# The State of Enterprise AI The State of Enterprise AI Source: https://www.youtube.com/watch?v=FzGPptXjM0w Recap page: https://rapidrecap.app/video/FzGPptXjM0w Generated: 2025-12-11T16:07:40.365+00:00 --- ## Quick Overview Enterprise AI adoption is accelerating and deepening, with recent reports from OpenAI and Menlo Ventures showing massive growth in usage, shifting focus from surface-level experimentation to deeper integration and workflow automation, while also revealing a growing gap between AI leaders and laggards, especially regarding the adoption of custom models and advanced AI architectures. **Key Points:** - Enterprise AI adoption is accelerating, evidenced by OpenAI reporting weekly messages growing 8x and ChatGPT Enterprise seats increasing 9x year-over-year since November 2024 (0:00:06). - Frontier workers (95th percentile in adoption intensity) generate significantly more messages (6x overall, 17x for coding) than median workers, highlighting a growing usage gap (7:44). - The AI application layer shows startups gaining ground on incumbents, capturing 63% of new revenue for every $1 earned by incumbents, up from 36% last year (11:44). - Enterprises are increasingly prioritizing buying AI solutions (76% purchased in 2025) over building them internally (24% built), a shift from 2024 when the split was 53% purchased vs. 47% built (13:46). - Coding is identified as Generative AI's first 'Killer Use Case,' with projected departmental AI spending hitting $7.3 billion in 2025, $4.0 billion (55%) of which is for coding (10:01). - Horizontal AI (Copilots) spending ($8.4 billion) dwarfs Agent spend ($750 million), with Copilots dominating 86% of that category (15:16). - Real production AI architectures remain surprisingly simple, with only 16% of enterprise deployments qualifying as 'true agents' (15:27). ![Screenshot at 0:07: The animated graphic illustrating 'The State of Enterprise AI' shows a boardroom meeting with human executives on one side and robotic counterparts on the other, symbolizing the growing presence and integration of AI in business decision-making and closing the gap between leaders and laggards.](https://ss.rapidrecap.app/screens/FzGPptXjM0w/00-00-07.png) **Context:** This video discusses findings from two recent reports on Enterprise AI: one from OpenAI and another from Menlo Ventures, both published around December 2025. The content focuses on the rapid adoption, usage patterns, and maturation of AI within businesses, contrasting the usage between AI leaders ('frontier workers' or firms) and laggards, and detailing where the AI budget is being allocated across different application layers like coding and agents. ## Detailed Analysis The video synthesizes findings from recent reports on Enterprise AI from OpenAI and Menlo Ventures, confirming rapid and deepening adoption. OpenAI data shows weekly enterprise messages grew 8x and ChatGPT Enterprise seats grew 9x year-over-year since late 2024 (0:00:06). A significant gap exists between AI leaders (frontier workers, 95th percentile) and median workers; frontier users send 6x more overall messages and 17x more coding-related messages (7:53). In the AI application layer, startups are gaining ground on incumbents, capturing 63% of new revenue for every $1 earned by incumbents, up from 36% previously (11:44). The trend shows a clear shift toward buying AI solutions over building them internally, with purchased solutions rising from 53% in 2024 to 76% in 2025 (13:46). Coding is highlighted as the first 'killer use case' for GenAI, projected to account for $4.0 billion (55%) of the $7.3 billion departmental AI spending in 2025 (10:01). Furthermore, horizontal AI spending ($8.4 billion) vastly outweighs agent spending ($750 million), with Copilots being 10x bigger than Agents for now (15:16). However, despite the excitement, the reports indicate that true, complex AI agent architectures remain nascent, with only 16% of enterprise deployments qualifying as true agents, suggesting most deployed solutions are still simple, fixed-sequence or routing-based workflows (15:27). Open-source LLM adoption is lagging behind frontier models, with Meta's Llama share dropping from 19% to 11% due to stagnation (14:24). ### Adoption Growth Metrics - OpenAI reports 9x year-over-year growth in ChatGPT Enterprise seats (0:00:06) - Frontier workers send 6x more messages than median workers overall, and 17x more coding-related messages (7:53). ### Startup vs. Incumbent Dynamics - Startups captured 63% of new revenue for every $1 earned by incumbents in the AI application layer, up from 36% last year (11:44) - Incumbents have entrenched advantages like data moats and sales teams (11:52). ### Build vs. Buy Trends - Enterprises are shifting towards purchasing AI solutions, with purchased solutions rising from 53% in 2024 to 76% in 2025, while built-in-house solutions dropped from 47% to 24% (13:46). ### Budget Allocation - Departmental AI spending is projected at $7.3B in 2025, with Coding dominating at $4.0B (55%) (10:01) - Horizontal AI (Copilots) spending ($8.4B) is 10x larger than Agent spending ($750M) (15:16). ### AI Architectures - Real production architectures remain simple; only 16% of enterprise deployments qualify as true agents (15:27) - Customization patterns like Prompt Design dominate production usage over complex methods like RAG or RL (15:34). ### Open-Source LLMs - Adoption is lagging for open-source models, with Meta's Llama share declining from 19% last year to 11% today (14:24). ![Screenshot at 0:07: Illustration of a boardroom meeting featuring human executives opposing robotic counterparts, symbolizing the theme of enterprise AI adoption and the leader/laggard gap.](https://ss.rapidrecap.app/screens/FzGPptXjM0w/00-00-07.png) ![Screenshot at 11:44: Bar chart comparing startup revenue capture \($2 in revenue for every $1 earned by incumbents\) vs. incumbent revenue capture \(63% market share\) on the AI application layer.](https://ss.rapidrecap.app/screens/FzGPptXjM0w/00-11-44.png) ![Screenshot at 13:46: Donut charts showing the shift in AI adoption methods from 2024 \(53% purchased\) to 2025 \(76% purchased\), indicating preference for buying over building.](https://ss.rapidrecap.app/screens/FzGPptXjM0w/00-13-46.png) ![Screenshot at 15:16: Bar chart illustrating that Horizontal AI spending \(Copilots at $7.2B\) is 10x greater than Agent spending \($750M\) for now.](https://ss.rapidrecap.app/screens/FzGPptXjM0w/00-15-16.png) ![Screenshot at 15:38: Bar chart comparing customization approaches used in production, showing Prompt Design \(39% enterprise, 31% startup\) as the most common method.](https://ss.rapidrecap.app/screens/FzGPptXjM0w/00-15-38.png)