Core Framework
Why AI Strategy Is Different¶
Most enterprise AI programs do not fail in the model. They fail in the mandate. The technology works; the intent behind it is unclear, unfunded, or unowned. That is a leadership problem, and it is the most expensive kind.
>80% of AI projects fail — roughly twice the failure rate of non-AI IT projects, with the leading root cause being misunderstood or miscommunicated intent and purpose, not technical limitations (RAND, 2024) 95% of organizations are getting zero return on generative-AI investment; only 5% of integrated AI pilots are extracting real value, against an estimated $30–40 billion of enterprise GenAI investment (MIT Project NANDA, "The GenAI Divide: State of AI in Business 2025," 2025)
Read those two figures together and the picture is unambiguous. The binding constraint on AI value is not compute, talent, or tooling. It is strategy and leadership — knowing what to build, why it matters to the business, and who is accountable for the outcome.
This is why AI Strategy & Leadership sits at the top of the stack. It is the upstream track. Everything in the eight-track model — governance, data, talent, infrastructure, measurement — gets scoped against what this track decides. Its job is not to produce inspiration. Its job is to produce a funded set of priorities that every other track is scoped against. Without that artifact, the other seven tracks optimize in the dark.
The stakes are set by the sheer magnitude of spend now in motion.
$644 billion in worldwide generative-AI spending is forecast for 2025, up 76.4% from 2024; total worldwide AI spending is forecast at $1.5 trillion for 2025 (Gartner, 2025)
At that scale, the absence of strategy is not a neutral state. It is capital deployed against a hope. The purpose of this page is to define the artifact, the decisions, and the governance that convert that spend into returns — and to name the failure modes that quietly burn it.
For the full scope of this track, see the track overview.
What a Good AI Strategy Actually Contains¶
A good AI strategy is an artifact, not an aspiration. It is not a vision deck and not a list of trends. It is a document leadership can fund, sequence, and be held to. Most organizations do not have one.
Only 32% of enterprises actively using AI have a holistic, organization-wide AI strategy; another 27% have one only for limited or specific use cases, and 32% are still developing one — leaving the rest with no enterprise strategy at all (IBM Global AI Adoption Index, 2023)
A strategy worth the name contains four parts.
A value thesis tied to the business model. Before any initiative, leadership must state where AI changes the economics of this company — its cost structure, its products, its customer relationships, its moat. A generic "we will adopt AI" thesis is indistinguishable from a competitor's and defends nothing. The thesis is the spine; every initiative hangs off it.
A prioritized, funded portfolio of initiatives. Priorities without budget are wishes. The portfolio names specific initiatives, ranks them, and attaches real money and owners to the top of the list. The discipline here is concentration, not coverage. The evidence is blunt on this point.
8% of companies ("front-runners") are scaling AI at an enterprise level and embedding it into core business strategy; companies scaling just one strategic AI bet are nearly 3x more likely to exceed their ROI expectations (Accenture, "The Front-Runners' Guide to Scaling AI," 2025)
The lesson: fund a few bets to depth rather than many to nowhere.
Explicit success metrics. Each funded initiative carries a defined outcome — a P&L line it moves, a cycle time it cuts, a revenue stream it opens — and a date by which the result is expected. Metrics defined upfront are what later let leadership tell a winning bet from a comfortable habit.
A funding and operating model. The strategy states how AI work is paid for (central fund, business-unit budgets, or a blend), who decides reallocation, and how initiatives move from pilot to production. Absent this, even good ideas stall at the handoff between experiment and operations.
The practitioner guide to running an AI strategy process is the companion that walks through producing this artifact step by step.
The Decisions Only Leadership Can Make¶
Some decisions can be delegated to a platform team or IT. A specific, irreducible set cannot. These are decisions about the firm's appetite, posture, and boundaries — and when they get pushed down, the organization ends up with technical answers to questions that were never technical.
Risk appetite. How much model error, data exposure, or reputational risk the firm will accept in exchange for speed is a leadership judgment, not an engineering setting. And most firms are far better at seeing risk than acting on it.
Enterprises recognize AI risks more than they mitigate them — cybersecurity (66% relevant), regulatory compliance (63%), and privacy (60%) top the list, but mitigation lags in every category (e.g., IP infringement 57% relevant vs 38% mitigated); only 14% of organizations report dedicated AI-governance roles (Stanford HAI, AI Index Report 2025)
That recognition-versus-mitigation gap is a leadership gap. Closing it is a decision about appetite and resourcing that no platform team can make on its own.
