AI Agent UpdateDeloitte: 28% report AI revenue gains, but only 25% are redesigning work
Deloitte's survey of 3,500 business leaders finds 28% now report revenue gains from AI — up from 20% in January — but only 25% are actually redesigning work around it. Speed is proven. Transformation is not.
The News
Deloitte published its 'State of AI in the Enterprise' report on October 29, surveying 3,500 director-to-C-suite leaders across 25 countries in May and June 2026, and the headline is a split verdict. 28% of leaders now report revenue gains from AI, up from 20% in January — real money, not just pilot projects. And 81% report faster decision-making. But only 25% say they use AI to create new products or services, redesign core processes or change their business models. The enterprise AI story has reached its ROI reckoning: the returns are arriving, but the reinvention is not.
The second finding is the one that should worry governance teams. 40% of leaders now allow AI to execute decisions and actions without human approval at least moderately — agents approving, routing, purchasing and answering on their own. That is the agentic enterprise arriving before the controls for it: companies are letting AI act while only a quarter of them have redesigned the work those actions sit inside. It is the gap between deploying agents and governing them, measured in a single survey.
The read-through for enterprise software is direct. Salesforce, SAP and Microsoft — named here as read-through tickers — sell the platforms on which this transformation is supposed to happen: agent frameworks, control planes and business applications with AI woven in. Deloitte's numbers say their customers are buying speed first and reinvention second. The vendors' next growth leg depends on converting the 28% reporting gains into the 25% redesigning work — a much harder sale than a faster dashboard.
28% see revenue gains, 25% redesign work — enterprise AI is paying off faster than it is transforming, and that gap is the story.
Why It Matters
The survey matters because it is the largest reality check on enterprise AI this year — 3,500 senior leaders, not vendor case studies. The revenue-gain jump from 20% to 28% is the first broad evidence that AI spending is returning cash, which underpins the whole enterprise AI investment case. But Deloitte's framing — an 'ROI reckoning' — is deliberate: the industry spent two years promising transformation and has mostly delivered acceleration. Investors should price the difference.
The governance finding is the second reason. 40% of companies letting AI act without human approval is a remarkable adoption rate for autonomous systems — and it arrives alongside a control-plane product wave (kill switches, cost caps, audit trails) that exists precisely because unsupervised agents are risky. The survey suggests deployment is outrunning design: companies grant autonomy first and build the guardrails second. That ordering works until an agent makes an expensive mistake in public.
The third reason is what it means for the vendors. Salesforce, SAP and Microsoft are all pitching agent platforms as the next enterprise layer, and Deloitte's numbers define their sales challenge: the easy sale (faster decisions) is landing, but the transformational sale (redesigned processes, new business models) is not — yet. The vendors that convert the 28% into the 25% will own the enterprise AI decade. The ones that merely sell speed will compete on price. Deloitte just published the scoreboard.
Analyst's Take
The verdict is Mixed. Deloitte's survey is genuinely good news for the AI enterprise thesis — revenue gains rising from 20% to 28% in under a year is a real inflection, and 81% reporting faster decisions means AI has cleared the usefulness bar. But the 25% transformation figure caps the celebration: most companies are still using AI to do old work faster, which is a productivity story, not a reinvention story. And 40% letting agents act unsupervised while only 25% redesign the work is a governance gap waiting for an incident.
What to watch next is whether the transformation number moves. The agent-control products launched this autumn — Salesforce's Agent Fabric controls, announced October 27 — are aimed at exactly the gap Deloitte measured. If next year's survey shows the redesign share climbing toward the revenue-gain share, enterprise AI graduates from tool to transformation. If it stalls, the ROI story plateaus at 'faster, not different' — and the vendors' agent upsell stalls with it.
By The Numbers
FAQ
What is Deloitte's 'State of AI in the Enterprise'?
An annual survey of senior business leaders about how their companies use AI. The latest edition, published October 29, covers 3,500 director-to-C-suite leaders across 25 countries, surveyed in May and June 2026.
What does 'revenue gains from AI' mean?
Leaders saying AI has increased their company's revenue — not just cut costs or saved time. The share rose from 20% in January to 28% now, the clearest sign yet that AI is starting to pay in the top line.
Why does only 25% redesigning work matter?
Because it separates using AI from transforming with it. Most companies bolt AI onto existing processes; only a quarter use it to create new products, redesign core processes or change their business model. Deloitte's point: the gains so far come from doing old work faster, not new work differently.
What does 'AI executing decisions without human approval' mean?
Agentic AI taking actions on its own — approving a transaction, rerouting a shipment, answering a customer — without a person signing off. 40% of leaders now allow that at least moderately, which is why agent controls and audit trails matter.
Why are Salesforce, SAP and Microsoft named here?
They are named as read-through tickers: the enterprise software vendors whose platforms companies use to deploy AI agents. If only a quarter of companies are redesigning work, those vendors' agent products are selling into a market still early in transformation.
👁
Source: Deloitte