If you needed a number to confirm that the AI agent wave is real and accelerating, KPMG just provided one.

The global professional services firm released its Q2 2026 AI Pulse Survey this week, drawing on responses from 2,145 C-suite and senior business leaders across 20 countries. The organisations surveyed are not experimenting on a shoestring: most generate annual revenues above US $50 million, and a significant share run billion-dollar operations. These are the companies whose decisions set the pace for the rest of the market.

The headline finding is striking. Employee adoption of AI agents reached 56 percent of surveyed organisations this quarter, up from just 23 percent in Q1. That is more than a doubling in a single quarter, covering the period from late April through late May 2026.

Agents Are No Longer a Pilot Project

The 56 percent adoption figure matters, but a second data point tells a more important story. The percentage of organisations actively orchestrating multiple AI agents across workflows doubled from 9 percent to 18 percent in the same period.

There is a meaningful difference between having an AI agent and running several of them in coordinated workflows. Single-agent use is still early-stage: one tool, one task, limited dependencies. Multi-agent orchestration is operational AI, the kind where agents hand work between each other, manage exceptions, and operate without human intervention at each step. Doubling that figure in a single quarter suggests that the companies who started early are now advancing quickly beyond the pilot phase.

Overall agent deployment stayed broadly consistent with last quarter, at 53 percent versus 55 percent. The story is not that more companies are starting, it is that companies already in are going deeper.

Investment Is Holding, But Cost Visibility Is Not

Average AI spending across the survey held steady at US $188 million, a slight increase from $186 million in Q1. More significant: 79 percent of leaders now describe AI as a top investment priority, up from 74 percent in Q1. This number has moved persistently upward over the past year.

What is not moving fast enough is cost visibility. Only 26 percent of organisations report full, real-time visibility into what their AI systems cost to run. Two-thirds have monitoring dashboards in place, and 61 percent have spending approval processes, but the gap between having a dashboard and understanding actual operational costs remains wide. This explains a lot of the frustration showing up in executive conversations about AI ROI: money is going in, but the feedback loops on where it is going are still poor.

The Barriers Organisations Cite

When KPMG asked what is holding back deeper AI agent deployment, the answers aligned closely with what practitioners report on the ground:

• Data readiness — cited by 63 percent as a top barrier
• Complexity of agentic systems — 49 percent
• Human oversight skills — 41 percent
• Workforce resistance to change — 36 percent
• AI cost and economic literacy — 30 percent

These are not primarily technology problems. They are capability and governance problems. The models exist. The platforms exist. The thing that slows deployment is the absence of people who know how to evaluate AI systems, interpret their outputs, manage their errors, and design the workflows that give agents useful work to do.

That data readiness sits at the top of the list is consistent with everything Enterprise DNA has observed across its own client work. Organisations that have invested in their data foundations, clean pipelines, accessible reporting, and staff who can actually read and challenge the numbers, move through AI deployment faster and with fewer expensive detours.

What This Means for Business

The market is bifurcating. The organisations already running multi-agent workflows at 18 percent are compounding their operational advantage every month. The ones still evaluating their first use case are not just behind, they are falling further behind. The gap between the 18 percent and the 82 percent will not close on its own.

Cost visibility is the underrated risk. Companies spending $188 million on average in AI, with only 26 percent having real-time cost insight, are flying partially blind. As agentic workloads scale, the cost curves behave differently from traditional software subscriptions. Understanding token economics, compute costs, and agent-driven usage patterns is not optional; it is the skill set that separates leaders who can sustain AI investment from those who end up with budget crises.

The bottleneck is people, not technology. Data readiness and human oversight skills top the barrier list because AI deployment ultimately depends on teams who know how to work alongside these systems. This is why the upskilling argument has moved from “nice to have” to “operationally necessary.” Organisations that have invested in data literacy and AI capability have a structural advantage that shows up directly in their ability to move from pilot to production.

If your organisation is still on the wrong side of this curve, the practical path forward starts with an honest assessment of where your data, your team capabilities, and your governance actually stand. The Omni Advisory team works directly with business leaders on this kind of diagnostic and planning work. And if upskilling your team on data and AI fundamentals is the priority, Enterprise DNA’s learning platform is built specifically for that.

The Q2 numbers confirm what the Q1 numbers suggested: this is not a gradual trend. The organisations moving are moving fast.