Agentic AI scaling stalls at 15% as workforce readiness lags

The pilot ceiling, measured

Corporate interest in AI agents is not the problem. Turning that interest into working systems is. Deloitte's newest agentic transformation research found that only 15% of US organizations have reached what it calls scaled, orchestrated, multi-agent adoption. That means several agents coordinating work across functions such as customer service, IT and engineering.13 Of the rest, 42% are testing a handful of agents and 43% are extending deployments across functions.13 The survey covered 501 senior leaders who shape their companies' AI strategy or implementation.13 It was fielded between April and June 2026 across five industries, including financial services, life sciences, and technology, media and telecommunications.6

The main point of the research is less the low adoption figure than the explanation for it. Deloitte's conclusion, carried through ZDNET's coverage, is that companies cannot reach scale by bolting agents onto how they already work. They have to rebuild processes and invest in the people who will run them.13 Trade coverage of the same report reached the same conclusion. Most organizations will need to overhaul business processes, data and workforces before agentic AI can spread widely. Respondents put half of business processes being redesigned around agents at roughly three to four years away.19

Ambition runs ahead of readiness

The survey shows a large gap between what leaders expect and what their organizations can currently support. Almost two-thirds of leaders say advances in agentic AI are making them rethink their business models, but only about half have a clear picture of the agent-driven operating model they want.13 Looking ahead, 74% expect nearly half of business processes to be redesigned or rebuilt around agents by 2030. Sixty-one percent think most of their processes will eventually run on largely autonomous agents.13 Today, just 16% say their current processes are ready for agentic adoption.13

The readiness numbers are weak across the board. Thirty-six percent of leaders call their vision and strategy ready, 34% say the same of their technology infrastructure, 32% of their data foundation, and 26% of risk, security and governance.13 Workforce readiness is the lowest at about one in five.13 Even scaling does not resolve the issue. Among the 15% that have scaled, fewer than half (46%) think their processes are prepared, and across the whole sample only 5% describe their processes as highly prepared.6

The barriers leaders listed are partly technical and partly organizational. The most common were the lack of a unified, accessible data foundation (72%), the inability to trust and govern agents (70%), and the cost and complexity of integration (67%).13 Deloitte's separate State of AI in the Enterprise work points the same way on governance. It found that only 21% of enterprises have mature governance for agentic AI risks, while 74% expect to use agents at least moderately by 2027.20

Why layering agents on legacy work falls short

The research identifies a common error: adding AI on top of old processes instead of redesigning them. Deloitte does not dismiss layering. It argues that layering should be a bridge to redesign, and that companies should build process redesign as a repeatable skill rather than treat it as a one-time project.13 The timeline shows how slowly that is likely to happen. Only 31% of companies expect to have rebuilt their processes around agents by 2028, against 74% by 2030.13

Other research supports this. McKinsey's 2026 State of AI survey found that high performers are 2.7 times more likely than others to scale agents beyond coding.16 Nearly three-quarters of those high performers have fundamentally redesigned workflows, compared with about a quarter of everyone else.3 Consultants quoted in Anthropic's 2026 agents report made a similar case. Boston Consulting Group's Tom Martin said clients reach profit-and-loss impact faster when they rebuild systems end to end with agents at the center, rather than adding agents to legacy processes.15

The evidence supports a firm conclusion. Agents alone do not reliably produce returns. The value comes from redesigning the work around them, and that redesign is the slow, expensive part that most organizations have not started in earnest.

The workforce is the bottleneck

Workplace adoption is where the research is clearest and least comfortable for leaders. Half of the leaders Deloitte surveyed said their organizations are underinvesting in the workforce transformation that agent adoption needs.13 Forty-three percent expect major job disruption as routine, structured tasks move to autonomous agents.13 Many companies are working on the problem: 71% are building baseline agent literacy, and 65% are running upskilling and reskilling programs for roles that agents could affect.13 Literacy, though, is a much lower bar than redesigning roles.

Deloitte's broader 2026 data shows the size of that gap. Its research found that 84% of companies have not redesigned jobs around AI capabilities. Even so, 36% expect at least 10% of their jobs to be fully automated within a year.11 Employee attitudes add further difficulty. Only 13% of non-technical workers are highly enthusiastic about AI, 21% would rather not use it but will if required, and 4% actively distrust and avoid it.11 Access is also not the same as use. Worker access to sanctioned AI tools rose to about 60% in 2025, but fewer than 60% of those with access use AI in their daily workflow.11

Deloitte's US head of AI, Jim Rowan, has framed the challenge as getting the workforce behind the change and generating buy-in instead of skepticism. The firm warns that the larger risk is employees slipping back into old habits after workflows have been redesigned.15 McKinsey's survey adds detail on who carries the strain. Eight in ten respondents say AI has improved their own productivity, but mid-level managers and individual contributors are more likely than executives to report AI-related strain.16

This has practical consequences. Coverage of the Deloitte findings describes the human role shifting from doing tasks to designing workflows, auditing agent decisions, handling exceptions, and refining prompts and policies.12 Those are new jobs, and in most organizations no one is yet responsible for them.

Where the numbers diverge

Adoption figures vary widely from one study to the next, and the differences matter. McKinsey reports that 40% of respondents at organizations with more than $1 billion in revenue are scaling AI agents, up from 27% a year earlier, while smaller organizations are flat at 22%.16 KPMG says 53% of large US companies have deployed agents, and that orchestration of multiple agents doubled in a quarter from 9% to 18%.1 Capgemini, at the other end, found only 2% of large organizations running agents at full scale.8 Gartner put agent deployment at 17% and forecasts that more than 40% of agentic projects will be canceled by the end of 2027.3

Most of the difference comes from definitions. McKinsey counts scaling within a single function, which one analysis calls a low bar.3 Deloitte's 15% requires orchestrated, multi-agent deployment across core functions, which is a much stricter test.13 Read together, the figures are consistent. Many companies have agents working somewhere, but very few have agents coordinating work across the business. ZDNET's follow-up, which set Deloitte's results beside research from KPMG and Accenture with Wharton, concluded that agent adoption is real but production-grade deployments that generate value remain rare.14

KPMG's tone is more optimistic. It reports that 76% of businesses now see real value from AI and 71% say they are making good progress toward an integrated AI-human workforce.14 Those figures measure confidence, and they sit uneasily next to Deloitte's finding that only one in five organizations has a ready workforce.13 Leaders' self-assessments seem to be moving faster than the operational changes behind them. Accenture and Wharton's point that intelligence may scale while accountability does not fits that reading.14

Readers should also note who is presenting the findings. ZDNET's main write-up of the Deloitte report was written by Vala Afshar, Salesforce's chief digital evangelist.13 Salesforce sells agent technology. The underlying numbers come from Deloitte, but the framing of agentic AI as inevitable comes from someone with a commercial stake in it.

What it means for employers

Deloitte's recommendations are simple to list and hard to carry out. Companies should build an integrated agentic roadmap, treat layering as a bridge, fund workforce transformation properly, and define how humans and agents will work together.13 The research also warns that costs will rise in ways budgets may not expect, including training, tokens and process redesign. It suggests leaders may need new performance metrics, such as tokens used per automated unit of work.13

The main lesson for workplace adoption is that buying agents is not the hard part. Access to tools has grown quickly, agent counts have grown with it, and individual employees report real productivity gains.1116 The companies pulling ahead have redesigned the work and helped their people adapt to the redesign. For the 85% still testing or expanding, the next step is a large organizational change program, and the data suggests few have funded one at the scale it needs.