44% of Enterprises Now Scaling AI Across Functions, Up From 38% Last Year

Enterprise AI Is Scaling, But Most Deployments Remain Stuck in Pilot

McKinsey's 2026 global AI survey shows 44% of enterprises now report AI scaling across their organization, up from 38% a year earlier. More tellingly, 56% say AI is deployed in three or more business functions. That is the clearest signal yet that enterprise buying is shifting from isolated experiments to platform decisions with cross-functional budgets and governance requirements.

The shift matters because it changes what buyers need to procure. When AI was confined to a single department, a narrow copilot and a few API keys sufficed. When it spans customer service, software development, and knowledge management, enterprises need orchestration layers, unified admin controls, and data access policies that work across functions. That environment favors broad platform vendors like Microsoft, Google, OpenAI, and Anthropic, as well as enterprise AI search and orchestration products like Glean. It creates headwinds for point-solution copilots that do not integrate across workflows.

Worker Access Rose 50%, But Production Deployment Lags

Deloitte's 2026 enterprise AI report shows worker access to sanctioned AI tools jumped 50% in 2025, moving from under 40% of employees to about 60%. But only 25% of respondents have moved 40% or more of their AI experiments into production. That gap between access and operationalization explains why CFOs are scrutinizing renewals more closely.

The implication: enterprises are paying for licenses without yet seeing broad operational impact. Buyers should expect procurement pressure to shift from buying model access to funding deployment infrastructure, policy controls, and change management. Vendors that sell chat interfaces without deployment tooling or workflow integration are less likely to survive renewal cycles than those with admin controls and production-ready connectors.

Microsoft's $30 Per User Per Month Remains the Pricing Benchmark

Microsoft 365 Copilot remains publicly priced at $30 per user per month, making it the most visible benchmark for enterprise AI seat pricing. Competitors including Google Gemini for Workspace, ChatGPT Enterprise, Anthropic Claude for enterprise, and Glean must compete against that anchor even when they use quote-only or usage-based models.

Buyers can use Microsoft's sticker price as a reference point, but should budget above it. Enterprise deployments require base licenses, admin overhead, and integration work. The real cost of scaling AI across functions typically runs higher than per-seat pricing suggests, especially when factoring in data preparation, governance tooling, and workflow redesign.

Budget Pressure Is Rising Faster Than Proof of Value

One survey this week found 59% of companies investing at least $1 million per year in AI, but only 29% report significant returns. That mismatch between spending and measurable outcomes is tightening renewal scrutiny. Enterprise buyers should expect more demanding business cases, tighter pilot-to-production conversion metrics, and stronger interest in outcome-based pricing before approving broader rollout.

The environment rewards vendors with clear ROI stories and measurable outcomes, particularly in customer service, knowledge management, and software development. It weakens undifferentiated model wrappers and generic AI assistants that cannot prove workflow-level impact.

What to Watch

Expect procurement to consolidate around fewer vendors with broader platform coverage. Enterprises scaling AI across functions are less willing to manage dozens of disconnected tools. That trend will accelerate M&A among AI vendors and increase pricing pressure on products that cannot demonstrate cross-functional value.

Buyers should also prepare for governance to become a gating factor. Scaling across functions exposes security and compliance gaps that were tolerable in single-department pilots. Products with weak admin controls, unclear data handling policies, or poor auditability will face harder renewal conversations. The market is moving past the "try it and see" phase into governed enterprise rollout, and vendors without enterprise-grade deployment infrastructure will struggle to retain accounts.