Soumitra Dutta on How AI Changes Strategy in Businesses
A three- or five-year strategic plan used to be the guiding light of company
planning‚ reviewed at every board meeting and updated annually. But, according
to Soumitra Dutta, former dean of Oxford Said Business School and Yves Doz‚ an
Emeritus Professor of Strategic Management at INSEAD, this model is fading․
In an age of AI and AI agents, a firm requires a strategy that is updated
continually. “As AI becomes embedded in everything from data systems to
decisions, workflows and customer interfaces, strategy has now shifted towards
providing a continuous operating architecture.”
Leaders, write Dutta and Doz in a recent article on INSEAD Knowledge, "are
no longer just decision makers but need to become architects‚ guides and
governors". When AI agents are setting prices‚ routing customer complaints
and dynamically modifying inventory in real time, strategic intent has to be
built directly into the systems doing the work.
Dutta and Doz say it has to be translated "into constraints‚ decision
logic and infrastructure"․
Soumitra Dutta and Doz illustrate the point in their piece: the bank's promise on trust
should not just be a slogan‚ but be expressed through constraints on risk‚
triggers for escalation‚ customer policy‚ and human oversight requirements․ A
firm's promise to the customer for speed must be expressed in its inventory
systems‚ its routing logic‚ its fulfillment processes․ When this translation
occurs‚ Dutta and Doz argue‚ systems behave as intended․ When that translation
fails‚ they optimize locally‚ and subtly weaken the overall strategy․
Most firms haven’t done that translation, going by enterprise data. Research from MIT's NANDA project found that 95% of enterprise generative AI pilots in large corporations did not create any business value․ Another study found that 88% of agentic AI pilots failed․ Forrester and Anaconda attribute the poor results to governance failures and vague evaluation criteria‚ rather than poor models․ Viewed through Dutta and Doz’s lens, software was not the limiting factor, architecture was.
Soumitra Dutta
and Doz write in their article that the big question now includes
"how does our organisation become more intelligent‚ and how do we ensure
that intelligence helps us win?"
We would add that boards (most boards at any rate) are good at approving a five-year plan but not so good at checking whether the recommendation engine or pricing agent has drifted away from what the company intended․ This is the 'governance friction', the gap between articulated strategy and what actually gets done by AI systems.