The Hype Cycle: Agentic Tools Everywhere, Productivity Gains Nowhere
Agentic AI has become a marketing staple across project management and service platforms, but measurable AI productivity gains remain rare. McKinsey’s research shows access to AI project management tools growing 50% year on year, yet only 1% of companies describe themselves as mature in AI deployment. Even at the executive level, just 19% of respondents report revenue increases above 5% from AI, and only 23% see any cost improvement. Meanwhile, platforms like Monday.com, Asana, ClickUp, and Adobe Workfront are rebranding around AI agents that join meetings, create tasks, and execute workflows. In theory, they promise fewer status updates, less manual coordination, and earlier risk detection. In practice, most enterprises are still stuck in pilot mode, struggling to move beyond experimentation into scaled AI agent deployment. The result is a widening gap between ambitious roadmaps, aggressive vendor messaging, and the day-to-day realities of enterprise AI readiness.
Why Enterprise AI Readiness Is Lagging Agentic ITSM Innovation
In IT service management, leading vendors are already shipping agentic ITSM automation tools, but adoption is lagging. Ivanti’s autonomous service desk agent can create incidents, submit requests, and mine knowledge bases without analyst intervention. McKinsey cites a multinational that automated up to 80% of roughly 450,000 annual tickets and redeployed half its service team after moving to agent-led resolution, achieving a customer satisfaction score of 4.8 out of 5. Yet such outcomes hinge on re-engineered workflows and customer journeys designed for agents from the outset. Most organisations are nowhere near that level of readiness. McKinsey reports 62% are still in the piloting stage, with no more than 10% scaling agents in any single function. Their infrastructure—the “plumbing”—remains built for ticket-based, human-led processes, not for agents that must traverse systems, act via APIs, and operate under strict governance.
The Hidden Cost of Building Internal Agentic AI Platforms
Many technology teams respond to the agentic AI wave by building their own platforms, assembling open-source models, orchestration layers, and internal gateways. This instinct to build fuels experimentation and learning, but it also recreates the fragmented toolchain problems seen in early DevOps. Organisations end up managing numerous point solutions that were never designed to work together, diverting engineering effort from business outcomes to glue code and integration. In regulated industries, the complexity multiplies. Building means owning the full stack: agentic frameworks, governance controls, compute, storage, databases, and networking. The organisation effectively becomes the platform vendor, responsible for unifying models, tools, and policies across the software lifecycle. By contrast, buying a purpose-built platform turns the organisation into a platform consumer, leveraging pre-integrated orchestration and governance. The real challenge isn’t the model itself but the orchestration logic that decides which tools to invoke, in what sequence, and under which constraints.
From Task Automation to Knowledge Work Automation
The core reason many AI agent deployment efforts stall is that enterprises are still optimising for task automation while the real bottleneck is interpretation. Traditional automation excelled at structured workflows and repetitive tasks, but modern knowledge work is dominated by unstructured information across emails, documents, and disparate systems. Studies indicate professionals spend more than half their time searching for and processing information, while more than eighty percent of enterprise data is now unstructured. Agentic AI and knowledge work automation aim to close this gap by applying AI to decision-making itself. Instead of just executing predefined steps, these systems understand context, synthesise data, surface insights, and recommend actions. Generative AI interfaces make interaction more intuitive, while intelligent agents learn from outcomes over time. Success, however, depends on embedding these capabilities into real workflows, not treating them as sidecar assistants bolted onto legacy processes.
What Actually Works: Context, Integration, and Pragmatic Scaling
The enterprises seeing real AI productivity gains share several patterns. First, they treat infrastructure and workflow redesign as prerequisites, not afterthoughts. In the ticket automation example, agents succeeded because customer journeys and processes were rebuilt around agent-led resolution. Second, they prioritise context-rich integration over flashy features. Platforms like Monday.com and Asana are re-architecting data models so agents can understand dependencies and execute work end-to-end, rather than generating isolated suggestions. Third, they avoid sprawling DIY platforms in favour of cohesive architectures—whether built or bought—that centralise orchestration and governance. Finally, they scale deliberately, moving beyond endless pilots by targeting specific high-volume, rules-light domains such as ITSM and project coordination, then iterating from there. Agentic AI adoption will not be unlocked by more tools alone; it will come from enterprises that align plumbing, data, and decision-making flows so agents can participate meaningfully in how work actually gets done.