layers grouped by what they're for. intent plane is where the board makes promises. production plane is where engineers discover the promises don't work.
A grouping of layers by operational function. The AI Enterprise Control Index defines five planes: Intent (L1–L2), User (L4), Production (L5), Knowledge (L6), and Integration (L7, L8). Planes provide an executive-level view of governance maturity, enabling boards and senior leaders to assess coverage across operational domains without requiring layer-level technical detail.
Why It Matters
Eight layers is too many for a board paper and too few for an engineering specification. Planes solve this by grouping layers into five operational domains that map to how organisations actually think about AI: what did we promise (Intent), what do users experience (User), how is it built (Production), what knowledge does it use (Knowledge), and how does it connect to external systems (Integration).
The practical value is triage. When an organisation begins its governance journey, it cannot mature all layers simultaneously. Planes enable prioritisation. A regulated financial institution may start with the Intent plane (risk appetite, ethics) and the Production plane (model evaluation, deployment gates). A technology startup may start with the Production plane and the User plane (application-level controls). The plane structure allows organisations to make deliberate, defensible sequencing decisions.
Planes also expose the most common governance failure: misalignment between what is declared and what is operational. The Intent plane (L1–L2) contains strategy and ethics. The Production plane (L5) contains engineering. When these two planes are not connected — when the board’s ethical AI principles have no corresponding engineering controls — the organisation has a strategy-execution gap that no amount of policy documentation will close.
The Stress Test
Your board has approved an AI ethics statement, a risk appetite framework, and a set of responsible AI principles. The Intent plane is at Maturity Level 3. The board asks for a report on Production plane maturity. You discover the Production plane is at Level 1: no model evaluation pipeline, no deployment gates, no post-deployment monitoring. The Knowledge plane is at Level 0: no data governance at all.
The organisation has intent without capability. The planes expose this gap instantly. Without the plane view, the board would see a mix of layer-level scores and conclude governance is “progressing.” With the plane view, the board sees that three out of five operational domains are ungoverned.
In the Wild
Google published AI Principles in 2018, establishing an Intent plane governance position. Internal research by Timnit Gebru and Margaret Mitchell on the environmental and social costs of large language models revealed a gap between the Intent plane (published principles) and the Production plane (how models were actually trained and deployed). The subsequent firing of both researchers demonstrated that Intent plane declarations without Production plane mechanisms create a governance structure that cannot withstand internal challenge.
The Intent plane said “responsible AI.” The Production plane had no mechanism to operationalise it. The gap between planes is where the governance failure lives.
Robinhood’s gamification of trading created a User plane experience optimised for engagement. The Integration plane — connecting user actions to market infrastructure — lacked the controls to manage the consequences. During the GameStop trading event, Robinhood restricted trading because its Integration plane (clearing house requirements, capital reserves) could not support the volume generated by its User plane. The $70 million FINRA fine and Congressional hearings followed.
A User plane optimised for engagement without an Integration plane designed for resilience is a business model that works until it doesn’t.
A 2025 survey by the European AI Office found that 68% of organisations subject to the EU AI Act had established governance principles and risk assessment frameworks (Intent plane) but fewer than 20% had implemented technical controls for model evaluation, bias testing, and post-deployment monitoring (Production plane). The survey concluded that the AI Act’s requirement for a “risk management system” would be met on paper by many organisations but met in practice by few.
An entire regulatory ecosystem with mature Intent planes and immature Production planes. The plane-level view makes this visible at a glance.
How to Govern It
Planes make governance legible to the people who fund it.
Within the AI Enterprise Control Index, planes serve a specific governance function:
- Intent Plane (L1–L2) — Strategy and Ethics. Board-level declarations: risk appetite, ethical principles, posture decisions (build vs. consume), and regulatory strategy. Owned by the board and C-suite.
- User Plane (L4) — Applications and Agents. User-facing controls: output disclaimers, confidence thresholds, human-in-the-loop mechanisms, and application-level circuit breakers. Owned by product and application teams.
- Production Plane (L5) — AI Engineering. Technical controls: model evaluation, training pipelines, deployment gates, and rollback mechanisms. Owned by AI engineering leads.
- Knowledge Plane (L6) — Data and Knowledge Management. Data governance: quality, provenance, lineage, consent, and retention. Owned by data governance officers.
- Integration Plane (L7–L8) — Interoperability and External Interfaces. System integration: API governance, third-party connections, and cross-system data flows. Owned by enterprise architecture.
When It's Relevant
Planes are most relevant in board reporting, governance programme design, and maturity assessment. They provide the level of abstraction that enables strategic decision-making without requiring technical depth. A CISO presenting to the board uses planes. An engineer implementing controls uses layers and components.
Plane-level governance becomes critical when:
- The board needs a legible view of AI governance maturity
- The organisation is prioritising which governance domains to mature first
- Strategy-execution gaps need to be identified and communicated
- Multiple teams need to coordinate governance across operational domains
- The organisation is reporting governance status to regulators or auditors
Related Terms
References
- [1] Bender, E.M., Gebru, T., McMillan-Major, A. and Shmitchell, S. (2021) ‘On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?’, Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT), pp. 610–623.
- [2] FINRA (2021) FINRA Fines Robinhood Financial $70 Million. Financial Industry Regulatory Authority.
- [3] European AI Office (2025) AI Act Implementation Survey: State of Readiness Among EU Organisations. European Commission.
- [4] ISO/IEC 42001:2023. Information Technology — Artificial Intelligence — Management System. International Organization for Standardization.
- [5] NIST (2023) Artificial Intelligence Risk Management Framework (AI RMF 1.0). National Institute of Standards and Technology, U.S. Department of Commerce.
- [6] European Parliament and Council (2024) Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (AI Act). Official Journal of the European Union.
- [7] Kaplan, R.S. and Norton, D.P. (2008) ‘Mastering the Management System’, Harvard Business Review, 86(1), pp. 62–77.
AI Enterprise Control Index v6.0 · Glossary · June 2026 · i-DEPOT 158508 (BOIP) · CC BY-NC-ND 4.0
By Jeroen Janssen, Apparens