01 · Executive AI Strategy¶
At the executive level, "AI strategy" stops being a set of project choices and becomes a capital-allocation and competitive-positioning question: where does AI change what's possible in your industry, where is it merely table-stakes efficiency, and how much of the company's scarce budget and attention should go toward each. Getting this wrong in either direction is expensive — over-investing chases hype into commoditized capability, while under-investing cedes a genuine structural advantage to a competitor who moves first. This module gives you the frameworks to make that allocation call deliberately, and to communicate it credibly to a board.
1. The AI strategy positioning matrix¶
Before allocating a dollar, classify where AI sits relative to your business, because the right level of investment differs sharply by quadrant.
| Low differentiation potential | High differentiation potential | |
|---|---|---|
| Core to your business model | Efficiency play — automate to reduce cost, match industry baseline (e.g., AI-assisted customer support for most companies) | Strategic bet — this is where competitive advantage is actually won or lost; deserves disproportionate investment and executive attention |
| Peripheral to your business model | Buy, don't build — commodity vendor tools, minimal internal investment | Selective build — only if a peripheral capability could become core later (rare; verify before over-investing) |
Most executive strategy mistakes are misplacing a capability in this grid — treating a genuine strategic bet as a peripheral efficiency play (under- investing in what actually differentiates you), or treating a commodity capability as strategic (over-building an in-house solution for something every vendor already does well).
2. The three-horizon AI investment model¶
Borrowed from general innovation strategy, adapted for AI's specific timeline and risk profile:
| Horizon | Time frame | Investment character | Example |
|---|---|---|---|
| Horizon 1 | 0-12 months | Efficiency and quality improvements to existing AI systems; low risk, fast payback | Improving an existing recommendation model's accuracy by 3-5 points |
| Horizon 2 | 1-3 years | New AI-enabled products or capabilities, built on proven techniques | Launching a new LLM-powered product line adjacent to your core offering |
| Horizon 3 | 3+ years | Exploratory bets on capabilities not yet proven at your scale or in your domain | Foundation-model fine-tuning as a core IP moat; agentic systems replacing a category of internal workflow |
A healthy portfolio allocates roughly 60-70% of AI investment to Horizon 1, 20-30% to Horizon 2, and 5-10% to Horizon 3 — adjust based on your industry's rate of AI-driven disruption, but be suspicious of a portfolio that's 90%+ Horizon 1 (you're purely playing defense) or heavily weighted to Horizon 3 (you're speculating without a proven base to fund it).
3. Setting strategy that survives contact with the board¶
A board-credible AI strategy answers four questions explicitly, in this order — skipping ahead to "here's our AI roadmap" without the first two is the most common reason boards push back.
| Question | What a weak answer looks like | What a strong answer looks like |
|---|---|---|
| Where does AI change our competitive position? | "AI is transforming everything" | "In our market, AI-driven personalization is becoming the primary switching cost; we're 8-12 months behind our closest competitor on this specific capability" |
| What's our realistic time horizon to matter? | No timeline, or an unfounded "6 months to transform the company" | "18 months to reach parity on personalization; 3 years to build a genuine moat via proprietary data flywheel" |
| What's the investment and what's the expected return? | A headcount number with no ROI model | The ROI framework from Level 3 Module 7, applied at portfolio scale |
| What's our biggest risk if we're wrong? | Unaddressed | "If personalization commoditizes faster than expected, this investment becomes table-stakes cost rather than differentiation — we've built in a 12-month re-evaluation checkpoint" |
4. Common executive strategy failure modes¶
- Strategy-by-headline. Committing to "become an AI-first company" because of competitive or investor pressure, without the positioning analysis in section 1 to justify where that investment actually goes.
- Treating AI strategy as IT strategy. Delegating the entire question to a CTO/CIO without executive-level input on which business capabilities matter — AI strategy is a business strategy question with technical execution, not a technical question with business framing.
- No re-evaluation cadence. AI capability and competitive dynamics move faster than a typical 3-year strategic plan cycle; commit to revisiting the positioning matrix at least annually, informed by real portfolio performance (Level 3 Module 7's ROI tracking) not vendor hype cycles.
Worked example¶
A regional grocery chain, Thornbury Markets, faced pressure from its board after a national competitor launched an AI-driven dynamic pricing and personalized promotion system. The initial instinct from several executives was to match feature-for-feature immediately.
Applying the positioning matrix first: personalized promotions were assessed as core to the business (directly drives basket size and loyalty) and genuinely high-differentiation in this market (few regional grocers had the data infrastructure to do it well) — a Horizon 2 strategic bet, not a Horizon 1 efficiency play matched reactively. Dynamic pricing, in contrast, was assessed as core but lower-differentiation in this specific market (customers in the served region were shown by internal survey data to be unusually price-sensitive to perceived fairness, making aggressive dynamic pricing a brand risk rather than an advantage) — reclassified as a "selective, cautious" investment rather than a race to match the competitor feature-for-feature.
The board strategy presentation led with this distinction explicitly: $2.4M committed over 18 months to personalized promotions (a genuine Horizon 2 bet, with the ROI model from Level 3 Module 7 built in before funding was approved), and a much smaller $300K exploratory Horizon 3 allocation to test dynamic pricing carefully in two pilot stores with close customer-sentiment monitoring before any wider rollout. The board approved the plan specifically because it distinguished where AI mattered strategically from where matching a competitor's headline feature would have been strategy-by-headline.
How It Actually Works¶
Thornbury's split verdict on dynamic pricing versus personalized promotions is a direct application of a mechanism worth naming precisely: two capabilities can look superficially similar (both "AI features a competitor just shipped") while having opposite risk-adjusted return profiles because differentiation potential and market-specific customer sensitivity are independent variables that happen to combine differently in each case. A capability's business value isn't an intrinsic property of the technology — personalized promotions and dynamic pricing use overlapping technical machinery — it's a function of how that capability interacts with this specific market's competitive structure and customer psychology. Reacting feature-for-feature to a competitor's launch implicitly assumes the technology itself is what matters, when the actual determinant is whether your customers and competitive position make that specific capability an advantage or a liability here — which is exactly why the survey data on regional price-sensitivity, not any property of the pricing algorithm itself, was the deciding input.
The 60-70/20-30/5-10 horizon allocation isn't an arbitrary convention — it mirrors a standard risk-and-information structure found in any portfolio under genuine uncertainty: Horizon 1 investments have the most available information (proven techniques, known ROI patterns) and the fastest feedback loops, so they can absorb the largest share of capital with the least risk of total loss. Horizon 3 investments have the least available information — you're placing a bet before you can validate the core assumption — so a small allocation limits the downside of being wrong while still buying the option value of being right early. A portfolio skewed too far toward Horizon 1 has optimized entirely for certainty and given up any chance at the disproportionate returns that come from being early on a real structural shift; skewed too far toward Horizon 3, it's spending capital on unvalidated bets faster than Horizon 1's returns can replenish it — which is precisely the failure mode "you're speculating without a proven base to fund it" describes mechanically, not just rhetorically.
Exercise¶
Take your own organization (or Thornbury Markets, pre-decision).
- Plot your top 3-5 AI initiatives or candidate initiatives on the positioning matrix in section 1. For each, state which quadrant and why — cite a specific competitive or market fact, not intuition.
- Estimate your current Horizon 1/2/3 split using section 2. Is it within the healthy range, and if not, what does that imply about your organization's risk posture (too defensive, or too speculative)?
- Draft the four-question board answer from section 3 for your single highest-priority AI initiative, in the specific order given.