Sprinthero
All resources
Business Model Transformation·Playbook

Building an ExO — the ExO 3.0 Skill (v25)

The complete ExO 3.0 skill pack: Intelligence Stack, REWRITE Playbook, Edge Twin & data-plane inversion — with templates and agent specs. Downloadable .skill file.

Attributed to ExO 3.0, the Intelligence Stack and the REWRITE Playbook are the intellectual property of their authors — Salim Ismail, Gary Boomer, Dea Csuba, Augusto Fazioli, Rob Gonda, Vinay Gupta, Charles Klasson, Kent Langley, Tony Manley, Vivek Matthews, Giovanni Morales, Marconi Pereira, Ann Ralston, Gary Ralston, Miguel Angel Rojas and Patrik Sandin. The Miura-Ko autonomy ladder is attributed to Ann Miura-Ko (Floodgate). 'Human capital / token capital' vocabulary follows Satya Nadella.

Download the skill pack

Get the full Building an ExO — the ExO 3.0 Skill (v25) file — import it into Claude, ChatGPT or Cursor and your AI assistant will run the method with you.

Download .skill

What it is

ExO 3.0 is the 2026 successor to the original 2014 Exponential Organizations framework (SCALE/IDEAS). Where ExO 1.0 described what exponential organisations look like, ExO 3.0 is the operating manual for rebuilding an existing firm around AI — in Direct Mode (≤50 people) or Edge Mode (>50 people, with a board-mandated Edge Twin).

This resource is the full 'building-an-exo' skill (v25 / OS Outline v25) packaged as a downloadable .skill file. It bundles the Intelligence Stack, the REWRITE Playbook (DRIVE → SHAPE → Edge Deployment), data-plane inversion, GOVERN/ASSURE guardrails, and a library of templates: Backcasting Canvas, Readiness Score & Score Interpretation Matrix, Task Decomposition Matrix, Capability Registry, Permission Envelope, Agent Blueprint specs, Trusted Evals, Searchable Logs, Granular Rollback and the Human Review Queue.

It is the same pack our coaches run inside the ExO 3.0 Sprint — AI-Native Rebuild. Install it in your own AI assistant (Claude, ChatGPT, Cursor) and it will walk a founder or CEO through the rebuild step by step, using the same language and instruments we use in the room.

When to use it

  • When a founder or CEO asks 'what would we build if we started today?' — and means it.
  • When a board mandates an AI-native rebuild of a legacy business (>50 people → Edge Mode with an Edge Twin).
  • When you need a governed path from pilot AI projects to an agentic operating core.
  • When your data strategy must become portable — make meaning portable before you move platforms (data-plane inversion).
  • When you want templates and agent specifications, not another slide deck about AI.
  • When preparing for — or running — the Sprinthero ExO 3.0 Sprint.

How to use it

  1. 1

    Download the skill file

    Use the download button below to get building-an-exo-skill.skill (a .zip-compatible skill archive, ~180 KB).

  2. 2

    Install it in your AI assistant

    Import the .skill into Claude (Settings → Skills), ChatGPT, or Cursor. The assistant adopts the ExO 3.0 operating system: vocabulary, canvases, gates and templates.

  3. 3

    Run the Readiness Score

    Start with the Readiness Score and Score Interpretation Matrix. The skill will not let you skip the gate — Direct Mode vs Edge Mode is decided here.

  4. 4

    Work the playbook

    DRIVE the destination (MTP-as-protocol, Backcasting Canvas), SHAPE the Intelligence Stack (agents, decision traces, logs you own), then deploy the Edge Twin on one high-stakes workflow.

  5. 5

    Govern as you go

    Apply GOVERN/ASSURE: Trusted Evals, Searchable Logs, Granular Rollback and the Human Review Queue before any agent touches a fiduciary wedge.

Key concepts

ExO 3.0

The 2026 successor to ExO 1.0 (SCALE/IDEAS) — an operating manual for rebuilding a firm around AI, not just describing exponential attributes.

Intelligence Stack

The firm's proprietary AI core: private data, evals, agents and the decision traces that compound learning. Own your logs — the log is the agent.

REWRITE Playbook

The staged transformation sequence — DRIVE (destination), SHAPE (stack design), Edge Deployment (twin build) — executed in waves with gates between them.

Edge Twin

An AI-native twin of one high-stakes legacy workflow, mandated by the board, that mirrors and then out-learns the legacy operation.

Data-plane inversion

Make meaning portable before migrating systems: demote — don't abolish — legacy platforms, and put a decision layer over your data first.

GOVERN / ASSURE

The guardrail set: Trusted Evals, Searchable Logs, Granular Rollback and a Human Review Queue — the four pillars that make agent autonomy auditable.

Common pitfalls

  • Skipping the Readiness Gate and jumping straight to tooling — the organisational immune system will reject the transplant.
  • Treating the Edge Twin as an IT project instead of a board-mandated business rebuild.
  • Letting vendor platforms own your decision traces and logs (the Vendor Shortcut) — you can outsource a task, never your learning process.
  • Scaling agents before Trusted Evals and rollback are in place.
  • Announcing 'AI transformation' as headcount reduction — it kills the Middle 60% you need for the Bridge Curriculum.

Further reading

Want a Sprinthero coach to apply this with your team?

Our coaches use this — and the rest of the Business Model Transformation toolkit — live with leadership teams every week.

Talk to a coach