Manifesto: Organisational Capability and the Shared Commons

1. What We Want to Do

We want to design and implement structural prompting commons and curatorial platforms for verified facts — turning individual AI interactions into shared, verifiable organisational capability rather than allowing them to disappear into private chat histories. This means organisations deliberately protecting capacity for narrated judgment: the explicit practice of documenting why decisions were made, how AI was prompted, and how outputs were verified — rather than immediately rushing to the next execution task. It also means ensuring that organisations, not platforms, own the prompting patterns, context definitions, and validated knowledge that make their AI work effective.

We recommend organisations protect a minimum of 15% of team capacity for this narrated, curatorial work. This is not overhead — it is the investment required to match the scale of the shift. The faster an organisation adopts AI, the larger this investment needs to be.

Critically, narrated judgment only becomes genuine — rather than a compliance exercise — when quality is elevated as a primary organisational metric. The human architecture of quality governance must be given real teeth: defined standards, visible measurement, and meaningful accountability. Without this, narration satisfies a requirement and transfers nothing. With it, the commons becomes the mechanism through which quality is maintained and defended at scale.

2. Why

The competitive pressure toward maximum automation creates a systemic race to the bottom. Individual practitioners use AI to execute tasks at breakneck speed, leaving no time for collective sense-making, shared reflection, or structural knowledge building. This feels like productivity. It is actually the accumulation of invisible debt.

Here is the mechanism: every AI interaction that goes undocumented is a piece of reasoning that disappears permanently. The prompt that produced a critical piece of code, the validation logic behind a clinical template, the contextual judgement that shaped an architectural decision — if these are not narrated and captured, they exist only in a private browser tab that will be closed. At small scale, this is a minor inefficiency. At the scale that AI now enables, it is an organisational crisis in slow motion.

The vulnerability does not stay constant. It grows in direct proportion to adoption speed. The faster an organisation uses AI to generate outputs, the faster its stock of unverified, undocumented, unowned artefacts accumulates. If key individuals leave, or if commercial AI services change their pricing or availability, the organisation is left holding complex software, clinical templates, and architectural systems that no single human has fully read, understood, or can defend. The capability gap between what the organisation can produce and what it can govern widens with every shortcut taken.

3. So That the Following Good Things Happen

• A Codified Knowledge Commons: Prompts, validation rules, and context definitions are curated and shared, turning fleeting individual AI interactions into a durable, compounding asset.
• Verifiable Bastions of Truth: Organisations establish curatorial platforms of verified factual content, providing a reliable gold standard of truth to ground and check AI outputs.
• Organisational Ownership: The prompting patterns, validation rules, and context definitions that make AI work effective belong to the organisation — not to individual accounts or third-party platforms — so that capability is portable, auditable, and cannot walk out the door.
• Quality Governance With Teeth: Organisations establish quality as a primary metric — with defined standards, visible measurement, and real accountability — so that narrated judgment is wired into the organisation's DNA rather than bolted on as a compliance layer, and the shared commons becomes the mechanism through which that quality is maintained and compounded.

4. And the Following Bad Things Don't Happen

• Organisational Hollowing-Out: We don't become organisations that can generate endless artefacts but cannot govern, debug, or evolve them — shells with high output and no genuine understanding of what they have built.
• Compounding Capability Debt: We don't allow the gap between what we can produce and what we can govern to widen silently — accumulating unvetted code, clinical notes, and architectural designs that carry hidden regulatory, clinical, and security risk that only surfaces when it is too late to address cheaply.
• Governance Theatre: We don't allow quality processes to become paper exercises that satisfy auditors but transfer nothing — narration without accountability is administration, not governance.
• Fragile Intellectual Property: We don't lose the organisation's unique domain expertise because our best prompting patterns and system context were lost in private personal accounts.