The Ambition of AI Agents and Operational Reality

Integrating AI agents into business operations is becoming a priority for many enterprises, but the path is far from straightforward. While 85% of organizations state their desire to adopt an "agentic" approach within the next three years, a significant portion, 76%, admits that their current operations and infrastructure are not ready to support such a change. This discrepancy highlights a lack of preparedness across people, processes, and workflows, creating a tangible barrier to fully realizing the potential of artificial intelligence.

The core issue, as highlighted by Prasun Shah, global CTO for workforce consulting and Chief AI Officer at PwC UK Consulting, lies in the adopted approach. Many companies tend to layer AI agents onto existing operations, rather than rethinking the operating model and how work needs to be rewired. This is akin to "adding sticky tape to parts of an operating model that is breaking," an approach that prevents organizations from capturing the intrinsic value of AI agents—their capacity to execute entire workflows with limited human input.

Agentic Business Transformation: A New Framework

To address this challenge, the agentic AI platform Ema, in partnership with HFS Research, has coined the term "Agentic Business Transformation" (ABT). This new framework aims to fill a gap in the existing lexicon about AI, offering enterprises a conceptual model to guide technology adoption. Surojit Chatterjee, CEO and founder of Ema, explains that while digital transformation was about moving from paper to software and AI transformation involved adding artificial intelligence to existing processes, ABT represents something categorically different: the integration of AI agents into the very fabric of the organization.

ABT is built upon three core pillars: an organization's technology stack, its workforce, and the metrics used to gauge success. According to Shah, a dedicated term like ABT "helps drive the need to redesign an organization in its entirety: its operating model, its workflows, decision rights, and performance management systems." The goal is to ensure that agents are active participants in value creation, rather than just point tools or productivity aids.

Redesigning the Technology Stack and Workforce

The first pillar of ABT requires a profound overhaul of the technology stack. Existing architectures were designed for human-operated, application-centric workflows. With AI agents operating at machine speed across multiple systems simultaneously, a rethinking is necessary. AI agents should not be just another layer in an existing technology stack, but rather a "connective tissue" capable of moving between or across layers to coordinate complex tasks, retrieve, and interpret data from multiple discrete applications. This capacity for contextualization can generate a significant competitive advantage. For those evaluating on-premise deployments, this implies a re-evaluation of existing infrastructures, potentially requiring investments in specific hardware for inference and integration with legacy systems, to ensure data sovereignty and granular control.

The second pillar concerns the workforce. Traditional hierarchical structures, a legacy of the industrial era, clash with the ability of AI agents to execute, coordinate, and optimize tasks without constant managerial oversight. Managers will need to take on new responsibilities, managing hybrid teams and addressing issues of trust, explainability, and psychological safety. McKinsey predicts that by 2030, three-quarters of current jobs will require redesign, upskilling, or redeployment, making urgent interventions in recruitment, retention, and remuneration.

From Output Focus to Outcome Focus

The third and final pillar of ABT is the redefinition of success metrics. With AI agents assuming greater ownership of core enterprise processes, traditional metrics based on activity or output (such as the number of calls handled or reports filed) lose their meaning. Chatterjee explains that an AI agent can handle thousands of customer interactions in the time it takes a human to handle ten; measuring success solely by interactions could lead to misleading conclusions, ignoring the real impact on customer satisfaction or revenue.

Enterprises must therefore develop new metrics that focus on outcomes rather than mere outputs. For example, an Ema customer tripled their ROI from agentic AI in two quarters by shifting from query-cost-based metrics to those focused on the percentage of contracts reviewed without human escalation. This transition also requires a reconfiguration of reward and talent management processes, as well as clear accountability. While ethical and fiduciary responsibilities will likely remain with human employees, operational accountability will become more diffused, raising new questions for leadership teams regarding managing AI agent errors and establishing safeguards for customers.