AI is getting more capable and cheaper every year. According to Stanford University's AI Index Report 2025, querying a model at GPT-3.5 level became more than 280 times cheaper between late 2022 and late 2024. Yet in a National Bureau of Economic Research (NBER) survey of about 6,000 executives in the US, UK, Germany and Australia, nine in ten report that AI has changed neither productivity nor employment at their firm. According to S&P Global, 42 percent of companies abandoned most of their AI initiatives, up from 17 percent the year before. In Germany, a third of AI adopters report costs above plan (Bitkom, 2026).

Horváth's white paper explains this gap. What every competitor can buy, none can win with. AI adds one new source of competitive advantage: the operational learning a company gains from using AI, retains and uses to improve decisions. It sits in the decision record, the evaluation standards and the operational memory. A company that leaves these assets with its AI provider rents its own experience back.

Gabriel Senftle and Kai Sowka turn this into an AI agenda for management and supervisory boards to 2035. They show which AI spending is strategic, how to choose the operating model for each business domain, how a board can tell whether AI pays, and which questions the supervisory board should ask at its next meeting.

Key findings

• Standard AI is becoming a commodity. What every competitor can buy, none can win with.
• The new advantage is the learning a company keeps. It sits in the decision record, the evaluation standards and the operational memory.
• Only one of four types of AI spending is strategic. Rent standard capability, standardize necessary control work, own company-specific learning, fund unproven activity only on proof.
• Authority stays with people. Every consequential automated decision needs a mandate, an evaluation standard, an escalation path and a named accountable person.
• One number shows whether AI pays. The cost per verified outcome, reported per material workflow.