What’s holding enterprises back from agentic AI?

Only one in four respondents say AI is integrated across workflows, tools and teams, or is helping coordinate work end to end. At the same time, 77% want AI to play a greater role in connecting work across teams and systems. That is a clear signal: the workforce is ready for more connected ways of working, but most organizations have not yet built the operating conditions to support them.

This is how organizations drift into pilot purgatory. Pilots prove that AI can create value in a controlled context, but they do not prove that the enterprise is ready to scale that value through live systems, real governance, risk controls, dependencies and accountability. Experimentation is necessary, but it becomes limiting when it replaces the harder work of redesigning how work is planned, connected, approved and measured.

The more useful leadership question is no longer, “Where are we using AI?” It is, “Where does AI-created value get stuck?” In many organizations, the answer sits between teams, tools, approvals, data flows and decision rights that were never designed for agentic execution.

Agentic AI barriers: the system around the tool is the bottleneck

The biggest agentic AI barriers are not only technical. They are structural, operational and organizational. Enterprises rarely stall because they lack another AI tool. They stall because AI is being introduced into systems of work that were not designed for agents to coordinate, execute or escalate across them.

Agentic AI creates value when it has context, permission and connection. It needs access to the right data. It needs to operate across tools. It needs clear rules for when it can act and when it must escalate. It needs workflow ownership, governance, enablement and metrics that tell teams whether the work is actually improving. Without those conditions, AI remains useful but fragmented.

That fragmentation is already visible in many enterprises. One team uses AI for documentation. Another uses it for analysis. Another uses it for code generation, testing, content or reporting. Each use case may be valuable, but the organization still lacks a connected system of work. The value is real, but it remains local.

The workforce is already signaling what needs to change. When asked what would unlock the most value from AI, respondents prioritized integration of tools and data, governance and guardrails, and clearer human-versus-AI decision rights. Only 4.3% believe more AI pilots would unlock the most value.

That is a strikingly practical message. Teams are asking for the foundations that allow AI to work consistently where it already makes sense. They want the enterprise to connect the capabilities that already exist, rather than keep adding disconnected experiments.

Agentic AI risks: autonomy without accountability

The agentic AI risk conversation often starts with security, privacy, compliance, hallucination, bias and model reliability. But one of the most important risks is a management risk: autonomy without accountability.

As AI becomes more agentic, the problem is whether the organization has defined where it should act, where it should recommend, where it should escalate and where humans must retain final authority.

The report shows that teams are not rejecting AI taking on more work. In fact, 74% are comfortable with AI taking on bounded execution if humans remain accountable. At the same time, respondents draw clear lines around final approval, people decisions, strategic decisions and ethical or risk-sensitive decisions, all areas where human control remains highly valued.

This is not resistance to AI. It is a sophisticated view of where judgment still matters. Teams are increasingly comfortable with AI handling execution, coordination, repetition and preparation. They are far less comfortable handing over decisions that affect customers, employees, risk, ethics, architecture, strategy or accountability.

Agentic AI requires decision-specific governance designed into the workflow itself. Every important workflow should make clear what AI can access, where it can assist, where it can act, when it must escalate, what gets logged, who reviews outcomes and who remains accountable when something goes wrong. When governance is built into the workflow, it creates the confidence teams need to move faster.

Agentic AI opportunities: remove friction, unlock value

Removing the friction that prevents organizations from adapting, deciding and delivering at speed is critical.

The report shows that professionals want AI to assist where work gets stuck. 82% say at least one-fifth of team time is spent on repetitive, rule-based work that could be automated. The strongest demand is for AI to remove repetitive steps, speed execution, improve quality and connect tasks, tools and teams.

Every enterprise has an invisible backlog of friction: repeated checks, slow approvals, duplicated reporting, etc. These are the places where speed, quality and responsiveness are lost. Agentic AI can help because it is suited to the connective tissue of work such as summarizing, monitoring and escalating. But the value only compounds when leaders redesign workflows around the moments where AI can remove drag.

The opportunity goes beyond efficiency. 79% of respondents want more time for judgment and higher-value work. When asked how they would use time freed by AI, they pointed to strategy and direction setting, quality improvement, better decision-making, experimentation, innovation and customer understanding.

That is the real productivity dividend. More capacity for the work that differentiates the business. If AI removes low-value execution and coordination, but the organization fills that time with more meetings, low-priority tasks or reporting, the larger opportunity disappears.

This is where agentic AI becomes a growth conversation, not just an efficiency conversation. The report found that 89% believe better AI-enabled workflows would expand what their organization can deliver. The prize is a more adaptive enterprise, where information moves faster, decisions are better supported and people have more space to apply judgment where it matters.

What leaders should do now

The next phase of agentic AI will be won by organizations that turn AI from scattered assistance into governed execution, treating today’s blockers as a roadmap for how the enterprise needs to evolve.

• Start with the places where work slows down

These workflow moments reveal where agentic AI can create immediate momentum without becoming another disconnected layer.

• Define the authority boundary

Make explicit where AI can assist, where it can act, where it must escalate and where humans retain final decision rights.

• Connect tools, data and workflows

The goal is not another point solution, but a connected system where AI can operate across the flow of work.

• Turn saved time into strategic capacity

Leaders need to deliberately redirect capacity toward quality, experimentation, innovation, decision-making and customer understanding.

• Measure AI by outcomes, not activity

Measure AI maturity by quality, speed, decision-making, delivery capacity, adaptability, cost-to-serve, customer relevance and business-goal achievement.

From AI use to AI advantage

Enterprises need to move from scattered AI use to scalable advantage by designing the operating model around AI: integrated tools and data, clear decision rights, workflow redesign, governance, enablement, metrics and ownership. This is how isolated productivity gains become a more durable source of business performance.

The organizations that lead will redesign how people, agents, systems and decisions work together.