How much should we spend on AI as percent of revenue?

Median across enterprises in 2026: 1 to 3 percent of revenue. Top 25 percent: 3 to 5 percent. The ceiling is rising; expect 2 to 7 percent by 2028 for AI-active companies.

AI Strategy for CTOs and CEOs: A Working Decision Frame

TL;DR

AI strategy is not a five-year plan. It’s a portfolio of bets calibrated to the rate of capability change.

The frame: pick 2–3 wedges where AI changes your unit economics or product position; ship six-month proofs against each; reinvest into the wedges that work; kill the ones that don’t. Decisions reviewed quarterly. The five-year plan most CEOs are asked to produce is the wrong artifact for this moment.

AI strategy is a portfolio of bets, not a roadmap. The frame for decisions when capability is changing every quarter.

The pattern at most companies in 2026: a board asks for an “AI strategy”; the response is a 30-page document with a five-year roadmap, a target architecture, and a center-of-excellence plan; the document ages in 90 days; the next board meeting asks for a refresh. The strategy document and the strategy decision frame are different artifacts. This piece is the decision frame for executives who need to actually decide what to do.

Why traditional strategy frames don’t quite fit

Three reasons.

1. Capability changes faster than strategy cycles. Traditional strategy assumes a stable capability frontier you optimize against. AI’s capability frontier moves every quarter — what was hard in January is trivial by June. Strategy that assumes 2-year stability becomes irrelevant by month 6.

2. The unit economics shift mid-flight. A use case that’s $5/transaction in 2025 might be $0.50 in 2026. Strategy decisions need to be re-evaluated as costs change. A traditional 3-year plan can’t accommodate this.

3. The competitive question is bet selection, not execution. In stable categories, execution is the differentiator. In AI, the major mistakes are betting on the wrong wedges. Bet selection requires different decision frames than execution.

The consequence: strategy work that produces value in 2026 looks more like portfolio management than roadmap development.

The portfolio frame

Three categories of AI bets.

Category 1: Operational efficiency (table stakes)

What it is: AI to do existing work cheaper or faster. Internal productivity, ops automation, support deflection.

How to think about it: this is table stakes. Everyone is investing here. The question isn’t whether to invest, it’s how fast to compound. Don’t expect this to be a competitive moat; expect it to be the new baseline.

Budget allocation: 40–50% of AI investment for most companies.

Category 2: Product capability (differentiation)

What it is: AI in your product that customers can see and value. New features, new capabilities, new product lines.

How to think about it: differentiation comes from product, not from operational efficiency. The companies winning AI strategically have AI capabilities visible in their products that competitors can’t quickly match.

Budget allocation: 30–40% of AI investment.

Category 3: Business model bets (asymmetric)

What it is: bets that change your unit economics or business model — new revenue streams, autonomous services, AI-native product lines.

How to think about it: this is where the asymmetric returns are. Most don’t work; the ones that do change the company.

Budget allocation: 10–20% of AI investment, typically in 1–3 specific bets.

The five questions for CEOs and CTOs

The questions a working AI strategy answers.

1. Where does AI change our unit economics? Specific. Not “AI will help everywhere” — which 2–3 cost or revenue lines does AI specifically change?

2. Where does AI change our product? What capability becomes possible that wasn’t, or feasible that wasn’t economic? What new product lines does this enable?

3. Where does AI change our competitive position? Are competitors getting ahead in specific dimensions? Where can we get ahead?

4. What’s the failure mode? Where does AI go wrong? What’s the incident playbook? What’s the regulatory exposure?

5. What’s the operating model? Who owns AI strategy execution? How are bets reviewed? What gets killed when?

These questions, answered specifically, are the strategy. Not the 30-page document.

The wedge approach

Most companies should pick 2–3 wedges and execute hard.

A wedge is: a specific use case or capability that meets three criteria:

It changes your unit economics or product position materially.
You can prove it within 6 months.
It compounds — winning here creates positions for follow-on bets.

Examples:

A SaaS company picks “AI-driven onboarding” as a wedge. Reduces TTV from 30 days to 3 days; pricing power improves; expansion revenue compounds.
A logistics company picks “autonomous dispatch” as a wedge. Drops dispatch cost 60%; operations leverage compounds; competitive position improves.
A healthcare admin company picks “AI-assisted prior authorization” as a wedge. Reduces denial rates; payer relationships strengthen; expansion follows.

The wedge approach beats the broad approach because it concentrates resources where outcomes are measurable.

What to skip

Three things that show up in AI strategy decks and shouldn’t.

The “AI everywhere” framing. Generic, unfalsifiable, doesn’t drive decisions.

The five-year roadmap. Inappropriate for a capability that changes every quarter. See The Five-Year AI Plan Most Companies Should Refuse.

The center-of-excellence as the strategy. A CoE is an organizational structure; strategy is what the CoE does. Many AI strategy docs confuse the two.

What the board should see

Quarterly AI strategy review:

Wedge progress: status of each named wedge bet.
Investment allocation: dollars across the three categories.
Risk and compliance: incidents, regulatory developments, governance status.
Decisions made: kills, reinvestments, new bets.
Capability shifts: what’s changed in the AI landscape that affects strategy.

This is decision-oriented, not status-oriented. The board’s role is governance over the bet portfolio.

Each spoke covers a specific strategy decision in depth.

Why AI Strategy Is a CFO Conversation, Not a CIO Conversation — the financial frame
The AI ROI Question: How to Answer It Without Lying — what counts as ROI
The Cost Curve No One Tells You About — agent cost dynamics
Why Pilots Stall and How to Design Ones That Don’t — pilot design
The Build vs Buy Decision for AI — sourcing strategy
AI Vendor Selection: A Procurement Playbook — vendor work
Your AI Center of Excellence (When You Actually Need One) — CoE design
The Five-Year AI Plan Most Companies Should Refuse — sequencing
The AI Wedge Strategy — the wedge model in depth
AI and Your Competitive Moat — defensibility

What to do this quarter

Identify your 2–3 wedges. Use the three-criterion test. If you can’t name them, the strategy is too diffuse.
Assess current investment allocation. Across the three categories. Adjust if mostly Category 1.
Set the quarterly review cadence. Strategy decisions in AI need this; decisions made annually go stale.
Kill at least one bet that’s not working. Most portfolios have at least one. Stopping is part of strategy.

FAQ

How does AI strategy differ from digital strategy? Digital strategy in the 2010s was largely about software-enabled processes; AI strategy is about software-enabled judgment and decision-making. The frame, time horizons, and economics are different.

Should we hire a Chief AI Officer? Depends on scale and maturity. For most mid-large enterprises, an AI program lead reporting to the CTO works better than a separate CAIO. The CAIO role makes sense at large scale or when AI is a core product strategic asset.

How much should we spend on AI as % of revenue? Median across enterprises in 2026: 1–3% of revenue. Top 25%: 3–5%. The ceiling is rising; expect 2–7% by 2028 for AI-active companies.

Should we pause AI strategy until the technology stabilizes? No. Capability stabilization is years away; companies pausing are losing positions to companies executing. The right response is portfolio management, not pause.

How do we handle strategy at the BU level vs. enterprise level? Enterprise sets the frame, allocates capital, owns governance. BUs execute within the frame, own the wedges in their domain. The frame should be tight; the execution should be local.

Working with JAIN on AI strategy? We help executive teams build the wedge portfolio and operating model that produces decisions, not documents. Book a 30-minute call.

Related reading: