Most important take away

AI isn’t just automating individual tasks within existing org structures — it’s eliminating the coordination overhead (meetings, PRDs, status updates, handoffs) that comprises 60% of knowledge work. The real opportunity is that humans get to spend more time on high-value work like product vision, architecture thinking, and customer relationships instead of coordination busywork.

Chapter Summaries

The Wrong Questions About AI and Jobs

The standard approach of decomposing roles into tasks and asking “which can AI do?” fundamentally misunderstands the situation. Organizations aren’t fixed structures — most knowledge work is coordination overhead, not value creation.

The Coordination Tax

Research shows 57-60% of knowledge worker time goes to “work about work” — meetings, documents, handoffs. These exist because the execution layer is human, not because they’re inherently valuable. The value is the shipped product; everything else is the cost of producing it with humans.

The Org Is Moving to Code

When agent harnesses can go from insight to code directly, translation layers between people get deleted. PRDs, sprint planning, status meetings, and design handoffs evaporate — not because they’re automated, but because the problems they solved no longer exist.

The Flywheel Effect

Removing coordination roles makes remaining work more verifiable by agents, which enables agents to do more, which reduces coordination further. This creates a compounding loop that pushes organizations toward humans managing agent teams.

What Survives — The Hard Skills

Product vision, brand thinking, genuine customer relationships, systems architecture, and designing agentic systems themselves. These are the highest-value skills that coordination overhead has been starving us of time to do properly.

Block/Dorsey — The Real Story

Dorsey framed layoffs as “AI automation” but was really correcting for over-hiring. The standard “automate tasks and stop” narrative understates AI’s organizational impact by 2-3x.

Agency and Ramp

The two qualities that distinguish people thriving in this transition: high agency (“I can figure this out”) and high ramp (curiosity-driven rapid learning).

Summary

Actionable Insights

• Audit your coordination tax: Review your calendar and count hours spent transferring information vs. creating value. If you’re at the typical 60/40 split, those coordination hours are the ones AI will eliminate first.
• Get hands-on with agentic tools now: The people thriving have two qualities — high agency and high ramp (rapid learning through curiosity). Start iterating directly with AI coding tools to experience the “translation layer deletion” firsthand.
• Reframe your role around value creation: Identify which parts of your job are genuine judgment and vision vs. coordination overhead that feels like judgment. Double down on the former.
• Think structurally, not task-by-task: Don’t ask “which of my tasks can AI do?” Ask “if coordination overhead disappeared, what would my role look like?”

Career Advice

• The most important operational competency in tech within a few years will be designing and managing agentic systems — task decomposition, verification criteria, debugging failure modes.
• Organizations will shift from 20% of people touching the product to 100% — position yourself to be close to the product, not the coordination layer.
• Coordination-heavy roles (project management, sprint planning, status reporting) are most at risk — not because they’re automated but because the problems they solve will stop existing.