Friday, 18 September 2026
Companies are preparing for a significant expansion in the use of autonomous AI systems, but many have yet to establish the management structures needed to control their deployment at scale, according to research from Deloitte.
The consultancy’s research indicates that 74% of respondents expect their organisations to use AI agents to at least a moderate extent over the next two years. However, only 21% said their companies currently have mature governance arrangements for managing the risks associated with the technology. The difference highlights a growing challenge for businesses moving from experimental applications towards wider operational use.
Deloitte argues that chief operating officers are likely to play an increasingly important role in this transition because the introduction of AI agents extends beyond technology implementation. Companies may need to reconsider how work is organised, where responsibility sits and which activities should remain under direct human control.
Rather than introducing autonomous systems into existing processes without changing them, businesses are being encouraged to examine complete workflows. This could mean determining which activities are best performed by employees, which can be delegated to AI and where human approval or intervention remains necessary.
The approach also changes how companies assess the return on AI investment. Deloitte argues that businesses should concentrate less on the number of agents deployed or individual tasks automated and more on whether the technology produces measurable improvements in business performance.
Sameer Shetty, Group Head of Digital Business and Transformation and Strategic Programs at Axis Bank, said companies need to consider the wider operating model when making these investments. “You can spend a lot of money on agentic AI without achieving real, valuable outcomes,” he said, adding that organisations need to understand both the expected return and the changes required internally to achieve it.
Coordination becomes more complicated as companies move beyond individual applications towards networks of AI agents operating across different departments and systems. This creates questions around accountability, escalation procedures, monitoring and the points at which employees should intervene in automated processes.
Alessio Marras, co-founder and Head of Organization at Italian digital bank AideXa, said the complexity increases as companies progress from individual applications towards interconnected systems. “Agents that are very efficient in single use cases can become much more difficult to coordinate and govern when they operate as part of an orchestrated system,” he said.
The transition could also influence corporate outsourcing strategies. Companies have spent decades moving selected activities to external service providers, but Deloitte suggests that AI could make it economically practical to bring some capabilities back inside organisations. Businesses may therefore reconsider which functions they own directly and which continue to be provided externally.
Another difference from previous large technology programmes is the ability to introduce AI agents gradually. Companies can begin with individual applications and expand or modify their systems as experience develops. This provides flexibility but also requires operating structures capable of adapting as the technology and its responsibilities change.
Workforce requirements are expected to evolve alongside the technology. As AI takes over more routine execution and analytical work, employees could devote more time to decision-making and higher-value activities. At the same time, companies will need people capable of supervising AI-supported processes, assessing results and understanding when human intervention is required.
Performance measurement may consequently need to change. Traditional corporate systems frequently assess employees according to activity or volume, while wider adoption of autonomous AI could place greater emphasis on the final business result. Deloitte argues that organisations should establish measurable objectives before scaling deployments and continuously assess whether the technology is delivering those outcomes.
For chief operating officers, the challenge is therefore becoming less about adopting individual AI tools and more about redesigning how organisations function. As agentic AI moves further into corporate operations, the companies that successfully scale it will need to combine technology investment with governance, workforce development, clearly assigned responsibility and measurable business objectives.
Source: Deloitte