Build, buy, or partner posture. Whether the firm builds proprietary capability, buys off the shelf, or partners shapes cost, speed, and defensibility for years. It is a strategic stance, taken initiative by initiative against the value thesis.
| Posture | Use when | Trades away |
|---|---|---|
| Build | AI is core to the moat; differentiation matters more than speed | Time, capital, scarce talent |
| Buy | The capability is commoditized; speed beats uniqueness | Differentiation, control |
| Partner | Capability is strategic but the firm lacks the means to build it alone | Some control and margin |
Resource allocation. Concentrating capital and talent on the few bets that matter — and starving the rest — is the act that separates a portfolio from a wish list. Only leadership can say no to the well-liked project that does not earn its place.
Where not to use AI. This is the most underused decision right and the most valuable. Naming the processes, decisions, and customer moments that stay human — for reasons of risk, trust, or simple economics — prevents scattershot deployment and protects the brand. Many firms are not even positioned to make these calls yet.
31% of organizations say they are not ready to deploy AI (Deloitte, "Governance of AI: A critical imperative for today's boards," 2025)
Readiness to deploy is itself a leadership decision. Drawing the boundary of where AI does not go is half of drawing where it does.
Connecting AI Investment to Business Outcomes¶
A strategy earns its keep when its spend can be traced to the P&L. The honest starting point is that, for most enterprises, it cannot yet.
74% of companies have yet to show tangible value from AI; only 26% have moved beyond proofs of concept to generate tangible value (BCG, "Where's the Value in AI?", 2024) 39% of organizations attribute any EBIT impact to AI, and for most that impact is under 5%; only ~6% are high performers attributing more than 5% of EBIT to AI, and nearly two-thirds have not yet begun scaling AI across the enterprise (McKinsey, "The State of AI," 2025)
The two figures measure different things — BCG's "tangible value" is a higher bar than McKinsey's "any EBIT impact," which is why the percentages differ — but they point the same way: a large majority of enterprises cannot yet trace AI spend to results.
The fix is to manage AI as a portfolio across horizons, not as a pile of disconnected pilots.
| Horizon | Bet type | What it moves |
|---|---|---|
| Efficiency | Cost and productivity | Near-term margin; fast, measurable payback |
| Growth | Revenue and new offerings | Top line; medium-term |
| Transformation | Business-model change | The moat; long-term, lumpy returns |
Each horizon needs its own success metric and its own patience. Holding a transformation bet to an efficiency bet's payback clock guarantees it gets killed before it matures — and the payback clock for AI is genuinely long.
15% of organizations using generative AI report already achieving significant, measurable ROI; only 10% see significant measurable ROI from agentic AI; 85% increased AI investment in the past year and 91% plan to increase it again — even though AI payback periods run longer than the 7–12 months typical of technology investments (Deloitte, "State of Generative AI in the Enterprise: AI ROI paradox," 2025)
Two implications follow. First, value attribution must tie each initiative back to a specific P&L line, or the portfolio cannot be steered. Second, returns concentrate. As the front-runner evidence shows, companies that drive a single strategic bet to scale are nearly 3x more likely to beat their ROI expectations — value pools in a few deep bets, not across a wide, shallow field.
This is where strategy depends on measurement. The attribution discipline, the value-realization tracking, and the signal that tells leadership which bets to double down on and which to cut all live in Track 08, Measurement & Value Realization. Strategy sets the intent; measurement closes the loop back to it. A strategy with no feedback loop is a strategy that cannot learn.
Governance Above the Program¶
Above the eight tracks sits a leadership layer. Its job is to hold intent steady, arbitrate trade-offs, and redirect capital as evidence arrives. This is not the same as risk governance — and conflating the two is a common mistake.
The handoff is clean. Track 01 owns intent: what the firm is trying to achieve with AI and where it places its bets. Track 02, AI Governance & Risk owns the guardrails: the policies, controls, and review gates that keep deployment safe and compliant. Strategy decides what and why; governance decides within what limits. One sets direction, the other sets boundaries, and the steering structure connects them.
The weak point in most organizations is ownership at the top. AI oversight too often lives well below the level where strategic trade-offs actually get made.
