AI Strategy: The Execution Gap
Toward an AI Strategy Framework That Connects Ambition to Delivery
The AI Strategy Playbook Landscape
This article’s life began with a curated list by Rubén Domínguez Ibar, a strategy analyst at BBVA. His piece, 16 AI Strategy Playbooks You Shouldn’t Miss, brings together a wide range of resources from consultancies, vendors, standards bodies, and strategic & delivery framework providers.
The term “playbook” is used loosely. Not all of the resources are structured strategy frameworks. Some are maturity models, governance guidelines, or delivery artefacts. But taken together, they offer a useful scan of how different actors are shaping enterprise AI narratives. Each reflects its own set of incentives, from transformation branding to vendor ecosystem lock-in.
These resources come from five broad types of organisations:
Consultancies, including McKinsey, Accenture, BCG, Bain, Deloitte, and PwC
Vendors, such as Microsoft, Google, IBM, and Amazon
Standards, such as NIST, which provides governance and risk frameworks for responsible AI
Strategic and delivery framework providers, such as aiSTROM, the Enterprise AI Canvas, and Scaled Agile
Multi-stakeholder platforms, such as the World Economic Forum
These players and their strategic perspectives are mapped in The Enterprise AI Landscape, which shows how different roles influence how AI strategy is defined and communicated.
From this broader set, five playbooks stood out for how clearly they present themselves as comprehensive, enterprise-wide AI strategies: McKinsey, Accenture, BCG, Microsoft, and IBM. These form the core of this synthesis.
While many of the reviewed playbooks predate the generative AI surge, or treat it as a subset, recent updates from Microsoft, McKinsey, and Accenture increasingly foreground GenAI use cases, risks, and tooling. This synthesis, however, treats GenAI as part of the broader AI strategy challenge. The focus is on what enterprises must build regardless of model type or hype cycle.
From this ecosystem, a proposed Enterprise AI Strategy Framework takes shape. It surfaces not just shared priorities across the leading playbooks, but also where they consistently fall short: in delivery systems, organisational rewiring, and feedback loops.
What the Five Playbooks Say AI Strategy Is
From the broader review of 16 AI strategy playbooks, five stand out for how directly they frame the challenge of enterprise-wide AI adoption. Rather than just offering recommendations, each presents a complete view of what AI strategy is, what it should enable, and who it is for. They position AI strategy as a coordinated, organisation-wide effort that connects leadership, data, talent, and governance to business transformation.
These five, from McKinsey, Accenture, BCG, Microsoft, and IBM, target different executive audiences, but share consistent patterns in how they conceptualise the strategic work of AI adoption:
McKinsey – The Executive’s AI Playbook (2023) A strategic roadmap for aligning AI with business value, driven by transformation mandates and capability building.
Accenture – The Art of AI Maturity (2022) A staged model of AI maturity across leadership, technology, and organisational enablers.
BCG – Transforming with AI (2023) A transformation-first lens on AI, emphasising reinvention, leadership alignment, and operating model change.
Microsoft – The CIO’s Generative AI Playbook (2023) Technical and strategic guidance for CIOs, with a focus on infrastructure, risk, and enablement at scale.
IBM – The CEO’s Guide to Generative AI (2023) A high-level leadership framing of AI strategy, centred on investment logic, value potential, and board-level imperatives.
Across all five, six components surface repeatedly:
Leadership and Sponsorship Strategy begins at the top. Whether through CEO narratives (IBM) or transformation mandates (McKinsey), executive ownership is positioned as the engine behind AI maturity.
Strategic Alignment AI is only strategic if it is mapped to business value. This shows up as maturity models (Accenture), use case frameworks (McKinsey), or vision-aligned reinvention paths (BCG).
Data and Infrastructure Foundations While not always foregrounded, each playbook assumes scalable, secure, well-governed infrastructure as a baseline enabler, with Microsoft providing the most detail here.
Talent and Operating Model The role of cross-functional teams, reskilling, and change leadership is a common thread. Accenture and BCG in particular frame this as a core lever for sustainable transformation.
Responsible AI and Governance Every playbook names responsible AI, though with varying levels of depth. Some stay at the level of leadership principle, others (notably Microsoft) provide operational frameworks.
Progression or Maturity Pathways Most playbooks imply, if not explicitly map, a path from experimentation to embedded capability. Accenture leads here with a formal maturity model. Others suggest evolution through capability building or architectural scaling.
From these themes, a working definition emerges:
AI strategy is the deliberate alignment of leadership, data, talent, and governance to embed AI into the operating model and value creation logic of the organisation.
It is a clear, compelling vision. But as the next section shows, it is also incomplete.
