions on foundations that looked clear—but weren’t.
And there’s another problem:
We’re probably not far off from AI becoming a meeting facilitator, not just a scribe. That means we could all be subject to AI’s biases (which we just call “programming.”)
- If AI is optimized for clarity, we risk oversimplification.
- If AI is optimized for productivity, we risk glossing over a messy conversation that leads to a breakthrough.
- If AI is optimized for harmony (i.e., making sure everyone speaks), we risk disrupting important flow.
What if misremembering is a feature, not a bug?
It’s easy to think of mismatched memories as a mistake. A failure of listening. A lack of clarity. But maybe we’re looking at it the wrong way.
Maybe different recollections are a sign that people were paying attention to different things—things that matter in different ways to different people.
Maybe those differences in memory are telling us where the team’s understanding is still forming. Where alignment is still thin. Where someone heard something that didn’t register for the rest of us—but should have.
Instead of assuming someone’s memory is flawed, we could get curious. “What stood out to you?” “What felt unresolved?” “What do you think we missed?”
Memory divergence isn’t a bug to squash. It’s a clue to follow.
Maybe the goal isn’t to remember meetings the same way. It’s to notice where—and why—we don’t.
So what do we do with all this?
I don’t think the answer is to remember everything the same way—or even to try. But I do think we need to get better at surfacing the differences before they harden into confusion or resentment.
That might look like:
- Normalizing end-of-meeting check-ins that aren’t just, “Any questions?” but “What are you walking away with?”
- Giving language to our level of certainty: “Are we aligned? Or just adjacent?”
- Acknowledging that what feels clear to one person might still feel unresolved to another—and that both experiences are valid.
And maybe most of all, we need to rethink what we’re using our meetings for.
Are we really gathering to get alignment? Or are we gathering to begin it?
If it’s the latter, then memory divergence isn’t a flaw in the system. It’s a signal that there’s more work to do—and that we still have time to do it.