- Gartner projects agentic AI spending will jump 139% to $206.5 billion in 2026, yet forecasts more than 40% of those projects will be canceled by late 2027 due to unclear ROI and weak risk controls.
- Only 38% of organizations had a formal AI governance policy as of May 2026, according to Managed FinOps, and 63% of firms that suffered AI-related breaches lacked one.
- BCG’s Applied AI Index, published September 2026, found nearly half of companies now generate meaningful value from AI, up from roughly 5% in 2025, with gains concentrated in narrow, well-instrumented use cases.
On October 5, 2026, PwC and AI company Cohere announced a global alliance to help organisations deploy secure enterprise AI and translate investments into measurable business value. The announcement lands at a moment when the central enterprise AI problem has shifted: whether AI models can outperform traditional forecasting is largely settled. The harder questions now are whether organisations can govern these systems, integrate them at scale, and prove a return.
Governance Demands Formal Policy
AI adoption has outpaced governance implementation by a wide margin, and the gap carries direct business risk. As of May 2026, only 38% of organisations had established a formal AI governance policy, according to Managed FinOps. Of those that experienced AI-related breaches, 63% lacked such a policy and 97% did not have proper AI access controls in place. As AI embeds itself in critical business functions, AI risk and business risk are converging.
Regulations including the EU AI Act, the NIST AI Risk Management Framework and ISO/IEC 42001 are converting “ethical AI” from a discussion point into a compliance requirement. Spending on dedicated AI governance platforms is projected to reach $492 million in 2026, according to Gartner, reflecting demand for tools that manage AI inventories, assess risks, enforce policies and maintain compliance evidence.
The ROI Problem
McKinsey data for 2026 shows 88% of organisations use AI in at least one business function. The returns, however, are concentrated in a small cohort. BCG and KPMG surveys independently put the share of companies reporting measurable, at-scale ROI at between 5% and 8%. An MIT finding that 95% of generative AI pilots yield zero measurable P&L impact points to workflow integration failures rather than model performance as the primary drag. PwC’s 2026 Global CEO Survey reported that 56% of CEOs see no revenue or cost benefits from their AI investments.
BCG’s Applied AI Index, published in September 2026, offers a counter-reading: nearly half of companies are now capturing meaningful value from AI, up from roughly 5% in 2025. The gains are concentrated among companies that targeted narrow, well-instrumented use cases with clear cost baselines, such as AI agents in customer service or finance. The divergence between the struggling majority and the scaling cohort suggests the bottleneck is execution discipline, not capability.
Agentic AI’s Governance Gap
Agentic AI has moved beyond experimental chatbots to autonomous systems executing complex, multi-step business workflows, and the governance infrastructure has not kept pace. Gartner projects spending on agentic AI software will grow 139% to $206.5 billion in 2026. Yet only 21% of companies report mature agent governance in place, even as nearly 75% plan to deploy agentic systems within two years.
Gartner anticipates more than 40% of agentic AI projects will be canceled by end of 2027, citing unclear value, cost overruns and weak risk controls. A further concern is what a Forbes analysis from October 2026 termed “cognitive concentration”: heavy reliance on shared AI intelligence across critical workflows creates correlated failure modes that standard monitoring may not catch. An IBM Institute for Business Value survey conducted in 2026 found 91% of senior executives admitted to not fully understanding their AI dependencies across vendors, models and infrastructure.
Where Integration Breaks
Gartner research puts the share of AI projects that move from prototype to deployment at 41%. The failures tend to cluster at the execution stage: data pipelines break, models produce biased outputs, or pilots cannot scale to production volumes. Poor data quality and integration are cited as the top obstacle by 52% of businesses surveyed.
Info-Tech Research Group’s blueprint “Define Your Target AI Infrastructure,” released October 5, 2026, addresses this directly, advocating for workload-driven architecture decisions to improve AI performance, scalability and cost control. The talent dimension compounds the technical one. IBM research suggests 29% of employees will need reskilling for entirely new roles between 2026 and 2028, with 53% needing retraining for their current positions. CIOs and Chief AI Officers consistently identify the “frozen middle” of management and entrenched workflows as a harder barrier than the technology itself.
Boards Want Proof, Not Pilots
Boards now expect demonstrable AI ROI within three months, not 18, compressing the timeline for outcome-based measurement. The practical implication is that AI deployments need to be connected to trusted data and systems from day one, with governance and auditability built into every layer rather than retrofitted. Deloitte’s 2026 report found that board-level AI value reporting is currently practiced by only 4% of respondents, yet the firm expects it to become standard for public companies and large enterprises by end of 2026. The 96-percentage-point gap between current practice and expected standard is the clearest indicator of where enterprise AI governance work is still ahead.



