Deloitte's AI Institute has published its eighth annual State of AI in the Enterprise report, surveying 3,235 business and technology leaders between August and September. The headline finding is one that should matter to anyone tracking where AI-related jobs are heading: agentic AI — AI systems that can take multi-step actions rather than just answer questions — is now rated as the single highest-impact emerging capability by the leaders surveyed, ahead of other generative AI applications.

Adoption is real but still early

Only around 23% of surveyed organizations report using agentic AI at a moderate level today. That's a modest current footprint for a technology generating this much attention. But the forward-looking number is the one worth paying attention to: 74% of leaders expect moderate-to-full agentic AI integration within the next two years. If that holds, the gap between today's limited deployment and near-future expectation is where most of the hiring and reskilling activity is likely to concentrate.

Where companies are actually using it

The report identifies customer support, supply chain management, R&D, and cybersecurity as the leading use cases for agentic AI inside the organizations surveyed. These are functions where a system taking autonomous multi-step action — resolving a support ticket end-to-end, rerouting a shipment, running a research query across multiple tools — has a clearer, faster payoff than in more judgment-heavy or relationship-driven roles.

The governance gap is the real story

The number that should concern job seekers and employers alike is this: only about 21% of organizations say they have mature governance models in place for autonomous AI agents. That's a wide mismatch against the 74% expecting significant agentic AI integration within two years. Deployed agents that aren't well-governed are a business risk — unchecked actions, unclear accountability, audit gaps — and companies know it.

What this means for jobs

This governance gap is itself a hiring signal. As companies move from piloting agents to running them at scale, demand is rising for roles that didn't clearly exist a few years ago: AI governance and risk specialists, agent-output auditors, prompt and agent quality assurance, and "human-in-the-loop" oversight roles embedded directly in customer support, supply chain, and security teams. If you work in any of the four leading use-case functions Deloitte named, expect your role to increasingly include some layer of agent oversight rather than being replaced outright by one.

• If you're in customer support, supply chain, R&D, or security, expect agentic tools to show up in your workflow before they show up anywhere else — these are the functions companies are prioritizing.
• Governance and oversight skills are becoming a genuine differentiator. Understanding how to monitor, audit, or set guardrails for an AI agent is a specific, learnable skill that most of the workforce doesn't yet have.
• "Moderate use today, major use in two years" is a narrow hiring window. Companies that move early on both deployment and governance will likely have a head start building the internal expertise to manage it.

Full methodology and findings are available via Deloitte's State of AI in the Enterprise report.