For much of modern management history, strategy creation was seen as an episodic activity. Leaders analysed markets, hired management teams, allocated resources and translated plans into action through structures, incentives, information and decision processes. Strategy was formulated at periodic intervals and the plan implemented over time.

That model is starting to look obsolete. 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. Decisions are increasingly made or shaped by systems and AI agents. Learning happens in real time. Ecosystems form and re-form around digital interfaces, while protocols automatically guide responses to specific opportunities.

This shift calls for what we term winning intelligence: the capacity of a firm to turn intelligence into sustained advantage. It isn’t simply about having better data, more advanced models or superior AI tools, but about the ability to design systems in which intelligence flows smoothly, learning deepens and autonomy is governed, while actions remain aligned with strategic purpose.

The central question for leaders is no longer: What is our strategy? It’s now also become: How does our organisation become more intelligent, and how do we ensure that intelligence helps us win?

Strategy becomes architecture

In traditional firms, strategic intent was typically translated into action by people. Managers interpreted priorities, made trade-offs, escalated exceptions and adjusted execution. In AI-enabled firms, much of this translation work must now be built directly into systems, which are intelligent and agentic.

A bank’s strategy on trust, for example, cannot remain a slogan. It must appear in risk constraints, escalation rules, customer policies, compliance checks and human-review requirements. A retailer’s promise of speed must be an integral part of inventory systems, routing decisions, fulfilment workflows and service interfaces.

Strategic intent must be translated or programmed into constraints, decision logic and infrastructure. When this translation works, systems act in ways that reflect leaders’ intentions. When it fails, systems may optimise locally but could ultimately undermine the firm’s broader strategy.

Winning intelligence therefore depends not just on intelligence alone, but also on architecture: the design of the systems through which intelligence becomes action.

The new responsibility of leadership

If strategy becomes systemic and continuous, the role of leaders will change. They are no longer only primary decision makers but also designers and governors of the systems through which decisions are made, learning occurs and advantage is built. 

This shift creates four essential leadership tasks:

1. Define purpose

AI systems can optimise, but they can’t decide what’s worth optimising for. That remains a human responsibility. Although a recommendation system can learn what increases engagement, it can’t determine whether that engagement is healthy, trustworthy or strategically desirable. A pricing system can maximise revenue, but cannot decide whether the resulting customer experience supports the brand and its values. 

The first responsibility of leadership is therefore to define purpose clearly enough so that it can guide how systems are designed: what the organisation values, what trade-offs it will accept and what outcomes it refuses to pursue.

2. Design the intelligence architecture

The second task is to design the architecture through which intelligence flows. AI advantage doesn’t come from individual tools. It comes from the integration of data, models, decisions, workflows and interfaces, a combination that we term as the firm’s intelligence stack. A company may have strong data but weak decision logic. It may deploy powerful models but fail to redesign workflows. It may control customer interfaces but be unable to convert interactions into learning. Leaders must understand where intelligence is created and where it can potentially get stuck.

The same architectural logic applies to learning. AI-enabled firms don’t act first and then learn later – they increasingly learn while acting. Every customer interaction, transaction and machine-to-machine exchange can become part of a broader learning surface. The strongest firms will build systems that are able to learn faster, at high density and at scale. In practice, this means designing organisations where feedback travels quickly from action to insight, and from insight back to changed behaviour.

3. Calibrate autonomy

As AI becomes more agentic, systems will be able to sequence tasks, negotiate trade-offs, reroute resources, trigger exceptions and coordinate with other systems. The firm begins to resemble a complex responsive system consisting of human and artificial agents.

This requires a new approach to control. Leaders cannot supervise every action, but neither can they delegate blindly. They must define a strategic envelope: the boundaries within which agents are able to act, adapt and learn. The envelope contains three elements: purpose, constraints and feedback. Purpose defines what the system is trying to achieve. Constraints detail what it must not do. Feedback allows leaders to detect drift, unintended consequences and emerging patterns.

The goal isn’t to suppress autonomy, but to make autonomy safe enough to scale. Leaders must decide where agents can act freely, where they need approval, where they must escalate issues and where human judgment is indispensable.

4. Interpret emergence

In AI-enabled organisations, strategy goes beyond a leader’s intentions to include what the system produces. A firm may intend to compete on trust, but its systems gradually optimise for speed or engagement. It may intend to build resilience, while supply chain agents repeatedly favour cost minimisation. It may intend to improve customer experience, while service bots optimise closure rates rather than relationship quality.

Leaders must therefore constantly compare intended strategy with produced strategy. What patterns are emerging from the interaction of people, models, agents, workflows and partners? Are these patterns aligned with the firm’s original purpose, or is local optimisation creating system-level drift?

This becomes even more important as competition moves beyond the firm. Increasingly, value is created through temporary alignments of firms, platforms, data providers, agents and protocols – what we call “swarms”. In such environments, firms compete not only through assets or alliances, but through their ability to be selected, trusted and effective in real-time configurations.

This also changes incentives. Static compensation systems and contracts are often too slow for dynamic, AI-enabled ecosystems. Firms need adaptive ways to align behaviour across human stakeholders and AI agents. Incentives become programmable when contributions can be measured, translated into rewards, recognition, access or rights, and then adjusted as behaviour changes.

The strategic challenge ahead

AI places greater demands on leaders. They must still exemplify values, make choices, set strategic direction and exercise judgment. But they must also design the systems through which intelligence can flow, decisions are made, learning accumulates, autonomy is bounded and behaviour is aligned.

Winning intelligence is the capability to do all this in a coherent way that gives the organisation an advantage. The firms that win won’t be those that simply adopt the most advanced AI tools, but those that build the architecture through which such intelligence becomes a strategic asset.