a16z: Codex Non-Technical User Growth in Legal Sector Reaches 108x, Sales 41x, Recruitment 41x
a16z data shows that heavy AI users are emerging outside of technology. Since February, the fastest adoption rates for Codex have been in: legal at 108x, sales at 41x, recruitment at 41x, marketing at 26x, and healthcare at 24x.
Codex has expanded from a coding tool to an autonomous digital colleague capable of decomposing tasks and running for extended periods.
Legal practitioners use it to handle large volumes of documents and contract comparisons, shifting knowledge work towards programmability.
Adoption in non-technical roles is accelerating, with AI penetration in high-value white-collar positions increasing.
Capital and attention are flowing towards cross-industry AI agent applications, with the proliferation of event-driven tools accelerating, traditional professional service processes under pressure, and vertical AI and agent platforms benefiting.
Source: Public Information
ABAB AI Insight
a16z continues to track enterprise AI adoption data, previously noting that coding, customer service, and search are major use cases. This time, it focuses on the explosive growth of Codex in non-technical sectors, continuing its observation of AI's expansion from developer tools to general knowledge work.
The legal sector shows the highest growth rate, motivated by the document-intensive nature of the work, which aligns well with AI, strategically replacing repetitive analysis and comparisons with agent capabilities.
Rapid growth of legal AIs like Harvey, along with the rise of vertical tools in healthcare and recruitment, indicates that we are currently in an accelerated phase of AI diffusion from within technology to across all industries.
Essentially, this is a form of technological substitution. Coding agent capabilities are spilling over into non-coding tasks, reducing professional barriers through natural language task decomposition, and reshaping white-collar workflows.
ABAB News · Cognitive Laws
The fastest-growing users often lie outside the initially designed domains.
Document-intensive industries are the first to be rewritten by agents.
Once tools can decompose tasks, user boundaries will disappear.