Skills aren’t blocking agentic AI adoption; operating conditions are 

This HFS Data Viewpoint is for service provider leaders and enterprise transformation executives diagnosing why agentic AI delivery stalls despite a trained and capable workforce.

HFS Research surveyed 93 employees of service providers that enterprises rely on for agentic AI delivery. The reality is uncomfortable: their capabilities are largely in place, but the operating conditions required to turn a trained workforce into a productive one are missing. These include protected experimentation time, production-grade tools and platforms, governance clarity, and visible recognition for agent builders.

All 93 respondents are senior service provider practitioners (86% have more than 10 years of work experience) interviewed in the Agentic Services Horizon, 2026 report. They are spread across three role bands: strategy and governance leaders (30%), project leads who deliver agentic AI services (34%), and builders who design and configure agentic systems (28%). All respondents answered both questions, so the chart reflects a layered view from leaders, delivery leads and practitioners, not two separate audiences.

Around 92% of employees say they received formal agentic AI training in the last 12 months, yet 90% say they would benefit from additional training. Capability is not the limiting factor. The delay in value delivered comes from capacity, complexity, and tooling, the same constraints that show up in enterprise teams running their own agentic programs. Enterprises that continue to overinvest in training rather than operational enablement risk slowing ROI from AI by attacking a bottleneck that is no longer the primary constraint.

Here are the top six takeaways from the above two charts:

• Talent capacity, not capability, is the bigger bottleneck. Across all three role bands of employees, technical complexity (52%) and lack of time (42%) are the top two barriers in agentic AI adoption. Skill gaps rank lowest at just 4%. For enterprises and service providers alike, the emerging constraint is delivery bandwidth, not access to talent.
• Limited tool access blocks the work itself. Twenty-two percent of practitioners cite limited access to tools and platforms as the third-largest barrier. Project leads (the role responsible for delivery) flag limited tool access roughly four times more often than strategy leaders or builders. Access to production systems decides whether agentic AI work scales or stalls.
• Governance ambiguity is a quiet throttle. Fourteen percent of practitioners cite unclear organizational policies and governance as an adoption barrier, ahead of peer resistance (11%). Without clear rules on what’s allowed and how decisions get audited, delivery teams hedge.
• Continuous and advanced training is the most-prescribed fix. Twenty-three percent of practitioners want their organization to invest in training on best practices, industry use cases, and foundational agentic AI as the top adoption lever. Employees want sharper and more applied training, but training alone will not remove the operational bottlenecks slowing adoption.
• Workforce motivation is emerging alongside technology as an adoption lever. Fourteen percent of practitioners want their organization to invest in incentives, recognition, and promoting internal wins. And another 8% want their organizations to frequently share experiences and learnings across the enterprise.
• Centralized AI structures get the least support. Only 2% of practitioners want a dedicated CoE or hub, and 3% want expanded partnerships. These are the two lowest-ranked items in the data. Practitioners indicate appetite for distributed enablement rather than additional centralized AI infrastructure.

The Bottom Line: Enterprises that fund the operating conditions (i.e., time, tools, governance clarity, and recognition) will be able to convert agentic AI capability into competitive advantage. Those that keep funding more training instead of operational enablement will continue delaying AI returns despite having trained workforces already in place.