Tooling · 4 min

Structural
Institutional Note

Published

27 April, 2025

Author

Francesco de Leo Kaufmann

Climate action is only as strong as the data that informs it. But most data stacks weren’t designed with emissions, supply chains, or climate modeling in mind. Teams are often stuck retrofitting existing systems or relying on brittle workarounds to generate insights. It’s time to rethink our infrastructure—starting with the foundation.

Two kinds of AI investment

The decisive variable in the velocity economy is not how much an organisation has invested in AI. It is where the learning accumulates. Organisations that embed intelligence into their core operations — into the systems that move people, manage energy, deliver services — build a learning curve that compounds with every operational cycle. Organisations that procure AI externally, deploying vendor platforms and third-party models, generate learning that accumulates inside a supplier’s infrastructure, not their own. Both appear identical in a technology budget. They produce fundamentally different strategic outcomes.

Rented ground

There is a straightforward diagnostic for any organisation evaluating its position in this bifurcation: where does your intelligence live?
If the most valuable AI capabilities in your organisation are provided by an external platform vendor, you are building learning on rented ground. The inference is happening inside your operations. The learning — the model refinement, the pattern recognition, the accumulated operational knowledge — is happening inside someone else’s cloud. When the contract ends or the vendor’s priorities shift, the learning does not stay with you.

The critical strategic question is not ‘are we using AI?’ It is ‘where is the learning accumulating — inside our organisation, or inside a supplier’s infrastructure?’

What your board has not yet asked

The governance of the intelligence layer — who controls model updates, who owns operational data, who holds audit rights — is a strategic decision that belongs at board level. In most organisations, it is sitting in the IT budget. Three questions the board must be able to answer directly:

Is the learning accumulating inside the organisation? Not the output of AI — the learning itself. Sensor data linked to outcomes. Maintenance records correlated with failure modes. Operational knowledge that improves every system cycle. If this is structured and retained internally, the organisation is building a strategic asset. If it is being processed and retained by a vendor platform, the organisation is generating data for someone else’s model.

Who governs the intelligence layer? Who sets the learning objectives for the organisation’s AI systems? Who has the authority to audit algorithmic decisions affecting service, safety, and regulatory compliance? These questions are not delegable to management. They are fiduciary responsibilities — and most board agendas have not yet named them as such.

Can the organisation migrate without losing accumulated learning? If the answer is no — if switching providers means losing the operational intelligence that has been built over years of deployment — the board is accepting a form of strategic dependency that may be more consequential than any financial risk currently on the balance sheet.

A board that cannot answer these questions is not governing its organisation’s most consequential strategic asset. The bifurcation does not wait for the strategy review.