Contexts

Artificial Intelligence Cannot Fix What the Institution Has Spoiled

Every institution begins with an announcement about artificial intelligence in a similar way: choosing the model, data centers, and perhaps governance of use. The assumption underlying all of this is that institutional intelligence is a natural result of the new technology if properly selected and used. The matter is more complex than using and installing the technology if concerns of technological sovereignty cast their shadows over you.

An institution can assign an assistant to every official to summarize regulations, search procurement archives, and draft policies in seconds, but it will be unable to answer important questions not found in documents, such as why payments stopped in the month of Safar, why the institution exempted the operations department from the hiring ceiling for that year but not previous years, and why the HR director general is likely to extend this exemption to next year? The documents recording decision-making processes exist, but the contexts explaining them are missing. This gap is the issue of sovereignty over artificial intelligence that almost no one talks about.

The problem is not the information in the systems, which requires us to collect and organize it, but rather the information not in the systems and never has been. The missing information that AI needs is the institution's logic that accumulates to explain its work and culture. It is the institutional memory distributed in the work methodology and those who carry it out, represented by problem-solving methods, alternative search methodologies, and decision-making logic. You will not find this information saved in any existing system.

The danger lies in AI succeeding in absorbing and extracting the available information and filling the knowledge gap with its inferences, which begin to redirect decisions if given the opportunity. If you direct a capable model to extract fragmented information and undocumented practices, it will not stop at the deficiency and declare its inability. The opposite will happen: it will produce a confident and well-written answer without any basis, and if it passes without review, it will enter the fabric of the institution and become difficult to separate. In contrast, AI is the most capable tool to bridge this gap if placed within the digital strategy. AI will be required to rebuild the institution's memory to provide an explanation for its actions sufficient to support strategic and executive decision-making.

The same technology can widen or bridge gaps. The model's role is limited if the necessary information is not provided to it, and the real challenge is whether institutional memory exists so that it can be extracted from those who have the institution's memory and those who have fragments of it.

This reformulates what technological sovereignty means. The question of who owns the model or who hosts the data center diminishes in importance compared to who understands the workflow that produced the data, who interprets the logic of the system design and the context that produced it, and how flexible the institution is in facing supplier challenges and their stability in the market locally and globally under geopolitical pressures.

The truth we must deal with is that institutional intelligence is not produced by models or data alone. The quality of the model depends on the quality of the information the institution possesses. In the AI race, the strongest model will not tip the scales or carry contestants the farthest, as long as institutional knowledge is lacking. The competitive advantage will be in the hands of those who can grasp their institutional knowledge and employ it optimally.