What the Five Playbooks Miss
Many AI strategies present a strong vision. But across the five core AI strategy playbooks, the same blind spots appear. Not as isolated oversights but as structural gaps in how execution is addressed:
Use case identification is treated as a one-off exercise
McKinsey and Accenture mention identifying use cases, but usually as an early-stage activity. None of the five address how use cases evolve, get refined, or connect to feedback loops in practice.Strategy is focused at the top, not through the organisation
Executional friction from middle managers, functional leads, or frontline teams is barely acknowledged.End-states are described, but paths to get there are not
Scaling is a common goal, but there is little discussion of retraining loops, prompt drift, or feedback cycles.Workflow integration is mostly overlooked
The five rarely show how AI fits into actual processes: document review, service journeys, audits, reporting.Local adoption dynamics are missing
The playbooks assume centralised control, with little attention to how innovation often emerges from departmental instigators or peripheral teams.Strategy is assumed to be persuasive on its own
There is limited recognition that strategy must earn belief and be translated into artefacts, rituals, and metrics that drive behaviour.
These gaps reflect a common shortfall in the third element of good strategy: coherent action. While all five offer a compelling vision and a guiding policy, they provide limited support for turning that policy into sustained, real-world execution.
Applied AI Strategy That Enables Execution
While only five AI strategy playbooks were selected for the core synthesis, based on how explicitly they frame themselves as organisation-wide AI strategies, four additional artefacts from the broader set offer something the core five do not. These were initially excluded because they are not end-to-end strategy playbooks. But each contributes essential guidance where the dominant narratives fall short: in operationalising strategy, managing risk, integrating workflows, and enabling delivery.
What they lack in breadth, they make up for in executional depth. Together, they help fill critical gaps in the mainstream strategy narrative.
The four are:
PMI Generative AI Framework (2023) Provides delivery artefacts such as canvases, role impact maps, and prioritisation tools for integrating GenAI into projects.
NIST AI Risk Management Framework (2023) Offers a structured model for addressing trust, traceability, and governance across the AI lifecycle.
Enterprise AI Canvas (2024) A strategic mapping tool linking value propositions, infrastructure, and operating models to practical planning.
aiSTROM Framework (2024) A lifecycle model for building and iterating LLM applications, focused on scoping, feedback, and continuous delivery.
These artefacts are not substitutes for strategic vision. But they strengthen its ability to be executed. The Enterprise AI Strategy Framework that emerges from this synthesis is built not just from high-level intent, but from the delivery realism these tools enable.
A Strategic Scaffold for Enterprise AI
This section outlines a strategic scaffold: a working set of dimensions intended to support more coherent and executable AI strategy. It builds on what the five core AI strategy playbooks get right and integrates missing delivery mechanisms from the four applied AI strategy playbooks.
These 11 dimensions do not represent abstract themes. Each one is a concrete area of capability that an organisation must define, support, and connect to deliver AI in practice. Together, they form a structured base that organisations can build on to connect ambition with execution.
The scaffold includes 11 strategic dimensions, grouped into three focus areas:
Strategic Direction
Leadership Commitment
Senior ownership, active role modelling, resourcing.Strategic Alignment
Clear linkage between AI initiatives and business priorities.AI Governance
Decision rights, risk policies, compliance mechanisms.Ethics and Trust
Frameworks for transparency, fairness, and public confidence.
Capability Foundations
Data and Infrastructure
Availability, accessibility, quality, and observability.Talent and Teams
Skills, roles, cross-functional collaboration.Use Case Management
Prioritisation, business case, value realisation.Workflow Integration
Embedding AI into tools, processes, and decision loops.
Execution and Adaptation
Delivery Ownership
Clear accountabilities for delivery, iteration, and scaling.Feedback Loops
Metrics, experimentation, organisational learning.Stewardship and Scaling
Connecting local wins to enterprise value.
How to Use This as an AI Strategy Playbook
This scaffold is the backbone of a more complete AI strategy playbook. While most playbooks focus on high-level ambition, this one bridges strategy with delivery. You can use it to:
Frame a Complete AI Strategy
Use the 11 dimensions as the basis for defining your strategic agenda. Each focus area captures what must be intentionally built or connected for AI to deliver enterprise value.Audit Existing Plans
Map your current initiatives against the scaffold. Identify where you are strong and where assumptions are being made without ownership or capability.Structure Strategic Dialogue
Use the scaffold to guide leadership conversations, board papers, or investment proposals. It helps bring coherence to fragmented efforts.Support Playbook Creation
If you are writing an internal AI strategy, this scaffold can be your starting point. Each dimension prompts a critical line of inquiry, from governance and talent to feedback and integration.
When Does a Scaffold Become a Framework?
