NextFin News - A stark disconnect has emerged between corporate ambition and operational reality as enterprises rush to deploy autonomous artificial intelligence agents. While eighty-five percent of organizations plan to adopt agentic AI within the next three years, a striking seventy-six percent admit their current infrastructure and workflows are entirely unprepared for the shift, according to a study by process-mining firm Celonis. This gap highlights a fundamental design flaw: companies are attempting to shoehorn autonomous digital workers into organizational structures built for humans, risking operational friction rather than achieving the promised productivity gains.
Prasun Shah, the global chief technology officer for workforce consulting and chief AI officer at PwC UK Consulting, describes this widespread practice as a temporary and fragile fix. Shah, who has long advocated for systemic organizational restructuring over piecemeal technology adoption, argues that embedding AI employees into a traditional human operating model is akin to adding sticky tape to parts of an operating model that is already breaking. In his view, the true value of agentic AI—which lies in its ability to execute entire workflows, make independent decisions, and coordinate complex tasks with minimal human intervention—cannot be unlocked without a complete overhaul of how work is designed and managed.
The financial stakes of resolving this operational bottleneck are substantial. Research by the Boston Consulting Group estimates that deploying AI agents at scale could accelerate business processes by thirty to fifty percent, while reducing the time spent on low-value tasks by twenty-five to forty percent. Yet, achieving these efficiencies requires moving beyond simple automation. Surojit Chatterjee, the chief executive officer and founder of enterprise platform Ema, argues that organizations must transition toward what he terms agentic business transformation. Chatterjee, whose firm specializes in deploying AI employees across corporate systems, distinguishes this shift from previous waves of digitization. While digital transformation moved paper to software, and co-pilots assisted humans with specific tasks, agentic transformation integrates autonomous agents directly into the fabric of the enterprise.
This integration requires a fundamental redesign of three core pillars: the technology stack, the workforce, and success metrics. Traditional corporate technology stacks were designed for human-operated, application-centric workflows where employees manually move data between separate software programs. AI agents, however, operate at machine speed across multiple systems simultaneously. To leverage this capability, companies must treat AI agents as a connective tissue that spans different software layers to retrieve, interpret, and act on data. Chatterjee notes that when organizations make this architectural shift, the time required to configure and deploy a new business workflow drops from months to days, allowing companies to bypass the lengthy development cycles of traditional software vendors.
Redesigning the workforce presents an even greater cultural and structural challenge. Modern corporate hierarchies, which have changed little since the early days of industrialization, rely on standardized processes and clear divisions between strategic business units. Autonomous agents disrupt these boundaries by executing and optimizing tasks without requiring traditional managerial coordination. Consequently, the role of middle management must shift from task supervision to the oversight of hybrid human-AI teams. Managers will need to navigate complex dynamics surrounding trust, explainability, and psychological safety within their teams. The scale of this transition is immense; McKinsey projects that by 2030, three-quarters of current jobs will require significant redesign, upskilling, or redeployment.
Perhaps the most immediate operational hurdle is the revision of performance metrics. Traditional key performance indicators focus on volume-based outputs, such as the number of customer service calls handled or reports generated. When applied to AI agents, these metrics become obsolete. An AI agent can process thousands of transactions in the time a human handles a dozen, but high volume does not guarantee customer satisfaction or revenue growth. Chatterjee points to an Ema client that tripled its return on investment within two quarters simply by shifting its metrics from tool-based measures, like cost per query, to outcome-based measures, such as the percentage of contracts reviewed without requiring human escalation. This shift forced the company to deploy AI where its decision-making capacity added the highest strategic value.
Despite the optimistic projections from technology vendors, significant hurdles remain, and some industry observers urge caution. Critics point out that diffusing operational accountability across hybrid teams introduces unprecedented legal and fiduciary risks. While ethical and legal responsibilities ultimately rest with human employees, determining liability when an AI agent makes a costly operational error remains an unresolved challenge. Furthermore, the high capital expenditure required to rebuild legacy IT infrastructure means that smaller enterprises may find themselves priced out of the initial wave of agentic transformation, widening the competitive gap between industry giants and smaller players. Whether organizations can successfully manage these risks while retraining millions of workers remains an open question.
For now, the transition remains in its infancy. As corporate leaders grapple with these systemic questions, the immediate challenge is to move beyond pilot programs and begin the difficult work of organizational redesign. PwC's Shah emphasizes that senior leadership must actively address who is accountable when an AI employee makes a mistake, and how to establish robust guardrails to safeguard customers. The speed at which companies can resolve these structural questions will ultimately determine whether agentic AI becomes a genuine engine of productivity or merely another expensive, taped-on corporate experiment.
Explore more exclusive insights at nextfin.ai.
