Squinting Strategy: Seeing the Pattern Without the Detail
What 'squinting strategy' means — and how deliberate blur helps leaders read markets, portfolios and AI bets in 2027.
Squint at the AI agent market until the vendor logos blur, and two numbers survive. Gartner expects 40% of enterprise applications to ship with task-specific AI agents by the end of 2026 — and it expects more than 40% of agentic AI projects to be cancelled by the end of 2027. Both forecasts come from the same firm, and both are probably right. Detail hides that pattern; every vendor deck is a success story. The blur reveals it: a technology spreading everywhere and failing often, at the same time, for structural reasons. That is what squinting is for.
The idea
Detail is where strategies go to hide. A portfolio review with 200 line items, a market map with 40 segments, a dashboard with 60 KPIs: each item is defensible, and the whole is unreadable. Squinting is the deliberate opposite move — reduce resolution until only the load-bearing patterns remain: Which three customer problems produce most of the value? Which two capabilities produce most of the advantage? Where is the one dependency that can kill the plan?
How to practise it
- The one-page test. If the strategy cannot survive compression to one page, the compression is the strategy work you skipped.
- Squint at the portfolio quarterly. Group everything by the customer problem it solves; orphans and duplicates announce themselves.
- Squint at your AI bets. The same blur that clarifies strategy clarifies agent portfolios: most estates collapse into a handful of functions — IT, knowledge work, support and operations — fund them in that order of evidence.
More analysis in the topics hub, including the 2027 trends briefing — a squint at the whole field.
A squint test, applied: your agent portfolio
Try the exercise on the noisiest portfolio most companies now own — their AI bets. Spread every pilot, copilot and "agentic initiative" on one page and squint. The detail says forty projects, twelve vendors, nine business cases. The blur says something simpler, and the market data backs it: agents are becoming a standard feature of enterprise software — Gartner expects 40% of enterprise applications to include task-specific agents by the end of 2026, up from under 5% in 2025 (Gartner, 2025). You will not choose most of your agents; they will arrive inside tools you already buy. The portfolio question is only about the few you build or bet on deliberately.
Squint again and the second pattern appears: attrition. Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, killed by escalating costs, unclear business value and inadequate risk controls (Gartner, 2025). McKinsey's deployment data shows why the middle looks so crowded: 62% of organisations are experimenting with agents but only 23% are scaling one anywhere, and in any given function no more than 10% have scaled (McKinsey, 2025). Experimentation is the default state; almost nothing has yet earned its budget.
So the squint test for an AI portfolio has three questions, and they are the same three questions this essay asks of any strategy. Which two or three agent bets map to a customer problem that produces measurable value — not a demo? Which single dependency — one vendor, one data source, one team's attention — can quietly kill the rest? And which projects exist because the detail was never compressed: defensible line by line, indefensible as a whole? Gartner's cancellation forecast is, read properly, a squinting forecast. The projects that die will not be the ones with the worst technology. They will be the ones nobody ever stepped back from.
Frequently asked questions
What does squinting mean as a strategy technique?
Squinting is the deliberate opposite of detail: reduce resolution until only the load-bearing patterns remain. Detail is where strategies go to hide — a portfolio review with 200 line items, a market map with 40 segments, a dashboard with 60 KPIs can be defensible item by item and unreadable as a whole. The blur answers three questions: which three customer problems produce most of the value, which two capabilities produce most of the advantage, and where is the one dependency that can kill the plan?
How do you run a squint test on a strategy or portfolio?
Start with the one-page test: if the strategy cannot survive compression to one page, the compression is the strategy work you skipped. Then squint at the portfolio quarterly: group everything by the customer problem it solves, and orphans and duplicates announce themselves. Finally, apply the same blur to your AI bets: most estates collapse into a handful of functions — IT, knowledge work, support and operations — so fund them in that order of evidence.
Why are so many agentic AI projects expected to be cancelled?
Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, killed by escalating costs, unclear business value and inadequate risk controls (Gartner, 2025). McKinsey's deployment data shows the crowded middle: 62% of organisations are experimenting with agents but only 23% are scaling one anywhere, and in any given function no more than 10% have scaled (McKinsey, 2025). Experimentation is the default state; almost nothing has yet earned its budget.
How can Gartner predict rapid AI agent adoption and mass project cancellation at the same time?
Both are probably right. Gartner expects 40% of enterprise applications to ship with task-specific AI agents by the end of 2026, up from less than 5% in 2025. Most agents will arrive inside tools you already buy. It also expects more than 40% of agentic AI projects to be cancelled by the end of 2027. Expansion and shakeout are the same trend: the technology spreads everywhere, and the projects that die are the ones nobody ever stepped back from.
Sources
- Gartner. (2025). "Gartner Predicts 40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026, Up from Less Than 5% in 2025". Gartner Newsroom.
- Gartner. (2025). Press release predicting more than 40% of agentic AI projects will be canceled by end of 2027 (June 25, 2025), as documented in "AI Agents in Production: Why Adoption Stats Conflict". AI Invasion.
- McKinsey & Company. (2025). "The State of AI in 2025: Agents, Innovation, and Transformation". McKinsey Global Survey (n=1,993).