Agentic AI demands a new operating model

Most firms want AI agents to run end-to-end, but their systems aren’t ready.

A wave of agentic AI is hitting enterprise doors, promising to coordinate tasks, adjust to changing conditions, and execute workflows with limited human input. Yet a chasm separates ambition from capability: 85% of organizations say they want to be agentic within the next three years, while 76% admit their current operations and infrastructure can’t support that shift. That mismatch is more than a budgeting problem; it’s a design problem. PwC UK Consulting’s Prasun Shah calls the current approach “embedding AI employees into what is a human operating model,” a strategy that often feels like gluing sticky tapes on a house whose wiring is already breaking. The risk is collective disillusionment when agents fail to deliver end-to-end value.

The core insight from industry voices is blunt: the real payoff lies not in sprinkling AI across existing processes, but in reimagining the operating model so agents can autonomously manage whole workflows. When agents can coordinate tasks across functions, make independent decisions, and iterate performance, organizations unlock a much larger lever for change. Early proof points are encouraging in customer service, HR, and sales, with estimates that AI agents could accelerate business processes by 30% to 50% and reduce low-value work time by 25% to 40% when scaled, yet only if the orchestration and governance keep pace with automation.

That framing has given rise to what Ema calls agentic business transformation, or ABT. The term, developed with HFS Research, signals a deliberate shift from tinkering with AI features to redesigning how work flows across people, processes, and platforms. In practice, that means enterprise platforms must evolve from additive AI widgets to systems that coordinate across data sources, human agents, and policy rails. The technical report behind these ideas emphasizes not just capabilities, but the governance, data quality, and workflow design required to sustain them at scale.

Analysts and practitioners are circling two big themes. First, readiness is a bottleneck. Even with exciting demonstrations, most organizations still struggle to align people, processes, and workflows with a primarily autonomous operating model. Second, the value equation is real but contingent. The gains cited hinge on comprehensive redesign, without it, benefits erode as friction points, data silos, and inconsistent decision rights pull agents off course. The promise is large, but it’s not a plug-and-play upgrade. It’s a strategic overhaul.

For product teams and platform builders, the quarter ahead will likely tilt toward enabling end-to-end workflow orchestration rather than standalone agent features. Vendors will need to foreground governance, cross-system orchestration, and policy-driven control that keep autonomous agents safe and compliant. Customers will demand capabilities that allow ABT pilots to scale with clear ROI signals, not just pilot-test results. The risk to watch: partial adoption that yields partial gains but ignores the broader organizational rewiring, leading to the same old bottlenecks resurfacing as soon as pilots end.

Analogy helps here: trying to run a city with a new power plant while you still lay down old, tangled wiring. The plant can generate abundant energy, but without rewiring the grid, voltage spikes, outages, and risky cross-currents will undermine every smart outage alert or automated service desk. In other words, agentic AI’s value depends on a holistic, enterprise-wide upgrade where governance, workflows, and culture move in lockstep with technology.

If you’re shipping this quarter, expect a sharper focus on orchestration frameworks, cross-functional workflows, and governance mechanisms that enable agent autonomy without losing sight of risk, compliance, and human-in-the-loop boundaries. Pilot programs will increasingly measure end-to-end impact across departments, not just isolated tasks, and leadership will push for a shared blueprint, an ABT blueprint, that turns promise into measurable, scalable performance.