Enterprises Aim for Agentic AI but 76% Say Operations Can't Support the Shift

A surge in enterprise plans to deploy agentic AI is colliding with a major readiness shortfall: 85% of organizations say they intend to be agentic within three years, while 76% report that current operations and infrastructure cannot support that shift. That gap forces companies to redesign operating models rather than simply bolting agents onto existing processes if they want agentic systems to meet expectations. Prasun Shah of PwC UK Consulting warns this often amounts to treating AI like “sticky tape” applied to a breaking operating model.

Vendors and researchers are responding with new frameworks aimed at systemic change. Last year Ema, together with HFS Research, coined the term “agentic business transformation” (ABT) to describe integrating AI agents across the organization instead of merely adding intelligence to legacy workflows. Ema CEO Surojit Chatterjee contrasts ABT with digital transformation and co-pilot models, saying ABT embeds agents into the fabric of work. Ema describes ABT as resting on three core pillars — the technology stack, the workforce, and the metrics used to judge success — and Shah says it requires redesigning operating models, workflows, decision rights, and performance — management systems so agents become active participants in value creation.

The technical role of agents is central to the argument for redesign. Agents can execute end-to-end workflows with limited human input: coordinating complex tasks, making independent decisions, adapting to changing conditions, and iterating their performance. Chatterjee emphasizes that agents act as connective tissue, moving between layers and across multiple systems to retrieve, contextualize, and act on data held in discrete applications.

Early deployments point to measurable benefits alongside clear risks. In initial proving grounds such as customer service, HR, and sales, agentic deployments are estimated to accelerate business processes by 30 — 50% and cut time spent on low-value work by 25 — 40% when scaled. But simply layering agents onto unchanged operations can produce disappointment and underdeliver on those potential gains.

For builders and architects the practical implication is an architectural and organizational shift: expose agents to multiple datasets and applications, design systems to surface higher — quality decisions, and enable tacit knowledge acquisition. Chatterjee warns that organizations that make this shift become genuinely more adaptive — a change that will determine whether agentic AI delivers competitive differentiation or remains confined to limited assistance roles.