Deloitte’s agentic AI survey exposes a business-readiness gap

Deloitte’s August 2026 research found that enthusiasm for agentic AI had travelled considerably further than operational readiness: only 5% of surveyed leaders described their organisations’ business processes as highly prepared. The findings came from 501 US respondents whose organisations were already at least piloting the technology. They measured executives’ assessments and expectations, not independently verified performance or investment returns.

What happened

In its release published on 12 Aug 2026, Deloitte reported a striking mismatch between ambition and preparation. Within four years, 74% of respondents expected nearly half their business processes to be redesigned or rebuilt around AI agents. Yet only 21% considered their processes prepared or highly prepared for adoption.

Deployment was not the same as readiness. According to Deloitte, just 15% reported scaled adoption involving coordinated agents working across business functions. Even among scaled adopters, only 46% believed their processes were prepared. Deloitte also said many scaled deployments were being used in applications with limited risk and limited returns. Scale, in this account, was no guarantee of economic significance.

The survey’s seven readiness measures made the imbalance clearer. Vision and strategy was the only area where a majority—52%—reported being prepared or highly prepared. Technology infrastructure followed at 48%, then data foundations at 42%, risk, security and governance at 39%, ecosystem partnerships at 34%, workforce at 25% and business processes at 21%. The boardroom had a destination; much of the organisation still lacked transport.

The workforce figures added a spending problem to the operational one. Deloitte found that 43% expected significant workforce disruption within 12 to 18 months, rising to 72% over two to three years. Nevertheless, half said their organisations were not investing adequately in AI-related workforce transformation. These were expectations of changing work, not reported redundancies or a forecast of a particular number of jobs lost.

The background

Deloitte fielded the survey between April and June 2026. Respondents ranged from senior managers to C-suite executives, were based in the United States and were involved in their organisation’s agentic AI strategy or implementation. Every participating organisation had to be at least piloting agentic AI. The results therefore describe organisations already engaged with the technology, not the readiness of American businesses generally.

The sample covered five industry groups: consumer; energy, resources and industrials; financial services; life sciences and health care; and technology, media and telecom. Technology-focused functional leaders accounted for 55% of respondents, with business and other functional leaders making up 45%. Deloitte supplemented the survey with 20 interviews involving executives and AI and data-science leaders at large organisations.

Those boundaries matter for interpretation. The release did not establish a UK readiness rate or provide a small-business breakdown. Nor did its self-reported preparedness scores demonstrate that any particular operating model produced better commercial results. For buyers, that is an important distinction between a management survey and performance evidence—the broader purchasing question explored in the evidence gap in AI benchmarks.

The reported obstacles were more specific than a general shortage of enthusiasm. Deloitte found that 72% said they lacked unified, accessible data; 70% did not feel able to trust and govern agents; and 67% regarded integration as too costly and complex. These measures concern different barriers and should not be added together. Collectively, they place much of the difficulty outside the agent itself, in the systems and controls it must use.

Operating-model uncertainty compounded those technical constraints. Nearly two-thirds of respondents said their organisations were undertaking a wholesale reassessment of their business models, but only about half had a clear view of an agent-powered operating model. Reconsidering how a business works was evidently easier than specifying how responsibilities, decisions and workflows would change.

What people are saying

The reactions in the release came from Deloitte’s own leaders. China Widener, its vice-chair and US technology, media and telecommunications industry leader, argued that value would depend on redesigning products, work, financial models and governance, rather than on agents alone. Her position shifted the investment question from acquiring software to changing the business around it.

Laura Shact, Deloitte’s US technology, media and telecommunications AI growth leader, similarly argued that isolated fixes and quick gains would be insufficient. She called for investment in work and organisation design, leadership and workforce preparation alongside technology and data strategies. This was Deloitte’s prescription for the readiness gap, not a causal finding established by the survey.

Respondents nevertheless saw a reason to proceed incrementally. Deloitte reported that most interviewed leaders were concentrating on adding agents to existing processes, partly because doing so might deliver earlier returns. Fewer than one-third of survey respondents—31%—expected a majority of their processes to be redesigned around agentic AI within two years. Near-term budget logic and longer-term transformation ambitions were pulling in different directions.

Executives’ expectations also left a substantial role for people. Deloitte found that 61% expected most agents in use to be generally autonomous within four years, with humans providing oversight. At the same time, 75% agreed that collaboration between people and agents created more value than agent-powered automation alone. These were stated beliefs about value, not a measured comparison between the two approaches.

Murray Wilkinson, founder of Found Revenue, an AI business services provider with three UK locations, says firms that see value usually pick one process, measure it before and after, and keep a named person accountable for the outcome.

What happens next

The release set out preparation priorities, not subsequent results. Deloitte recommended redesigning processes and workflows, strengthening workforce readiness and trust, and developing new operating models and organisational structures. Nothing in the supplied evidence establishes whether respondents later made those changes or achieved their expected returns. The survey’s forecasts remain anchored to its 2026 findings.

For data and infrastructure, the practical implication is a broader investment case than the price of an agent. Respondents’ concerns about inaccessible data and expensive integration suggest that connecting systems and making information usable could determine whether a deployment advances beyond a pilot. The survey did not quantify those costs, so it cannot supply a reliable implementation budget or payback period.

Governance presents a different preparation task: defining what autonomous operation means in practice. The combination of expected autonomy and reported distrust makes authority and oversight central to Deloitte’s argument. Questions about which actions an agent may take, and when a person must intervene, become particularly concrete in automated purchasing and prompt-injection risk. That is a related control problem, not an incident reported by this survey.

For workforce preparation, Deloitte identified changing job requirements, additional learning and development, new working arrangements and new roles. The 25% workforce-readiness score does not isolate a shortage of technical recruits; the reported problem was broader organisational preparation. This distinction matters when assessing narrower uses of AI, such as Morgan Stanley’s adviser-note case study: adopting assistance for a task and redesigning responsibility across functions are different undertakings.