Part II — Strategy and Leadership

AI Strategy That Actually Works

Most enterprise AI strategies are not strategies. They are lists of use cases, organized by business unit, with optimistic timelines and no delivery model. They answer "what could AI do for us?" but not "how will we deliver AI value given our specific organizational capabilities, data infrastructure, and risk tolerance?" A strategy that cannot answer the second question will not survive contact with implementation.

13.1 What AI Strategy Must Answer

A working AI strategy answers five questions:

Where will AI create value? Not "what are all the things AI could do" but "which specific problems, solved with AI, will create the most business value given our current organizational capabilities?" This is the portfolio selection problem, and it requires applying Signal Capture discipline — not just enthusiasm.

What capabilities do we need to build or acquire? AI strategy is not just about buying tools. It is about building organizational capabilities — data engineering, ML engineering, product management for AI products, change management for AI adoption. Which of these does the organization have? Which must be hired? Which can be accessed through partners?

What is the delivery model? How will AI be built and deployed — entirely in-house, through system integrators, through vendor platforms, or some combination? The delivery model determines timeline, cost, and control.

How will we measure success? Which metrics will determine whether the AI strategy is working? Business outcomes (not model metrics) that are measurable, attributable, and meaningful to the people funding the strategy.

What are the risks, and how will we manage them? Regulatory risk, data risk, model risk, operational risk, reputational risk. An AI strategy that does not address risk is not complete.

13.2 The Portfolio Model

AI investment should be managed as a portfolio — a mix of quick wins, medium-term capability builders, and long-term strategic bets — rather than as individual project decisions.

Quick wins (0–6 months): Small, well-bounded initiatives with accessible data, low integration complexity, and measurable outcomes. Purpose: build organizational confidence, demonstrate value, and learn about AI development in the organization's specific context. Examples: document classification, email summarization, internal search improvement.

Capability builders (6–18 months): Larger initiatives that build reusable infrastructure — data pipelines, model serving platforms, evaluation frameworks, MLOps practices — alongside delivering specific AI capabilities. Purpose: create the foundation that makes future AI development faster and cheaper.

Strategic bets (18+ months): Transformative AI applications that require new capabilities, new data sources, or significant organizational change. Purpose: create differentiated capability that is hard for competitors to replicate quickly. Examples: proprietary AI models trained on unique data assets, AI-enabled business model changes.

A portfolio without quick wins has no demonstrated value to show stakeholders. A portfolio without capability builders will find each project as hard as the first. A portfolio without strategic bets has no competitive differentiation.

13.3 The One-Page AI Strategy

A strategy that cannot be communicated on one page is too complex to execute. The one-page AI strategy has five sections:

Strategic intent (2–3 sentences): What will AI enable the organization to do that it cannot do today, and why does that matter?

Priority use cases (3–5 bullets): The specific, measurable problems that the AI portfolio will address in the next 18 months, in priority order.

Capability requirements (3–5 bullets): The people, platforms, and data infrastructure investments required to deliver the priority use cases.

Success metrics (3–5 bullets): The business outcomes by which the strategy will be judged, with baseline values and targets.

Top risks (3 bullets): The biggest risks to strategy execution and the mitigation approach for each.

This document is not a project plan. It is a shared reference that aligns executives on direction, gives product and engineering teams a clear mandate, and gives governance bodies the context to make investment decisions.

13.4 What Good AI Strategy Looks Like

Good AI strategy is honest about what the organization can deliver given its current capabilities, disciplined about prioritizing the problems with the clearest signal, and realistic about the investment required to build capabilities that do not yet exist.

It does not promise to deploy AI across all business units in eighteen months. It does not claim that AI will "transform" the organization without specifying how transformation will be measured. It does not treat vendor capabilities as organizational capabilities.

The organizations that build AI strategies that work combine ambition about what AI can do with honesty about what the organization is currently capable of delivering. The gap between the two is not a reason to lower ambition — it is a roadmap for capability investment.