The Automation Paradox: How AI Facilitates Economic Coexistence

In a groundbreaking analysis published in Harvard Business Review, Sangeet Paul Choudary, a senior fellow at the University of California, Berkeley, argues that artificial intelligence’s most significant economic impact lies not in automation or prediction but in reducing ‘translation costs’—the friction between teams, systems, and knowledge flows. This thesis challenges mainstream economic analyses of AI, which often emphasize efficiency gains from task automation. Choudary’s work, detailed in his book Reshuffle, positions coordination as the cornerstone of AI’s transformative potential.

The Core Thesis: Coordination Over Automation

Choudary’s central argument is that AI’s value lies in its ability to lower the costs of translating information, aligning workflows, and synchronizing actions across fragmented systems. While automation reduces the time and cost of executing tasks, and prediction lowers uncertainty, translation costs—defined as the expenses of aligning disparate teams, data formats, and operational standards—remain the most underappreciated factor in AI’s economic impact. By enabling seamless communication between systems, AI facilitates coordination that transcends individual tasks, creating new organizational structures and value chains.

For example, Choudary contrasts the traditional view of AI as a tool for automation with its potential as a ‘coordination layer.’ He cites the example of Toyota's Production System, which redefined manufacturing by aligning suppliers and workers through shared protocols, rather than automating individual tasks. Similarly, AI can act as a ‘scalable infrastructure’ for modular, on-demand economies, where teams collaborate without predefined workflows or consensus.

The Railroad Analogy: From Canals to Governance

Choudary draws a historical analogy to illustrate the shift from automation to coordination. He compares AI’s potential to the evolution from canals to railroads: canals optimized execution within fixed processes, while railroads redefined governance to enable complex, interlocking systems. In the context of AI, this means moving from task-level automation to building governance frameworks that allow agents to act cohesively across systems.

‘Agentic AI,’ a term Choudary emphasizes, refers to systems where AI agents make decisions and interact with one another and humans. The value of these systems lies not in how quickly tasks are executed but in how effectively workflows are synchronized, governed, and adapted to changing conditions. Poorly governed agents can amplify errors, while well-governed ones can detect problems early, synchronize across functions, and adapt dynamically.

Mainstream Analyses vs. Choudary’s Coordination Framework

Mainstream economic analyses of AI, as highlighted in reports from McKinsey and Goldman Sachs, often focus on automation’s ability to replace routine tasks and boost productivity. These studies predict AI could contribute 0.5–3.4% to global GDP annually by 2030, primarily through efficiency gains. However, Choudary argues that these analyses overlook the systemic risks of automation, such as ‘speed mismatches’—where AI-driven processes outpace human oversight, leading to fragmentation.

For instance, a marketing AI might generate insights faster than sales teams can act on them, creating misalignment. Choudary contends that without coordination infrastructure, automation alone cannot unlock AI’s full potential. Instead, organizations must invest in governance frameworks that align agents with organizational objectives and enable dynamic adaptation.

Real-World Implications: Reshaping Work and Power Dynamics

The shift from automation to coordination has profound implications for organizational structures and power dynamics. Choudary notes that AI’s uneven impact across functions will shift influence within companies. Roles reliant on repeatable knowledge work, such as IT support or customer service, may lose centrality, while teams managing complex, cross-disciplinary problems gain leverage.

This is exemplified in the case of Walmart, which leveraged ERP systems not just to cut costs but to rewire relationships with suppliers, creating a supply chain that could orchestrate inventory across an entire retail ecosystem. Similarly, AI’s ability to commoditize execution (e.g., coding, data analysis) will shift value to areas where human judgment, creativity, and integration remain scarce.

The Path Forward: Governance as the New Frontier

Choudary’s thesis underscores the need for organizations to move beyond incremental efficiency gains and embrace governance as the foundation of AI adoption. This involves rethinking workflows, standards, and decision-making processes to align with agentic AI’s capabilities. Leaders must ask not just how to automate tasks but how to design systems where agents can coordinate effectively.

As the Substack article The Problem with Agentic AI in 2025 notes, the challenge is not teaching agents to act but aligning their actions with organizational goals. Governance becomes the key determinant of success, defining goals, autonomy, exception escalation, and accountability. Organizations that master this shift will capture the transformative benefits of coordination, while those that cling to automation-centric strategies risk stagnation.

Conclusion

Sangeet Paul Choudary’s argument that AI’s big payoff lies in coordination, not automation, challenges conventional economic analyses and offers a forward-looking framework for understanding AI’s impact. By focusing on translation costs and governance, his thesis highlights the need for systemic rethinking of how AI integrates into workflows, ecosystems, and organizational power structures. As the AI era unfolds, the ability to coordinate across fragmented systems may prove as critical as the technology itself.