Meetings Aren't So Bad After All

Most knowledge workers have strong opinions about meetings. They're where real work goes to die. They're the price you pay for having colleagues.

But the meetings we hate most are simply bad meetings—the ones with no agenda, no decision, no outcome. The ones where you sit for an hour, talk in circles, and leave with the same confusion you walked in with.

Good meetings are different. A good meeting is a transition. People arrive with different information, different priorities, different assumptions. They leave aligned, because the meeting itself surfaced what needed to be considered and forced a resolution.

That isn't a waste of time. That's a coordination technology.

Here's the thing: that same good meeting / bad meeting distinction applies to AI work. There's a version that feels productive but goes nowhere—endless drafts, circular refinement, starting fresh every session. And there's also a version that leads to concrete outcomes. The difference isn't the tool. It's whether you're running a good meeting or a bad one.

How Good Meetings Work

The meeting itself is just the visible part. Most of the work happens before and after.

Before the meeting, work happens in parallel. People gather information, form positions, develop options—separately, asynchronously, in their own contexts. This is being efficient, each person working without requiring everyone else's attention.

This then converges in the meeting itself. Your assumptions conflict with someone else's constraints. The question you thought was settled is actually still open. Things move from being open to being defined and settled.

After the meeting, decisions persist. Someone writes them down, not just as a transcript of who said what, but decisions, action items, priorities that provide context for future actions.

Parallel exploration, then convergence, then binding. That loop is what meetings actually are. Not the calendar invite. The coordination around it.

What AI Changes

AI makes exploration cheap—infinite, really. That's genuinely powerful. It's also exactly what makes AI dangerous to use badly.

The trap is that AI's strengths become weaknesses when nothing constrains them. Infinite optionality feels like freedom, but it's also a way to never decide.

You've seen these failure modes before—they're the failure modes of bad meetings. But AI finds new ways into the old traps. A colleague would get tired, or push back, or have something else to do. AI will keep generating options and exploring tangents until you tell it to stop. The meeting that couldn't run forever with humans can run forever now.

Decision paralysis disguised as research. Ahead of a good meeting, people work separately, but toward a shared question. With AI, you can scale that up almost without limit. Explore this angle. Now this one. Now a third. It feels productive—more input should help, right? But each new thread also makes convergence on just one harder. At some point, more input actually starts making the decision less clear. This is the AI version of the meeting where everyone prepares so many slides that the meeting can never get to the decision.

Exploration without a deadline. In a good meeting, there's a limit on the clock. At the end of the hour, we've read whatever input we can and surfaced whatever issues there are, and now we have put our pencils down and decide, ready or not. That constraint forces decisions, and is force on us by the limits of human availability and scheduling. With AI, there's no external push for that fixed point. You can push further, add nuance, consider one more angle. A human collaborator would eventually say "we need to decide." If nothing else, they need to go home. AI can refine forever. This is the meeting that runs over, then over again, then gets a follow-up scheduled.

Progress without persistence. In a good meeting, someone takes notes. With AI, the chat just ends. You made real progress—then the session closes and the context evaporates. Next time, you start fresh, re-explaining things you already figured out. This is the meeting where everyone leaves with a different understanding of what was decided.

AI made exploration cheap. That means coordination—the thing meetings exist for—becomes the bottleneck. The discipline that meeting structure, and even just human limits like patience used to impose from outside, you have to supply yourself.

And this only intensifies as you delegate more. The better AI gets, the more your job becomes supervision. You're not doing less work. You're doing different work.

Running Your Own Meeting

What would it look like to treat your AI work like a well-run meeting?

The Agenda. Think of it less as a schedule and more as a document that names what's actually being decided and what you'd need to decide it well. AI is good at this part—exploring options, surfacing trade-offs, clarifying what's really at stake—but it needs that structure to be able to explore independently. It is filling in the gaps that exist in the agenda that leads to your decision. Skip this step, and you're in the meeting where everyone shows up wondering what they're doing there.

The Decision. When exploration is bounded by an agenda, the decision moment gets cleaner. You're not still generating options—you're picking among them. You call the question: "Okay, we've seen four approaches. I'm going with this one, for these reasons. We're not revisiting it today." Without this, you're in the meeting that ends with "let's schedule a follow-up to continue the discussion."

The Minutes. After the session, you write something down. Not just the decision, but the thinking behind it—context for future decisions. What you decided, what you explicitly rejected, what's still open. A few sentences that both the AI and your future self can load to continue moving. Without this, you're in the meeting where everyone leaves with a different understanding of what was agreed.

This isn't an elaborate process. A few minutes upfront to name the decision, decide what inputs impact that decision, and capturing the reusable lessons from the process.

Why This Is Hard

Think about how many meetings you've attended versus how many you've run. Ten people in a room, one person whose meeting this is. Ninety percent of your meeting experience is as a participant—someone else worries about the agenda, the clock, the documentation.

And participants get to ignore a lot. You don't need to track whether the decision actually got made or just felt made. You don't need to worry about what gets written down. Someone else handles that. Your job is to show up with your piece, contribute when called on, and leave when it's over.

Most professionals have attended hundreds of meetings but chaired a many fewer. We're fluent in participating—having opinions, responding to prompts, evaluating options. The management layer is someone else's job. We never learned it because we never had to.

With AI, there's no someone else. You're the only one in the room, which makes it your meeting whether like it or not. The participants are, by and large, the AI and its responses. But the interface doesn't feel that way. It feels like active discussion—you prompt, it responds, you react. Work is happening. It's just that no one's watching the clock, or noticing the loops, or writing anything down.

AI didn't eliminate the need for meetings. It made you the chair of every one. The question is whether you're running the good kind—with agenda, decision, and minutes—or the kind we all learned to skip.