AI agents have moved from boardroom curiosity to operating reality in less than 24 months. According to G2's 2025 AI Agents Insights survey, 57% of companies already have AI agents running in production, with another 22% in pilot. Yet KPMG, Deloitte, McKinsey, and Gartner all reach the same uncomfortable conclusion: most organisations have deployed faster than they have prepared. Readiness — not access to technology — is the variable that now determines who scales and who stalls.
The pace of change is unprecedented
The shift in adoption between 2024 and 2026 has no obvious historical parallel in enterprise technology. KPMG's Q3 2025 AI Quarterly Pulse survey found that the share of organisations with AI agents deployed had nearly quadrupled in two quarters — from 11% to 42%1. By Q4 2025, 67% of leaders said they would maintain AI spending even in the event of a recession, and projected outlays of approximately $124 million per organisation over the next year1.
The directional consensus across the major research firms is unambiguous. McKinsey reports that 88% of organisations now use AI in at least one business function and that 62% are at least experimenting with AI agents2. Gartner projects that 40% of enterprise applications will integrate task-specific AI agents by the end of 2026, up from less than 5% in 2025, and that 33% of enterprise software applications will include agentic AI by 20283,4.
The readiness gap is now the value gap
The numbers that should worry executives are not the adoption figures. They are the readiness figures. Independent research consistently exposes a sizeable gap between ambition and operational maturity.
- Fewer than 20% of organisations report mature data readiness, and more than 80% lack the AI infrastructure — monitoring, observability, audit, and control — required to govern agentic systems at scale5.
- Deloitte's 2026 State of AI in the Enterprise found that only one in five companies has a mature governance model for autonomous AI agents6.
- McKinsey reports that nearly two-thirds of organisations have not yet begun scaling AI across the enterprise, and that fewer than 10% have scaled agents in any single function2.
- PwC's CEO Survey found that 44% of business leaders report workforce efficiency gains from AI implementation, but only 24% report measurable profit impact — a 20-point gap between activity and outcome7.
The pattern is the same across every major data set: organisations are deploying agents into infrastructure that was never designed to govern them.
The Gartner warning
In June 2025, Gartner published a forecast that has since framed almost every serious enterprise AI conversation: more than 40% of agentic AI projects will be cancelled by the end of 20279. The firm cited three causes — escalating costs, unclear business value, and inadequate risk controls — and added that many "agentic" projects are not in fact agentic at all. Gartner used the term "agent washing" to describe the rebranding of existing assistants, RPA bots, and chatbots without substantive autonomous capability, and estimated that only around 130 of the thousands of self-described agentic AI vendors offer genuinely agentic systems9.
Most agentic AI projects right now are early-stage experiments or proof of concepts that are mostly driven by hype and are often misapplied. — Gartner, June 20259
The lesson is not that agents do not work. The lesson is that they do not work in environments that have not been prepared for them.
Where readiness is paying off
When AI agents are deployed into prepared environments, the productivity returns are substantial — and increasingly well-documented in peer-reviewed and vendor-published case studies.
A longitudinal study of an enterprise software organisation, published in 2025, recorded a 31.8% reduction in pull-request review cycle time after the deployment of multi-agent code review tooling, alongside an approximate 28% increase in production code shipment volume10. A peer-reviewed study published in Harvard Data Science Review documented a 92% reduction in audit report preparation time at industrial gases firm Linde following the deployment of a multi-agent audit system, and a corresponding scaling of B2B sales-negotiation scenario generation at Stora Enso11.
PwC's May 2025 AI Agent Survey found that 79% of US companies surveyed had already adopted AI agents in some form, and that two-thirds of those (66%) reported measurable productivity value7. ServiceNow has separately documented 80% autonomous handling of customer-support inquiries and a 52% reduction in time-to-resolution for complex cases, generating an estimated $325 million in annualised value12.
The common factor across every published success is not the agent technology. It is the readiness of the operating environment around it: clean data, clear workflow boundaries, governance with human oversight, and a measurement framework tied to a business outcome.
The eight dimensions of AI agent readiness
Drawing on McKinsey's six-dimensional Rewired framework2, KPMG's foundations-of-scale reporting1, and Deloitte's governance research6, eight readiness dimensions consistently separate the organisations that scale agents from those that do not:
- Data readiness. Agents are only as accurate as the data they retrieve. Gartner reports that eight in ten companies cite data limitations as a primary blocker to agent scaling. Modernised data architecture, retrieval-augmented generation, and clear data ownership are non-negotiable.
- Infrastructure and observability. Multi-step agent workflows require tool-calling infrastructure, prompt and output logging, telemetry, and rollback procedures. Most organisations admit they lack these.
- Governance and human-in-the-loop design. McKinsey identifies the existence of defined processes for when agent outputs require human validation as one of the strongest correlates of value capture2.
- Workflow redesign — not workflow overlay. Both McKinsey and Gartner are explicit: the highest returns come from rebuilding workflows around agents, not from layering agents onto legacy processes.
- Talent and operating model. KPMG reports that 64% of organisations have already altered their entry-level hiring approach due to AI agents, up from 18% one quarter earlier1. The skills shift is real and accelerating.
- Cybersecurity and identity. Agents that act on behalf of users introduce a new identity and access surface. Treat agents as first-class identities with scoped permissions, not as anonymous services.
- Measurement and ROI discipline. Tie every agent use case to a financial or operational KPI on day one. The "20-point gap" between efficiency and profit reported by PwC7 reflects an absence of this discipline.
- Strategic ambition. McKinsey's data is consistent: organisations that set growth and innovation targets — not only cost-reduction targets — capture significantly more value from AI2.
What "ready" looks like in practice
Readiness is observable. Organisations that are ready typically share five characteristics:
- One named, end-to-end agent workflow in production — not twenty proofs-of-concept.
- A documented data and tools registry that the agent can call, with versioning and access controls.
- An approvals matrix by risk tier — what the agent can do autonomously, what requires human approval, and what is forbidden.
- A measurement framework that tracks adoption, output quality, latency, and the business KPI the agent is expected to move.
- Executive sponsorship with clear accountability for the agent's performance and risk posture.
Deloitte's 2026 enterprise AI research is direct on this last point: enterprises in which senior leadership actively shapes AI governance achieve significantly greater business value than those that delegate the work to technical teams alone6.
The cost of waiting
The temptation to wait until the technology stabilises is understandable but expensive. KPMG reports that 93% of leaders agree generative AI investments to-date have already enhanced their company's competitive position, with planned investments rising to nearly $114 million per organisation in the year ahead1. Deloitte reports that worker access to AI rose by 50% during 2025, and that the number of companies with 40% or more of AI projects in production is set to double in six months6.
The window between "early adopter" and "permanent laggard" in agentic AI is closing faster than in any prior enterprise technology cycle. As Gartner has noted, the C-suite at software organisations has a three-to-six-month window to define agentic AI strategy or risk being outpaced3.
The bottom line
AI agents are redefining how businesses scale. The technology works — the published case studies are clear. What is also clear, from McKinsey, Gartner, KPMG, Deloitte, and PwC alike, is that the operating environment around the technology is what determines whether scale produces value or chaos. The 40%+ project cancellation rate Gartner forecasts is not a verdict on agents. It is a verdict on the readiness of the organisations deploying them. The organisations that win the next 24 months will not be the ones with the most agents in production. They will be the ones with the most prepared environments to receive them.