Enterprise AI stops being a pilot the moment it starts getting paid for like electricity, not an experiment — so when does that actually happen?

This explores the signals that show enterprise AI has stopped being an experiment and become a normal, recurring line item, and what has to change inside a company for that spending to pay off. The clearest signal in the corpus is where the money comes from. In a16z's 2025 CIO survey, AI funding moved out of experimental 'innovation' budgets, which fell from about a quarter of AI spend to 7%, and into core IT budgets that renew every year. Leaders also expect AI budgets to grow about 75% in the following year Are enterprises actually spending more on AI that works?. So the shift is less about one breakthrough moment and more about accounting: AI becomes mainstream when it is paid for like email or cloud storage rather than like a pilot.

A second signal is that companies stop building their own tools and start buying them. The same survey found enterprises turning toward third-party AI apps, because keeping homegrown tools current with each new model release cost too much in upkeep to stay worth it Are enterprises abandoning custom AI apps for third-party solutions?. That pattern is typical of a maturing market: once the technology is a commodity, the competitive edge moves from owning the stack to using it well. Adoption still spreads unevenly. OpenAI's enterprise usage data shows larger, R&D-heavy firms adopting first, and inside those firms marketers and early-career staff use the tools far more than executives do Who adopts enterprise AI first and how do they use it?. Anthropic's usage data adds a timing pattern: as adoption deepens, people move from handing AI whole tasks toward working alongside it and learning from it Does AI adoption follow wealth and mature over time?.

The less obvious finding is that mainstream spend and mainstream payoff are different milestones. McKinsey's 2026 survey found that 80% of respondents report personal productivity gains, but only 37% see enterprise profit improve. The roughly 6% of 'high performers' got there by redesigning workflows, keeping humans in the loop to check AI output, managing costs, and having senior leaders own the effort, not by picking a better model Why do individual AI gains not translate to business results?. A survey of executives points the same way: perceived gains run ahead of measured ones, partly because revenue shows up later than operational improvements Do AI productivity gains feel larger than they actually measure?. At the macro level, Brynjolfsson reads 2025 data as the AI productivity 'J-curve' finally turning upward, though other economists say the evidence outside tech leaders is still thin Is AI productivity finally showing up in economic data?.

What holds the transition back is mostly organizational. Evans argues that making tools easier to build doesn't solve two problems: most workers don't see their own tasks as automatable, and enterprise adoption depends on decisions that cross departments and budget cycles Does easier tool-building actually solve enterprise adoption problems?. Capability claims can also mislead buyers. An analysis of 960 real occupational workflows finds that agents that win benchmark contests still fail at long, multi-step professional work, so benchmark scores are a poor guide to economic readiness Why do agent benchmarks not predict real economic value?. Where substitution is already happening, it compounds. Firms with more AI exposure replace contract labor faster and more cheaply than others, which suggests that internal AI skill builds on itself rather than spreading evenly Do firms substitute labor for AI at different rates?.

In short, the corpus suggests AI became mainstream *spend* around 2025, when budgets moved into recurring IT lines. It becomes mainstream *value* only for companies that rebuild their workflows around it, and most haven't done that yet. The gap between those two milestones is where the most interesting open questions are.