This scaffold becomes a full framework when it is applied with intent and made actionable. That requires:
Definition
Each dimension must be clarified in your specific context. What does "workflow integration" mean in legal, operations, or customer support?Ownership
Assign clear accountability for progress along each dimension. Who is leading on responsible AI? Who is building reuse into your AI services?Connection
Link the dimensions across functions. Data infrastructure should support use cases. Talent strategy should align with delivery ownership.
Once the scaffold is used to shape real plans, decisions, and delivery structures, not just as a conceptual map, it becomes a framework. One that is alive, operational, and tailored to your organisation.
What Comes Next
This article provides a strategic foundation by synthesising the key dimensions that underpin effective enterprise AI strategy. While it outlines the what and why of these dimensions, the how (practical tools, diagnostics, and implementation guidance) will be covered in upcoming work.
Future articles and resources will:
Break down each dimension into specific capabilities, processes, and roles that organisations must build.
Offer self-assessment tools and checklists to help teams identify strengths and gaps.
Share real-world examples and case studies demonstrating successful AI strategy execution.
Provide templates and frameworks designed to make AI strategy actionable and measurable.
Use this article as your strategic map, a way to understand the landscape and spot critical focus areas. Then look to the follow-up content for detailed guidance on how to build, govern, and scale AI in your organisation.
Appendix: The 16 AI Strategy Playbooks Reviewed
Below is the full list of AI strategy playbooks reviewed, grouped by organisation type, with brief summaries to provide context for the synthesis above.
McKinsey – The Executive’s AI Playbook (2023)
Frames AI strategy as a business transformation roadmap, led by senior executives and focused on aligning use cases with enterprise value. Strong on capability building and maturity progression.
Accenture – The Art of AI Maturity (2022)
Presents a staged maturity model from experimentation to AI-powered reinvention. Highlights leadership, tech architecture, responsible AI, and organisational change as key enablers.
BCG – Transforming with AI (2023)
Takes a transformation-first view of AI, positioning it as a lever for enterprise reinvention. Emphasises cultural shift, leadership alignment, and cross-functional operating model design.
Bain – Scaling AI Like a Tech Native (2023)
Advocates for adopting tech-native behaviours to scale AI, including platform thinking, modular architecture, and empowered teams. Pragmatic and delivery-oriented.
Deloitte – State of AI in the Enterprise (2023)
More of a diagnostic benchmark than a strategy playbook. Offers insights from global enterprise surveys, with emphasis on adoption challenges, risk, and talent.
PwC – AI and the C-Suite (2023)
A leadership-focused piece on how executives can drive AI value. Highlights vision, investment logic, and risk framing. Less concrete on operating models or delivery.
Microsoft – The CIO’s Generative AI Playbook (2023)
Strategy as infrastructure-led enablement. Strong on governance, data architecture, and technical guardrails. Geared toward CIOs modernising enterprise platforms.
Google – AI Adoption Framework (2023)
A structured guide to scaling AI responsibly within organisations. Covers strategy, data maturity, deployment stages, and cultural enablement. Widely adopted in tech circles.
IBM – The CEO’s Guide to Generative AI (2023)
Frames AI as a CEO-level concern. Focused on investment decisions, vision alignment, and readiness assessments. More conceptual than operational.
Amazon – The AWS Generative AI Strategy Guide (2023)
Designed for AWS clients. Focuses on cloud-native approaches to GenAI, with reference architectures and customer stories. Biased toward adoption within the AWS ecosystem.
NIST – AI Risk Management Framework (2023)
A formal framework for identifying and mitigating AI risks. Emphasises trust, accountability, and measurable safeguards. Useful for compliance and operational governance.
World Economic Forum – Blueprint for Generative AI Governance (2023)
A high-level governance framework advocating for responsible GenAI use. Covers public trust, stakeholder roles, and global cooperation. Designed for policy and strategy audiences.
Strategic & Delivery Frameworks
aiSTROM – aiSTROM Framework (2024)
A delivery lifecycle model for building LLM-powered applications. Emphasises scoping, feedback loops, and scaling infrastructure. Strong on workflow integration.
Enterprise AI Canvas – Enterprise AI Canvas (2024)
A visual tool to structure AI strategy across value, architecture, and operating models. Helps bridge vision and execution through shared planning artefacts.
Project Management Institute (PMI) – Generative AI Framework (2023)
A practical guide for project and product leaders. Offers tools like AI canvases, impact assessments, and use case scoring to bridge intent and implementation.
Scaled Agile – SAFe for AI (2023)
Integrates AI thinking into the Scaled Agile Framework. Focuses on roles, release trains, and backlog integration. Tailored for SAFe practitioners looking to extend into AI.