Keeping Strategy Aligned as Execution Accelerates

I have been thinking about a growing mismatch in enterprise technology. Organizations are becoming much faster at making and executing decisions than at recognizing when those decisions start changing something larger. AI makes the mismatch easier to see, but it did not create it. Cloud, SaaS, APIs, modular infrastructure, and increasingly distributed operating models had already shortened the distance between deciding and doing. AI compresses analysis further, and agents increasingly compress execution as well.

Strategy should not require the C-suite to approve every application choice, workflow change, automation, or model selection. People closer to the work need enough authority to execute within the strategy without waiting for senior management to bless routine decisions.

McKinsey recently described strategic analysis that once took weeks as being compressed into days with AI, while BCG reports that 35% of surveyed organizations already have agentic AI in production and another 44% plan deployments. The capacity to act is expanding faster than many of the management mechanisms used to understand what those actions eventually affect.

An agent deployed to automate a process illustrates the problem. The team may be trying to remove manual work, improve response time, or increase capacity. Once operating, the agent can also alter data access, infrastructure demand, exception handling, security exposure, staffing requirements, customer commitments, or dependence on a model or provider. The automation can work exactly as intended while producing consequences outside the context in which the original decision was made.

Enough of those decisions can create strategic drift through local optimization. Teams make rational choices against the objectives, information, authority, and constraints available to them. A model performs better, an application solves the immediate problem, or an agent eliminates several manual steps. Each choice can succeed locally while their combined effects change enterprise economics, architecture, capacity, dependencies, risk, or strategic options. The enterprise begins operating differently from the strategy its executives intended without anyone explicitly choosing a new strategy.

Moving more decisions back up the hierarchy would sacrifice much of the speed and adaptability technology now makes possible. Strategy still belongs at the top: executives decide where the enterprise is going, how to allocate capital, which risks it is prepared to carry, and which capabilities and dependencies align with that direction. Execution increasingly happens elsewhere. The management problem is keeping those two connected when execution's consequences can change faster than the assumptions behind the original decision.

The technology industry has an irritating habit of rediscovering thirty-year-old ideas after adding GPUs. Decision Support Systems were created to help managers combine information, alternatives, analysis, and judgment before making consequential choices. That model assumed the difficult part was reaching the decision. Once the manager decided, the DSS had largely completed its job.

That assumption is becoming less useful. A decision can be well supported when it is made and still have consequences nobody could see clearly at the time. A modern DSS would stay involved as execution unfolds, comparing what happens with the assumptions, authority, dependencies, resources, and conditions that made the original decision reasonable.

It would need enough enterprise context to recognize material changes: effects spreading beyond the original process or function, new dependencies becoming important, local activity consuming capital or capacity required elsewhere, actual results contradicting the assumptions behind the decision, or several individually reasonable choices interacting in ways that were not visible separately. None of those conditions proves that the original choice was wrong. They indicate that the effective scope of the decision has changed and that the tradeoff may now belong somewhere else.

Governance establishes what is permitted, who has authority, and where established limits sit. A modern DSS would continue observing what happens after properly authorized activity begins interacting with the rest of the enterprise. Most decisions should never reach executive attention because nothing significant changed. The exceptions should reach the executive responsibility affected—capital, architecture, operations, workforce, security, compliance—or the C-suite when the consequences span several of them.

A consequence that is material and difficult to reverse may justify stopping execution while it is evaluated. The executive team would need to establish the authority to do that in advance; the DSS would exercise delegated authority rather than invent its own. The people closest to the work should also be able to challenge the evidence behind the intervention and provide context the system or executive function lacks. If that evidence shows that an assumption embedded in strategy is wrong, executives change the strategy. Local execution can challenge the evidence behind strategy without quietly becoming the mechanism for changing it.

The resolution should not disappear when execution resumes or stops. The enterprise needs to retain what changed, what evidence caused the intervention, which assumptions were challenged, the competing interpretations, what executives decided, and what happened afterward. A threshold that repeatedly interrupts harmless activity should probably change; a warning repeatedly overridden before the predicted consequence appears should influence the next decision differently. Over time, those decisions become institutional memory rather than another archive of approvals.

Experienced operators matter because identifying a change and understanding its significance are different capabilities. Technology can recognize when a dependency appears, consumption changes, authority expands, or exceptions accumulate. Experience helps you recognize when a technically replaceable dependency will be operationally difficult to unwind, when a small exception resembles the start of an earlier failure pattern, or when a locally successful decision conflicts with something the enterprise learned several technology cycles ago. A DSS can preserve more of that reasoning and make it available later without pretending that recorded history is equivalent to human judgment.

The original DSS improved the information available at the moment of decision. The version enterprises increasingly need would connect the decision with what happens during execution and what the organization learns afterward. Distributed execution can remain distributed, while executive judgment reenters when consequences begin affecting assumptions and choices that belong to enterprise strategy. The enterprise can recognize strategic drift through local optimization as it forms, rather than discovering it after a collection of individually reasonable decisions has already moved the company somewhere else.

Reading Resources

• McKinsey & Company — “The Judgment Bottleneck: Balancing AI and Human Speeds in Strategy,” October 15, 2026
  McKinsey examines the growing mismatch between how quickly AI can perform research, analysis, synthesis, and scenario generation and how much time executives still need to understand assumptions, challenge evidence, weigh tradeoffs, and own consequential decisions. It adds evidence to the timing problem behind this piece: accelerating analysis does not automatically accelerate judgment, and greater analytical capacity can make experienced executive attention more important rather than less.
  
• Boston Consulting Group — “The Authorization Gap: Why Yesterday’s Controls Won’t Work with Today’s Agents,” September 22, 2026
  This is the resource most directly related to the DSS argument. BCG examines how agents can remain within assigned permissions while producing consequences that traditional authorization models were not designed to anticipate. Its treatment of permission, purpose, conduct, and consequences adds useful depth to the distinction between an action remaining authorized and its effective scope changing during execution.
  
• Deloitte — “Rewiring the Enterprise Operating Model for AI Scale,” June 29, 2026
  Deloitte broadens the discussion from individual AI deployments to the enterprise operating model. Based on its 2026 Global Technology Leadership Study of 662 senior technology leaders at organizations with at least $1 billion in revenue, it examines changes in decision-making, capital allocation, risk, organization, and work as AI scales. It adds the enterprise-management perspective needed when consequences that originate within one function begin affecting several.
  
• Boston Consulting Group — “Five Ways That AI Front-Runners Change How Work Gets Done,” September 24, 2026
  This provides an adjacent perspective rather than simply supporting the argument. BCG examines organizations that deliberately move decision rights closer to the people nearest the problem and use factors such as reversibility to decide where decisions should sit. Distributed decision-making can increase speed and value; that makes it even more important to recognize when its consequences require broader judgment.
  
• Accenture — “When Everyone Has AI, Your Operating Model Will Set You Apart,” October 6, 2026
  Accenture approaches the issue through the relationship among strategy, decisions, execution, and organizational learning. It is most useful for the latter part of this argument: decisions and their outcomes can become part of the operating context used to improve subsequent choices rather than disappearing after approval or implementation.

What’s your take? We are fostering a community of executives who want to discuss these issues in depth. This newsletter is but a part of it. We welcome your feedback and look forward to engaging in these conversations.

If you are interested in exploring the full report, discussing the Board’s offering further, or have any additional questions, please contact me at [email protected], and I will be happy to connect with you.