Why meetings still suck (and how AI might finally save us) 🤖

Anyone who knows me knows I love productivity and efficiency tools. My current favourite for meetings is Granola. It’s smart, simple, and actually helpful. But let’s be honest, there are a lot of tools like this right now, and most of them won’t survive.

In a world where companies are under pressure to do more with less, or work smarter with what they already have, the real winners in this space will need to do something that feels counterintuitive: Help us have fewer, better meetings.

The silent killer of work

We’ve normalised a world where meetings are the default setting for getting anything done, and a lot of them are garbage.

This is not just my view, it’s backed by data (sources at the end):

• Only 37% of meetings have an agenda
• 64% of recurring and 60% of one-off meetings happen without one
• Executives consider 67% of meetings a failure
• Individual contributors often spend half their week in meetings they didn’t ask for and don’t understand

Quotes from personal sources:

“The amount of hours I spend in useless meetings is insane.”
—Product Leader, Enterprise

“I slack the organiser just to understand what it’s about.”
—IC PM, SMB

“We waste a lot of time scene-setting because the purpose wasn’t clear from the start.”
—Marketing Operator, Scale-up

We’re 25 years into the 21st century and still wasting huge amounts of time.

Everyone’s building the same AI. That’s a problem.

There are at least two dozen products that transcribe your meetings, summarise what happened, and tag follow-ups. They wrap it in a clean UI, store it somewhere, and call it productivity.

The result? They just make it easier to have more of them 🤦‍♂️.

But none of them fix the actual problem:
The meeting itself is broken.

Before AI can improve our meetings, we need to talk about how it thinks, and why context is everything.

Better inputs = Better outputs

Language models don’t just need data. They need context.

A study on how context affects LLM factual predictions showed that adding relevant information significantly improves model accuracy. AI isn’t magic. It just does better work when it knows what it’s supposed to focus on.

If your AI knows the meeting is a sales call, it can spotlight objections, buyer intent, and next steps. If it’s a design review, it’ll highlight feedback and action points.

But if there’s no title, no agenda, and no stated goal? AI is just guessing.

The missing link: Pre-meeting context

Everyone’s excited about agentic AI, tools that can take your meeting, distill the actions, and just…do the work. But here’s the catch: if the meeting was a mess, the AI output will be too.

Poorly defined goals, vague decisions, and bloated invites don’t magically translate into high-quality execution. They translate into AI agents doing mediocre things with misplaced confidence.

If we want agentic AI to actually be useful, we need to fix the upstream inputs. That means better meetings, or better yet, fewer of them.

Imagine your meeting assistant nudging you with:

• “What’s the goal?”
• “Can this be async?”
• “Why are these eight people here?”

Not just a note-taker, more like a gatekeeper. A pre-check before anyone hits ‘Join.’

The result? Meetings that are sharper. Context that’s crisp, and downstream agents that can actually do their job, because they’re not starting from chaos.

Want smart AI? Start with smarter meetings.

Final thoughts

AI tools are about to flood your calendar with summaries, tags, and follow-ups. But if the meeting itself was aimless? All you’re doing is polishing a turd.

The real frontier isn’t just post-processing. It’s upstream context.

If we want AI to take admin off our plates so we can focus on what matters, we need to feed it clarity before anyone clicks ‘Join.’

The smartest question your AI can ask might just be the one we avoid the most:

“Why are we meeting at all?”