Neil Hoyne, Chief Strategist at Google, author of Converted (Penguin Random House), and Executive Faculty at The Wharton School, delivered a keynote at GE Vernova Accelerate 2026 on how organizations move from data to decisions. Companies that mentioned AI in earnings calls saw their stock grow 42% faster than peers that did not, according to Factset Insights (2024). Yet CEOs themselves estimate that 35% of AI projects are designed to signal innovation to investors without delivering business value, per the Global AI Confessions Report: CEO Edition from Harris Poll (2026). Separately, 71% of CTOs and CIOs report that executive leadership has an unrealistic expectation around the ROI of AI, according to Solvd’s CIO and CTO Insights: AI Research (2026). For many executives, the driving force behind AI investments is simpler: a fear of missing out. The keynote presented four organizational shifts, starting with a reframe: the best companies do not ask “What is our AI strategy?” They ask “What in our company’s strategy can AI help us achieve?”
Build Hypotheses Instead of Blueprints
Most AI initiatives pass through the same organizational sequence: the boss decides whether to support the idea, the broader team weighs in, engineering asks whether it fits the next sprint, legal reviews for compliance, and design checks brand alignment. By the time a concept reaches execution, it has been shaped by internal navigation more than business potential.
Organizations that treat AI concepts as hypotheses to test, rather than solutions to defend, create more room to learn. Hypotheses invite exploration and challenge. Solutions create rigid ownership and expectations.
Three principles support this shift. Foster hypotheses over pitches and create room for the organization to explore. Map reality from the ground up, because leaders often miss daily friction; frontline employees and customers, those with the least organizational power, frequently have the best view of where processes suffer and where data may help. And reward innovation over navigation by recognizing people who identify new, audacious opportunities rather than those who work the system to drive adoption.
Define What Counts as Evidence
Two out of every three CEOs have ignored insights based on data because those insights contradicted their intuition, according to a KPMG Global CEO survey (2019). The data existed. The decision went the other way.
Before acting on any AI initiative, an organization needs to define what evidence it requires. Five levels of proof typically drive decisions in practice:
| Level | What Triggers the Decision |
| Existence | Someone found data and assumed it should be used |
| Boss | A board member thought the company should build something |
| Keynote | Someone heard a competitor doing it at a keynote |
| Science | Published, peer-reviewed research proved a point |
| Experiment | A reasonable geo-test validated the hypothesis |
Making the standard explicit accelerates every decision that follows. Organizations should ask whether AI initiatives operate under different evidence requirements than the rest of the business, and if so, why. The evidence threshold determines how fast the organization moves. Circulating findings broadly with colleagues and partners avoids redundant discovery, especially in a field as new as AI where collaboration is key.
Test What Is Close, Not What Is Perfect
Data that is close to clarifying the most promising hypotheses matters more than perfect data that answers everything at once. Organizations may not know everything, but they probably know enough.
AI tools can serve as testing engines before becoming full solutions. Generating more content, creating personalized creatives, and running faster experiments all help discover what drives value before the complexity of enterprise-wide deployments.
One less obvious application: large language models may replicate traditional survey panels accurately without the cost or time. While others chase short-term use cases, these capabilities can build deeper insights over time.
Lead With Growth, Not Mandates
AI adoption works as a movement, not a mandate. Redefining role descriptions goes beyond mapping new skills to existing positions. Understanding how employees currently identify and create value, then aligning those perspectives with AI roadmaps, reveals the path of least resistance for adoption.
Cost savings may please shareholders today but kill transformation tomorrow. Cutting costs is the refuge of companies that cannot prove ROI. A vision that emphasizes expansion and opportunity comes first.
Some compelling opportunities may exceed the organization’s current readiness. Forcing them through could cost more than they are worth. “Not yet” is strategic patience, and the organizational goal, as the keynote concluded, is to outpace rather than outplan.
This article is based on a keynote by Neil Hoyne, Chief Strategist at Google, author of Converted (Penguin Random House), and Executive Faculty at The Wharton School, delivered at GE Vernova Accelerate 2026. Lucian Fogoros of IIoT World attended the event. AI tools were used to help summarize and organize the content. Reviewed and edited by the IIoT World editorial team.
Related from IIoT World
- From Sensor to Decision: What Determines Speed in Industrial AI
- The Productivity Paradox: More Data, More Waste in Manufacturing
- Industry 4.0, Fifteen Years In: Where Are We Now?
Frequently Asked Questions
1. Why do AI projects fail to deliver ROI?
CEOs estimate that 35% of AI projects are designed to signal innovation to investors without delivering business value, according to the Global AI Confessions Report: CEO Edition from Harris Poll (2026). Meanwhile, 71% of CTOs and CIOs report that executive leadership has an unrealistic expectation around the ROI of AI, per Solvd’s CIO and CTO Insights: AI Research (2026). Many organizations start by asking “What is our AI strategy?” instead of asking which parts of the existing business strategy AI can accelerate. The result is investment driven by perception rather than measurable business outcomes.
2. How should organizations set evidence standards for AI investments?
Organizations should define what evidence they require before acting on an AI initiative, based on the risk and value of the decision. Five common levels of proof range from using any available data (existence) to running a controlled geo-test (experiment). Making the standard explicit accelerates decision-making because the evidence threshold determines organizational velocity. Organizations should also ask whether AI initiatives operate under different evidence standards than other business decisions.
3. What is strategic patience in AI adoption?
Strategic patience means recognizing when a compelling AI opportunity exceeds the organization’s current readiness. Forcing a project through before the organization can support it costs more than waiting. “Not yet” is a form of strategic patience, a deliberate decision to act when conditions allow the investment to succeed rather than when competitive pressure or executive FOMO demands it.
4. How can organizations avoid AI FOMO in investment decisions?
Companies that mentioned AI in earnings calls saw stock growth 42% faster than peers that did not, according to Factset Insights (2024). That stock premium creates pressure to announce AI initiatives regardless of business value. Organizations counter this by building hypotheses instead of blueprints, setting explicit evidence standards for investment decisions, and framing AI adoption around growth and opportunity rather than fear of falling behind.
About the author
