Reality check on AI deployment and maturity
Yet only a fraction of organizations have successfully leveraged the technology to achieve notable business results and measurable returns.
What’s slowing momentum? The answer is that enterprises are chasing the promise of AI-enabled intelligence and automation without radically rethinking work patterns.
For years, every department had its own system — Salesforce for sales, Workday for HR, Zendesk for support, and others. There hasn’t been a connecting solution to bring everything together, which has resulted in a patchwork enterprise that’s so fragmented it no longer functions as one. Companies rushed to fill the void with AI copilots bolted on to departmental SaaS systems. But these stopgap measures simply layer shallow intelligence on top of broken processes, only marginally moving the needle on productivity and efficiency gains. The initiatives were never designed to drive AI workflows and measurable ROI.
The problem isn’t lack of AI intelligence, it’s that current systems aren’t architected for AI action. AI is proficient at answering questions, generating content, or summarizing incidents. But most implementations aren’t mature enough to do the work of enterprise business. In fact, enterprise AI maturity has declined by 20%, gated in part by insufficient governance policies and controls, according to the ServiceNow Enterprise AI Maturity Index 2026.
AI models alone don't understand enterprise context or governance. They require enterprise systems to provide permissions, policy enforcement, and workflow execution. They need a reasoning layer that can interpret intent, plan across systems, and activate work within enterprise controls. As a result, AI doesn’t fix the chaos of fragmented systems and disconnected data; it simply exposes the dysfunction while potentially derailing desired gains.
The disconnect has major ramifications for business. The average enterprise runs 367+ applications across the employee experience alone, each with its own data model, security perimeter, and governance logic. With an estimated 1.3 billion AI agents expected on the job by 2028, according to IDC estimates, AI sprawl is set to become the new shadow IT. Copilots, AI models, agentic frameworks, and Model Context Protocol (MCP) servers are popping up across Microsoft, AWS, Google Cloud Platform, SaaS apps, and internal systems without the benefit of unified visibility and control.
Visibility and governance of agentic AI is a huge problem. Most organizations’ practices remain manual, inconsistent, and in many cases, non-existent. Many companies also lack an automated workflow that connects all stakeholders involved in planning and deploying AI. While close to three-quarters of companies are planning to deploy agentic AI within two years, only 21% report having a mature model for agent governance, according to Deloitte’s 2026 State of AI in the Enterprise report. Without that structure, organizations risk exposure and AI ventures may stall as it’s impossible to manage what they can’t see.
“AI maturity has gone backwards,” says Sean Regan, senior vice president of Product and Solutions Marketing at ServiceNow. “It’s not because AI isn’t powerful; it’s because enterprises are deploying AI on broken foundations.”
of respondents do not have a mature model for agent governance
Enterprise AI maturity has declined by
