Oil Fund expands AI use, Norges Bank keeps final decisions with humans

• The Oil Fund is using AI tools in research, information handling and internal workflows
• Fund executives say humans will retain decision-making authority
• The shift raises a governance question for a state-linked manager of public wealth: who defines the boundary between recommendation and decision?
• Automation may cut time spent on routine tasks while making internal processes harder for outsiders to inspect

Nicolai Tangen has made artificial intelligence a priority inside Norway’s Oil Fund, the world’s largest sovereign wealth fund, but the line he and his AI chief draw is blunt: the machine advises, the human decides. In VG’s reporting, the fund’s AI leadership describes a growing role for AI across the organisation while insisting that judgment on consequential matters remains with people.

That distinction matters more at the Oil Fund than in most offices. Norges Bank Investment Management, which runs the fund, manages public wealth on a scale large enough that small changes in process can affect enormous sums. If AI is used to sort information, summarise companies, compare signals across markets or assist analysts and portfolio managers, the gain is obvious: fewer staff hours spent on repetitive work, faster filtering of data, quicker internal responses. The harder question is where assistance ends and deference begins.

VG describes an organisation where Tangen’s enthusiasm for AI is pushing adoption forward, but where the internal message from the fund’s AI leadership is that full delegation is off the table. That leaves a narrower but more consequential issue: who inside the institution defines which tasks are safe to automate, which require human review, and what counts as a genuine decision. In a private hedge fund, that boundary would be a matter for owners and managers. In a state-linked vehicle built from petroleum revenue and held on behalf of the Norwegian public, the chain of accountability runs further, while visibility into internal systems remains limited.

The attraction of AI for a fund of this size is not hard to see. Investment organisations drown in documents, earnings calls, regulatory filings, broker notes and internal reporting. A tool that can compress that flood into something analysts can scan in seconds buys time. It may also standardise work that otherwise depends on who happened to read what. But standardisation has its own cost. Once a machine-generated summary, ranking or recommendation becomes the first screen through which staff see the world, the system starts shaping judgment before any formal decision is taken.

The Oil Fund’s answer, as presented to VG, is to keep a human at the end of the chain. That is a safeguard, but also a narrow one. A final sign-off is only one part of control; the upstream design of prompts, models, thresholds and permitted uses decides what reaches that sign-off in the first place. The more routine work disappears into internal AI systems, the more outsiders are asked to trust a process they cannot inspect.

For now, the fund’s public line is simple: AI can support, but not replace, the people in charge. At a sovereign fund built from offshore oil and gas revenue, the machine may draft the memo, sort the signals and compress the noise. The signature still belongs to a human being.

Källor: VG