Most organizations don't actually know how many AI agents are running across their business right now. A new research initiative from Improv, an Austin-based advisory firm, is launching the first industry-wide benchmark to measure AI adoption in human resources and workforce management, revealing a critical gap: companies are deploying AI faster than they can govern it.

What's Actually Happening With AI Adoption in HR and Workforce Management?

Improv has launched "The State of AI in HCM and WFM," an anonymous survey designed to create the first credible, cross-platform picture of where AI adoption actually stands in human capital management (HCM) and workforce management (WFM) systems. The research addresses a problem that has plagued enterprise leaders: no publicly available benchmark exists that shows real adoption rates broken down by company size, industry, and platform.

The survey targets HR professionals, HCM practitioners, and operations leaders across organizations using major platforms including UKG, Workday, ADP, and SAP. It covers 16 questions and takes approximately three minutes to complete, with all responses remaining fully anonymous and reported only in aggregate.

"AI is moving through HR, HCM, and Workforce Management faster than almost any other part of the business. Some organizations are already capturing real advantage from it. Others are still deciding where to start. The honest answer is that nobody has published a clear picture of where the market actually stands, broken out by company size, industry, and platform. That is the gap this research closes," said Vince Jackson, President of Improv.

Vince Jackson, President of Improv

Why Can't Enterprises See Their Own AI Deployments?

The visibility problem extends far beyond HR. Microsoft's recent analysis reveals what it calls "Agent Sprawl," a phenomenon where AI agents are being created across multiple platforms and departments without centralized oversight. Agents are being built inside Copilot Studio, Azure AI Foundry, partner platforms, open-source frameworks, departmental initiatives, hackathons, and citizen development programs, often operating outside formal governance structures.

According to Microsoft's assessment, most organizations don't have an AI problem; they have a visibility problem. When asked whether they can confidently tell how many AI agents are running across their organization, very few answer affirmatively. This represents what Microsoft describes as the first enterprise AI crisis, one that emerges not from deployment failures but from a fundamental inability to see what's already running.

The challenge is particularly acute because AI agents are fundamentally different from traditional software. Unlike applications or cloud infrastructure, agents possess identities, access enterprise systems, invoke tools, collaborate with other agents, execute workflows, and make decisions autonomously. Their behavior can also change over time, adapting based on new context and data, making traditional IT governance models insufficient.

What Key Areas Will the Improv Survey Address?

• Production vs. Pilot Status: Identifying which AI use cases are running in production versus still in pilot phases across different organizations and platforms.
• Platform-Specific Adoption Rates: Breaking down adoption differences across UKG, Workday, ADP, and SAP to show how AI readiness varies by system.
• Adoption Barriers: Understanding what is actually slowing teams down, including budget constraints, skills gaps, and trust issues.
• Governance Lag: Measuring how far governance policies are trailing actual deployment on the ground.

The survey will produce six specific outputs designed to help leaders understand and optimize their AI investments. These include a real baseline on AI adoption segmented by size and industry, platform-level data showing user perception on AI readiness, identification of the dominant barrier slowing adoption, a quantified picture of how far governance policy is trailing deployment, a ranked list of problems practitioners most want AI to solve, and a shared industry report available to all participating organizations.

How to Prepare Your Organization for AI Governance

• Conduct an AI Inventory: Start by discovering and documenting every AI agent currently running across your organization, including shadow AI systems built by departments without IT approval.
• Assign Clear Ownership: Establish trusted identities and designate clear ownership for each AI agent, ensuring accountability and governance responsibility.
• Establish Governance Policies: Develop policies that govern permissions, data boundaries, and agent behavior before deploying additional AI systems, rather than retrofitting governance after deployment.
• Monitor Agent Behavior: Implement continuous monitoring of agent behavior, outcomes, and anomalies to catch problems early and measure business value.
• Plan for Lifecycle Management: Create processes for managing the full agent lifecycle from design through retirement, ensuring agents don't become orphaned systems.

Microsoft has introduced Agent 365, a control plane designed to address the visibility and governance gap. Launched in May 2026, Agent 365 extends Microsoft's existing identity, security, compliance, and management capabilities to AI agents, providing centralized discovery and registry, policy-based governance, and cross-platform visibility. Notably, it reaches beyond Microsoft platforms to govern agents built on AWS Bedrock, Google Gemini, LangChain, and OpenAI.

The Agent 365 Agent Registry surfaces unmanaged local agents discovered automatically by Microsoft Defender, Entra, and Intune working together, covering more than 35 known agent types including coding agents, AI desktop applications, and local and remote MCP servers. This capability directly addresses the visibility problem by surfacing agents IT never knew existed.

The broader implication is clear: as AI agents proliferate across enterprises, the organizations that will succeed are those that combine technology solutions with a deliberate operating model for AI governance. IDC projects more than 1.3 billion active enterprise agents by 2028, making the visibility and governance challenge increasingly urgent.

For HR and workforce management leaders, the Improv survey offers a rare opportunity to benchmark their organization's AI adoption against peers in their industry and company size. Respondents who provide a business email receive the finished benchmark report before it goes public, giving early adopters a competitive advantage in understanding where their AI investments stand relative to the market.