AI investment is highly concentrated. While median firms spend trivially (around $12/employee/month), the top 1% spend thousands. This intense use by a few explains why AI's impact isn't yet visible in broad productivity statistics.
Companies deeply integrating AI are hiring more entry-level workers than expected. This strategy aims to bring in a workforce that is naturally proficient with the latest AI tools, valuing current skills over traditional experience.
Unlike traditional enterprise software where companies consolidate to a single vendor for better pricing, advanced AI users actively use multiple model providers. This suggests a "best tool for the job" approach prevails over single-vendor lock-in.
Unlike sticky enterprise software, the AI model market is highly contestable. Leadership between players like OpenAI and Anthropic can shift in months, driven purely by which company releases the better-performing model, posing a risk to long-term valuations.
While a dataset may skew towards tech-forward businesses, this is a feature, not a bug. These early adopters signal where the market is headed, allowing for predictive insights into future technology trends before they become mainstream.
True self-hosting of open-source AI models is rare due to complexity. Instead, companies pay for 'router' platforms that provide cheap, managed access to various open-source models, making their adoption trackable via spending data.
Fears of AI-driven job loss are misplaced for now. Research shows that only firms adopting AI intensely—using advanced agents and multiple models—experience accelerated headcount growth. Casual adoption has no discernible impact on hiring rates.
Transactional data on paid AI tools likely undercounts overall adoption. It fails to capture the widespread use of free tiers (like Gemini in Google Workspace) or employees using personal accounts, creating a blind spot in market analysis.