Why every organization needs an AI portfolio strategy
Treat AI as a governed portfolio of bets with owners and exit rules — not a pile of disconnected pilots competing for the same scarce budget.
$2.6–4.4T
annual GenAI value potential
~75%
of that value in 4 functions
41
capabilities in BCG's maturity model
25%
of execs built value by focusing narrowly
- Generative AI alone carries an estimated $2.6–4.4 trillion in annual value potential — but roughly 70–75% of it concentrates in a handful of functions, not everywhere at once.
- The organizations creating real value do it by focusing on a small set of initiatives and scaling them fast, not by running the most pilots.
- AI maturity is multidimensional — BCG assesses it across 41 distinct capabilities spanning strategy, technology, people, innovation, and outcomes.
- Even AI itself struggles with portfolio-level judgment: it improves market research, but hasn't yet moved the needle on high-stakes calls like M&A or capital allocation.
- Without an explicit portfolio — horizons, prioritization criteria, and a rule for what gets killed — organizations default to funding whichever pilot has the loudest sponsor.
Value doesn't spread evenly — so bets shouldn't either
McKinsey's analysis of 63 generative AI use cases across 16 business functions estimates $2.6 trillion to $4.4 trillion in annual value — but about 75% of that falls across just four areas: customer operations, marketing and sales, software engineering, and R&D. BCG's 2025 research finds a similar concentration: roughly 70% of AI's potential value sits in core functions like R&D, innovation, and digital marketing.
That concentration is the whole argument for treating AI as a portfolio rather than a technology rollout. If value pools this unevenly, an organization funding twenty scattered pilots across twenty functions is, by construction, spreading most of its resources across the 25% of use cases that don't matter much — while under-resourcing the few that do.
Value capture by tier (BCG, 2025)
Source: BCG, The Widening AI Value Gap — Build for the Future 2025 (n=1,250 senior executives).
A portfolio has horizons — most AI programs don't
Innovation-portfolio literature has long organized bets into three horizons: near-term efficiency plays with fast, measurable returns; medium-term initiatives that strengthen competitive position; and longer-horizon, higher-risk bets on new capabilities. Applied to AI, practitioners commonly split budget along similar lines — the bulk toward proven, fast-ROI applications, a meaningful slice toward emerging generative and predictive capabilities, and a smaller reserve for exploratory bets. The exact split is a judgment call for each organization; naming the horizons and sizing them deliberately is what most AI programs skip entirely.
| Horizon | Focus | Typical share of budget | Time to value |
|---|---|---|---|
| H1 — Efficiency | Automation, forecasting, fraud/anomaly detection | 55–65% | < 6 months |
| H2 — Position | Generative workflows, competitive intelligence, product features | 20–30% | 6–18 months |
| H3 — Transformation | Agentic/autonomous systems, new business models | 10–15% | 18+ months |
Illustrative allocation synthesized from portfolio-planning practice (Three Horizons framework, adapted for AI); exact splits are organization-specific.
BCG's "future-built" companies — the 5% capturing disproportionate value — don't just pick better use cases; they run a different operating model around them. They plan to spend 26% more on IT overall and dedicate up to 64% more of that IT budget to AI, and BCG estimates they see roughly twice the revenue increase and 1.4 times the cost reduction of other firms. Agentic AI is where this shows up first: future-built firms allocate 15% of their AI budget to agents, versus roughly 12% for firms that are merely "scaling," and almost none for the 60% still lagging.
"AI tools and agents have delivered consistent, positive impact for strategy leaders only in market intelligence and research."
— BCG, The Corporate Strategy Function in an AI-First World, 2026
Even AI can't do your portfolio judgment for you
There's a useful irony in BCG's 2026 research on strategy functions: the first half of 2025 alone saw more strategy-related AI tools launched than the previous two years combined, yet those tools have delivered consistent value only for market intelligence and research — the commodity layer of strategic work. The judgment-intensive decisions that actually define a portfolio — M&A, partnerships, capital allocation across competing bets — have not seen material improvement from AI. Separately, BCG's finance-function research found that of the 45% of finance executives who can even quantify their AI ROI, a third report returns under 5%, well below the 20%+ threshold many organizations are planning around.
AI can help you research your options faster. It still can't decide, for you, which few of them deserve your organization's limited money, data, and engineering time. That decision — the portfolio itself — remains a human governance function.
The data
| Finding | Source | Year |
|---|---|---|
| Generative AI use cases carry $2.6–4.4T in annual value potential; ~75% concentrates in 4 functions | McKinsey Global Institute | 2023–25 |
| 5% of firms are "future-built," 35% scaling, 60% capturing little value; maturity spans 41 capabilities | BCG, Widening AI Value Gap | 2025 |
| Future-built firms spend 26% more on IT, up to 64% more of it on AI, for ~2x revenue / 1.4x cost gains | BCG, Build for the Future Global Study (n=1,250) | 2025 |
| AI tools help strategy teams with market intelligence, not yet with M&A or portfolio decisions | BCG, Corporate Strategy in an AI-First World | 2026 |
| Only 45% of finance execs can quantify AI ROI; a third of those report under 5% return | BCG Center for CFO Excellence | 2025 |
Most organizations don't lack AI ideas — they lack a way to say no to most of them. Every pilot sounds reasonable in isolation, which is exactly the problem: reasonable ideas, funded without a shared framework for comparing them, will always outnumber the engineering time and clean data available to support them. A portfolio strategy isn't a bigger plan; it's a smaller, harder one — naming the few bets worth real investment, sizing them against a horizon, and setting the evidence bar that gets a bet killed rather than quietly re-funded. That's a governance exercise as much as a technical one, and it's usually missing not because organizations lack ambition, but because nobody owns the "no."
Where CometX fits
Building a defensible AI portfolio is a governance and evidence problem before it's a technology one — territory where three CometX practices apply directly:
01
Consulting & advisory
Evaluation frameworks and theories of change that turn "which AI bets do we fund" into a governed decision — horizons, prioritization criteria, and the evidence bar for scaling or killing each one.
02
Applied research
Rapid evidence reviews that compare competing use-case bets on real evidence — feasibility, expected value pool, data readiness — before budget is committed.
03
Training & capacity building
Grant-writing and data-literacy training that builds the internal capacity to run a portfolio review cycle yourselves, not just for one engagement.
Sources
- McKinsey Global Institute, "The Economic Potential of Generative AI: The Next Productivity Frontier", 2023 (figures still cited as the standing estimate through 2025).
- Boston Consulting Group, "AI Leaders Outpace Laggards with Double the Revenue Growth and 40% More Cost Savings", September 2025.
- Boston Consulting Group, "The Widening AI Value Gap: Build for the Future 2025", 2025.
- Boston Consulting Group, "The Corporate Strategy Function in an AI-First World", 2026.
- Boston Consulting Group Center for CFO Excellence, "How Finance Leaders Can Get ROI from AI", 2025.
- Three Horizons portfolio framework, as adapted for AI investment by ITONICS and industry practitioners, 2025–2026.