Jack Dorsey Says AI Will Replace Half Your Team. But Will It Replace Your Meetings?

When Jack Dorsey announced in late February that Block was cutting 4,000 employees — 40% of its workforce — because “intelligence tools have changed what it means to build and run a company,” the tech world erupted in debate.

Can AI really replace that many knowledge workers? Is this the beginning of the “AI jobpocalypse” that Microsoft AI chief Mustafa Suleyman warned about? Or is AI just convenient cover for old-fashioned cost-cutting after pandemic-era overhiring?

Those are important questions. But there’s a different question that matters for everyone still employed in the AI era: if AI is shrinking teams, how should those smaller teams communicate?

Because here’s what nobody in the AI discourse seems to be talking about: AI is very good at automating tasks. It’s not good at deciding which tasks matter. And deciding which tasks matter is exactly what meetings are supposed to do.

What AI Can Actually Replace

To understand what AI means for workplace communication, it helps to separate what AI does well from what it doesn’t.

AI excels at routine cognitive work: drafting emails, summarizing documents, generating code, processing data, creating reports, and answering well-defined questions. These are the tasks that Block’s internal AI tool, Goose, has apparently automated to the point where the company’s CFO reported a 40% increase in production code shipped per engineer.

AI also excels at meeting administration: transcribing conversations, generating summaries, extracting action items, and distributing notes. Tools like Otter, Fireflies, and Microsoft Copilot have made post-meeting documentation largely automatic. Microsoft reports over 100 million monthly active Copilot users, many of whom rely on AI-generated meeting summaries as a core workflow.

What AI has replaced, then, is a significant portion of the work that happened before and after meetings: the preparation, the note-taking, the follow-up documentation. It’s also replaced much of the work that meetings were nominally about: status updates (which AI dashboards can now generate automatically), information synthesis (which AI can compile faster than any human), and routine coordination (which AI scheduling tools can handle).

What AI Cannot Replace

But the reason meetings exist isn’t primarily to share information or coordinate logistics. Those activities have always been poorly suited to meetings — they were just forced into meeting format because better alternatives didn’t exist or weren’t widely adopted.

The actual value of meetings lies in activities that AI cannot replicate:

Judgment under ambiguity. When a team faces a decision with incomplete information, competing priorities, and no clear right answer, they need human judgment — informed by experience, values, and organizational context — to choose a direction. AI can present options and analyze trade-offs, but it cannot make the value judgment about which trade-off to accept.

Creative collision. The most valuable brainstorming happens when one person’s half-formed idea sparks a completely different idea in someone else’s mind. This combinatorial creativity emerges from the unpredictable dynamics of human conversation and depends on empathy, intuition, and shared context that AI doesn’t possess.

Emotional navigation. Delivering difficult feedback, navigating disagreements, building trust, and aligning people around a shared vision all require emotional intelligence that AI doesn’t have. These conversations are fundamentally about human relationships, not information exchange.

Stakeholder alignment. Getting five people to commit to a direction — especially when they have different priorities, different information, and different risk tolerances — requires real-time negotiation. AI can facilitate this by providing data, but the alignment itself has to happen between humans.

These are precisely the activities where meetings are most valuable. And they’re the activities that become more important, not less, as teams shrink and each person’s decisions carry more weight.

The Paradox of Smaller, AI-Augmented Teams

Here’s the irony of Dorsey’s vision: smaller teams using AI tools actually need better communication decisions, not fewer communication touchpoints. When a 10-person team becomes a 6-person team, each person covers more ground. The overlap zones — where two people’s responsibilities intersect and alignment is needed — don’t shrink proportionally. They often expand.

In a 10-person team, the engineer and the designer might overlap on two projects. In a 6-person team where each person handles more, that same engineer and designer might overlap on four or five. The coordination surface grows even as the headcount shrinks.

If those coordination conversations happen through unfocused, unnecessary meetings — the same status updates and FYI sessions that plagued the larger team — the smaller team drowns faster. There are fewer people to absorb the meeting load, so each unnecessary meeting consumes a larger percentage of each person’s available time.

But if those conversations happen through the right channel — async messages for quick questions, shared documents for feedback, short focused meetings only for decisions and creative work — the smaller team can operate with the speed and clarity that Dorsey is betting on.

The format decision — meeting or not — isn’t a luxury for small teams. It’s a survival skill.

A New Communication Framework for the AI Era

The AI revolution doesn’t eliminate the need for communication format decisions. It sharpens them. Here’s how the framework updates for a world where AI handles routine work and humans focus on judgment, creativity, and alignment.

Status updates and information sharing: Never a meeting. This was already true before AI. Now it’s even more clear. AI can generate dashboards, compile progress reports, and summarize project states automatically. If the information can be read, it should be read — not presented in a meeting.

Routine questions and coordination: Async message or AI-assisted. Questions like “What’s the status of the API integration?” or “When will the design review be ready?” can be answered by AI tools that pull from project management systems, or by a quick async message. No meeting needed.

Feedback on proposals or documents: Shared document, then optional meeting. Share the proposal. Let people comment asynchronously. Use AI to synthesize the feedback. Then hold a short meeting only if the feedback reveals genuine disagreements that need live resolution.

Decisions with real stakes: Short, focused meeting. When the team needs to commit to a direction that affects multiple workstreams, a meeting is justified — but only with the people who have decision authority, only with an agenda, and only for 25-30 minutes. AI can prepare the decision brief, present the options, and document the outcome. The meeting exists solely for the human judgment call.

Creative brainstorming: Meeting with async pre-work. Have participants submit initial ideas asynchronously (AI can help organize and cluster them). Then hold a focused session to build on the most promising ideas. The meeting is shorter and more productive because the “cold start” problem is already solved.

Relationship building and sensitive conversations: Always a meeting. AI cannot build trust, deliver empathy, or navigate emotional complexity. One-on-ones, mentorship conversations, and team connection moments should be protected as meetings — and treated as some of the highest-value time on the calendar.

The Real Question Isn’t Whether AI Replaces Jobs

Jack Dorsey may be right that AI will reshape the workforce dramatically. Or he may be using AI as cover for a correction to pandemic-era overhiring, as Bloomberg and others have suggested. Either way, the remaining employees at Block and at every other company adopting AI tools face the same practical question: how should I communicate this?

The answer has never been more important. In a world where routine work is automated and human attention is the scarcest resource, every unnecessary meeting is more costly than ever. Every communication forced into the wrong format — a meeting that should be a document, a document that should be a message, a message that should be a meeting — wastes the very human judgment that AI can’t replace.

Before scheduling your next meeting, take 30 seconds to ask: does this need to be a live conversation, or can it be handled another way? That question is free, it takes less time than creating a calendar invite, and in the AI era, it might be the most valuable decision you make all day.