IBM released new research on January 20, 2026, showing that global executives expect artificial intelligence to drive revenue by 2030. The study covers more than 2,000 C-suite leaders across 33 countries and matters as companies scale AI spending despite execution gaps.
What changed in executive AI expectations
IBM found that 79% of executives now expect AI to contribute to revenue within five years. Today, only 40% say AI delivers direct revenue. This marks a sharp shift in how leaders frame AI inside growth plans.
At the same time, clarity remains limited. Only 24% of respondents said they understand where future AI-led revenue will come from, according to the IBM study.
Context behind IBM’s AI findings
The research shows AI moving beyond efficiency tools. While 47% of current AI spending targets productivity gains, executives expect 62% of future investment to focus on innovation by 2030.
IBM said this change reflects how firms now treat AI as part of core strategy. Still, execution challenges persist across sectors and regions.
Impact on companies and leadership
IBM data shows growing concern inside boardrooms. About 68% of executives worry AI projects may fail because teams have not integrated them into core operations.
Meanwhile, leadership structures may shift. By 2030, executives expect one in four enterprise boards to include an AI adviser or co-decision-maker. In parallel, 74% believe AI will reshape leadership roles.
Supporting data from the IBM study
The IBM Institute for Business Value conducted the research with Oxford Economics. It surveyed more than 2,000 executives across 20 industries.
Executives expect AI to lift productivity by 42% by the end of the decade. Also, 70% plan to reinvest AI-driven productivity gains into growth initiatives rather than cost control.
How IBM sees the next phase
IBM reported rising uncertainty around technology choices. While 57% of executives see AI model sophistication as critical, only 28% know which models they will need by 2030.
Many expect multi-model systems to dominate. Several respondents also said smaller language models may outperform large ones for enterprise use cases.
