Soumitra Dutta: Rethinking Execution Strategies For Companies in the Era of Agentic AI
The traditional strategy practice carried on for decades was a simple formula: leaders defining the direction and communicating it to peers who execute. But now, with agentic AI not just processing commands but also acting on intent by understanding the desired outcome, it completely changes the old-school equation.
Soumitra Dutta (Portulans and NEXIVERIFY co-founder), in his "Programmable Strategy: When Strategy Becomes Code," which he wrote with Yves Doz (INSEAD), the second paper in their eight-part Winning Intelligence series, argues that strategy is no longer just devised at the leadership level and then implemented. It is compiled into the systems that execute it.
Why the old model needs an AI upgrade Why the old model needs an AI upgrade
When AI systems make thousands of decisions an hour across critical business functions, vague intent is no longer needed. Soumitra Dutta IIT graduate bluntly states, “In the age of continuous AI execution, strategy that lives only in documents and presentations is not strategy at all. It is aspiration.”
Compiling Successive Layers
Soumitra Dutta feels that in the era dominated by autonomous AI, the translation must be directly embedded in the software rather than shared with humans for better execution. He and the co-author suggest automation across four successive layers that translate purpose into system behaviour: Soumitra Dutta feels that in the era dominated by autonomous AI, the translation must be directly embedded in the software rather than shared with humans for better execution. He and the co-author suggest automation across four successive layers that translate purpose into system behaviour:
Intent: What the organisation’s core priorities are - growth, margin, market position, customer value - mention it precisely enough to be operationalised.
Control: The rules and regulations that define what AI systems are permitted to do in pursuit of that intent.
Decision Logic: Designing processes, creating rules, and agent workflows that actually run when a decision is made in real time.
Infrastructure: The compute, data pipelines, APIs and monitoring that execute the logic continuously at scale.
When compilation breaks, strategy reopens for interpretation
Companies may face roadblocks when trying to interpret new strategies with AI, and Soumitra Dutta feels that these hindrances need attention. AI does not understand generic, unclear goals like ‘be your best’ or ‘lead the industry’ and hence, interpreting these intents into action remains equally aspirational.
Soumitra Dutta Nexiverify Co-Founder emphasises why leaders need to ensure that their strategy is precise enough to guide AI-driven actions. He suggests that corporate leaders must see to it that systems are optimising for the right outcomes. They need to also make sure that the right entities are responsible for the translation of strategy into decision logic. Lastly, attention must be given to how quickly the rules can be updated as priorities change.
Comments