Analysis
I've spent this whole issue writing up nine- and ten-figure AI rounds -- Nvidia possibly anchoring Anthropic's $2 trillion IPO, Cognition and Harvey re-rating on real revenue, Instinct chasing a $10 billion mark a year after founding. Fortune's reporting this week is the number that should temper all of it: companies are on track to spend $2.5 trillion on AI in 2026, a 47% jump from 2025, and only about 5% of employees with GenAI access are actually using it effectively. Poor and average performers -- more than half of a typical organization -- dominate the initial adoption, not the power users the spending assumes.
This isn't hypothetical anymore; it's showing up in the data on the other side of the relationship too. Gen Z hopefulness about AI dropped to 18% from 27% year over year, and one study cited in the piece found AI polling less popular than ICE among younger workers. That's a genuinely bad signal for any company betting its AI-agent product on next-generation workforce enthusiasm carrying adoption forward.
“Gen Z hopefulness about AI dropped to 18% from 27% year over year, and one study cited in the piece found AI polling less popular than ICE among younger workers.”
Salesforce's own quiet retreat from Agentforce branding, covered elsewhere in this issue, is the clearest corporate tell I've seen all year that the FOMO-to-hangover thesis is real and not just an academic framing. You don't rebrand away from your own AI-agent strategy two years in unless the enthusiasm outran what customers actually wanted.
For founders and GPs, the read isn't 'AI adoption is failing' -- $2.5 trillion in spending doesn't happen on a failing thesis. It's that the current wave of enterprise AI deployment is mismatched to how most employees actually work, and the startups that win the next 18 months are the ones solving that mismatch -- repositioning AI as a thinking-enhancement tool rather than a task-automation threat -- rather than the ones selling more automation into an organization that's already resisting it.
Room for disagreement: every AI adoption curve in history has looked like a disappointment measured too early -- email, the internet and cloud computing all had multiyear gaps between spending and measurable productivity gains before the curve bent upward. $2.5 trillion in 2026 spending could simply be the up-front infrastructure cost of a curve that pays off in 2028 or 2029, and a 5% effective-use rate this early might be entirely normal rather than alarming. If GenAI's productivity numbers look meaningfully better a year from now, this take will look like it mistook an adoption lag for a structural problem.