28% of organizations using AI report that their CEO is responsible for overseeing AI governance; 17% say their board of directors oversees it (McKinsey, "The State of AI," 2025)
The trend, at least, points the right way.
31% of board directors say AI is not on the board agenda, down from 45% the prior year (Deloitte, "Governance of AI: A critical imperative for today's boards," 2025)
Boards are arriving, but slowly. An effective leadership layer needs three things: a steering structure with named C-suite ownership and clear decision rights; a strategy-to-governance handoff that keeps intent and guardrails distinct but coordinated; and a re-direction cadence — a standing rhythm at which leadership reviews evidence from measurement and reallocates against the portfolio. Without that cadence, a strategy is set once and then drifts, immune to what the business is learning.
Common Failure Modes¶
Four failure modes account for most stalled AI strategies. Each is a leadership failure wearing a technical costume.
1. Pilot purgatory. Pilots multiply, impress, and then quietly die before production. This is the single most common pattern, and the numbers are stark.
At least 30% of generative-AI projects will be abandoned after proof of concept by the end of 2025 — due to poor data quality, inadequate risk controls, escalating costs, or unclear business value (Gartner, 2024) The share of companies abandoning most of their AI initiatives rose to 42%, up from 17% a year earlier; the average organization scrapped 46% of AI proofs-of-concept before they reached production (S&P Global Market Intelligence, 2025)
The root cause is rarely the pilot. It is the missing funding and operating model that should have carried the winners into production.
2. Strategy-as-document. A polished deck is produced, presented, applauded — and never funded. It names ambitions but attaches no money, no owners, and no decision rights. It is indistinguishable from having no strategy at all, and it is worse, because it creates the comfortable illusion of one.
3. The loudest-function trap. A pattern this track names: AI ownership falls to whichever function shouts the most — the team with the most enthusiasm or political capital, rather than the function where value concentrates. The result is scattershot: effort flows to the loudest, not the highest, and the value thesis is never the organizing principle it should be.
4. The binding-constraint trap. Leadership funds the most visible layer — usually models and tooling, the layer that demos well — instead of the layer actually limiting outcomes, which is often data, talent, or governance. As the eight-track model makes explicit, value is gated by the binding constraint, and spending on a non-binding layer produces motion without progress. Leadership's job is to find the real constraint and fund that, even when it is unglamorous.
Strategic Readiness Checklist¶
- [ ] A written value thesis ties AI to this firm's specific business model, not generic industry trends
- [ ] A prioritized portfolio of initiatives exists, with the top priorities funded and owned
- [ ] Capital is concentrated on a few strategic bets rather than spread across many shallow pilots
- [ ] Each funded initiative has an explicit success metric and a target date tied to a P&L line
- [ ] A funding and operating model defines how initiatives move from pilot to production
- [ ] Risk appetite is set by leadership and matched by actual mitigation, not just risk recognition
- [ ] Build/buy/partner posture is decided per initiative against the value thesis
- [ ] Boundaries are defined for where AI will not be used, and why
- [ ] A named C-suite owner and a board or steering body oversee AI, with intent (Track 01) and guardrails (Track 02) clearly separated
- [ ] A standing re-direction cadence reviews measurement evidence and reallocates the portfolio
Score this readiness with the companion strategic readiness scoring diagnostic.
Sources¶
- RAND Corporation — The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed — 2024
- MIT Project NANDA — The GenAI Divide: State of AI in Business 2025 — 2025
- Gartner — Forecasts Worldwide GenAI Spending to Reach $644 Billion in 2025 / Worldwide AI Spending to Total $1.5 Trillion in 2025 (press releases) — 2025
- Gartner — Predicts 30% of Generative AI Projects Will Be Abandoned After Proof of Concept by End of 2025 (press release) — 2024
- S&P Global Market Intelligence (451 Research) — 2025 AI survey, reported by CIO Dive, "AI project failure rates are on the rise" — 2025
- BCG — Where's the Value in AI? — 2024
- McKinsey & Company — The State of AI — 2025
- Deloitte — State of Generative AI in the Enterprise: AI ROI paradox — 2025
- Deloitte — Governance of AI: A critical imperative for today's boards — 2025
- IBM — Global AI Adoption Index — 2023
- Accenture — The Front-Runners' Guide to Scaling AI — 2025
- Stanford HAI — AI Index Report 2025 — 2025