org-design-in-ai-era

Org Design in the AI Era

Summary

External thinking on how AI is reshaping organisational structure. Common thread: traditional hierarchies persisted for 2,000 years because humans had no alternative for coordinating large groups — AI now replaces the information-routing function of middle management, enabling dynamic, skills-based, fluid org models. The Block model (Dorsey / Botha) and the YC / Diana Hu “AI-native company” frame converge: company must be made queryable to AI, with humans at the edge guiding it rather than in the middle routing it.

Key Points

• The 2,000-year persistence of hierarchy. Roman span-of-control (8→80→480→5,000), Prussian General Staff, railroad org charts, matrix orgs — all variations on hierarchy because coordination at scale demanded it. AI is the first real alternative.
• Middle management was an information-routing layer. Once AI absorbs that function (workforce intelligence, real-time decision support, workflow automation), the structural reason for permanent middle layers weakens. Diana Hu / YC puts it more bluntly: “velocity is only as fast as information flow — every layer of human routing removed is a direct speed gain.”
• What replaces it — dynamic structures:
  • Fluid team formation around projects, not departments
  • Skills-based resource allocation, not title-based
  • Decentralised decision authority
  • Continuous feedback loops, real-time adjustment
• Block’s concrete model (Dorsey / Botha). Four components:
  1. Capabilities — atomic primitives without UIs
  2. World Models — dual representation of company ops + customer behaviour
  3. Intelligence Layer — composes capabilities into personalised solutions
  4. Interfaces — apps, hardware
• Three roles replacing hierarchy (Block / Dorsey, echoed by Diana Hu / YC):
  1. IC / Builder-Operator — directly makes and runs things. Not limited to engineers; everyone (ops, support, sales) builds. Meetings get working prototypes, not pitch decks.
  2. DRI (Directly Responsible Individual) — one person, one outcome, focused on strategy and customer outcomes. Not a classic manager — no hiding.
  3. Player-coach / AI Founder type — builds, coaches, leads by example. Founder must be at the forefront showing capability gains. Do not delegate AI strategy.
• Closed loops, not open loops (Diana Hu / YC). Old companies ran as open loops — decisions executed, outcomes not systematically fed back. AI-native companies run every important process as a closed loop: capture info → feed it back to intelligent systems → self-improve over time. Closed loops are the substrate for correctness and stability.
• Make the company queryable. Every important action should produce an artifact the intelligence layer can learn from. Concrete moves: record meetings with AI notetakers, minimise DMs and emails, embed agents throughout comms channels, build dashboards for everything (revenue, sales, engineering, hiring, ops). Sprint planning agent example: with access to Linear + Slack + Pylon + GitHub + Notion + sales calls + standup recordings, it analyses what was actually shipped vs customer needs and proposes more predictable sprint plans. Claim: 10x output, half the sprint time.
• Startup advantage vs incumbents (Diana Hu / YC). Startups can design org, workflows, culture around AI from day one. Incumbents must maintain live product while unwinding years of SOPs. Some incumbents succeed via small internal skunkworks (e.g. Mutiny) building AI-native systems separate from the core. This is the Intercom destruction model at the org layer — same shape, different lens.
• Token-maxing, not headcount. Be willing to run an uncomfortably high API bill — it’s replacing far more expensive headcount. One person with AI tools = what used to require a team. Implication for org sizing across engineering, design, HR, admin.
• AI as “connective tissue.” Makes dynamic coordination viable where it previously collapsed under overhead.
• Implementation friction is real. Cultural resistance, leadership redefinition, coordination complexity, data dependency, governance clarity. Not a tech-only change.

Implications for next chapter

• The “Head of Engineering” / “VP Engineering” title is itself a hierarchical artefact. Worth considering what the equivalent role looks like in a DRI / player-coach world.
• Skills-based deployment over role-based assignment aligns with the fractional / portfolio model already dominating the Melbourne exec market (see Melbourne Tech Leadership Market).
• The Block model is the most concrete operational blueprint surfaced so far — capabilities + world models + intelligence layer + interfaces. Worth a deeper read against own thinking on agentic engineering.

Sources

• 2026-05-12 — From Hierarchy to Intelligence (Dorsey & Botha, Block)
• 2026-05-12 — The End of Org Charts (techrseries)
• YC) — closed loops, queryable company, three archetypes, token-maxing, startup-vs-incumbent

Last Updated

2026-05-19