80% of CEOs expect AI to overhaul operations by 2028
A Gartner survey of 469 CEOs and senior executives found that four in five expect AI to drive significant changes to operational capabilities over the next three years. The shift moves beyond automating individual tasks toward what Gartner calls "autonomous business"-where self-learning software agents handle decisions and actions with minimal human oversight.
The survey, conducted through the fourth quarter of 2025, shows a clear trajectory. Today, 54% of organisations limit automation to specific tasks. By the end of 2028, only 13% expect to stay at that level.
Two paths forward
The remaining 87% split into two camps. Thirty-two percent plan to deploy self-learning AI that supports human decision-making. Twenty-seven percent expect their organisations to operate primarily without human intervention.
This divide reflects real choices operations teams will face: augment existing workflows with adaptive AI, or pilot autonomous agents in narrow domains and scale from there.
Revenue and customer risk
CEOs also flagged business exposure. Twenty-eight percent see transactional revenue as most at risk from AI-a signal that automated pricing engines, procurement systems, and negotiation agents could reduce intermediary roles. Seventeen percent expect major changes to their customer base.
These concerns align with public examples where agentic automation has already compressed margins in transaction-heavy sectors.
What operations teams need to prepare
Moving toward autonomous operations requires more than new software. Organisations typically increase investment in three areas: data infrastructure, model lifecycle tooling, and runtime governance. These capabilities support continuous learning systems while managing drift, reliability, and safety at scale.
Integrating self-learning agents with legacy systems raises complexity. Teams will need stronger monitoring, explainability controls, and formalised interfaces between human workflows and automated agents.
Practitioners should prioritise stable data pipelines and reproducible model training now. Organisations experimenting with machine-customer interactions should instrument those tests carefully-the survey suggests transactional revenue exposure will be material.
Watch how your organisation budgets for data platforms and model operations over the next 12 months. Those line items signal whether leadership is serious about autonomous capabilities or treating this as incremental automation.
For more on building operational AI systems, see AI for Operations and AI Agents & Automation.
Your membership also unlocks:
