The legal and governance department is the biggest artificial intelligence user among company departments, according to research conducted by AI governance and control platform Harmonic Security.
Legal and governance accounted for 19.5 percent of AI use hours, above departments like go-to market (17.7 percent), design and development (13.3 percent), strategy (11.9 percent), infra and security (9.4 percent), finance (8.9 percent), and operations (5.1 percent). Procurement used AI the least, accounting for just 1 percent of AI use hours.
Harmonic described the statistics as “striking” given the legal profession’s reputation for sluggishness in technology adoption. It acknowledged that AI significantly aided text-heavy, reference-intensive tasks like contract review, regulatory research, policy drafting, and compliance analysis; nonetheless, it cautioned in a press release that legal work included “among the most sensitive categories of content an employee could put into an AI prompt”, such as contracts, litigation strategy, regulatory analysis, and IP review.
Legal departments on enterprise plans made up 32.3 percent of AI hours, and Harmonic noted that legal teams have generally adopted enterprise AI. ChatGPT is most widely used by this department, accounting for 67 percent of legal AI activity; Harmonic said it was “by far the largest function-tool concentration in the dataset.”
Claude is more widely adopted by strategy (41 percent) and finance (40 percent) departments. Deepseek is generally used by the go-to market department, while Microsoft Copilot accounts for 32 percent of infrastructure and security departments’ AI use.
Harmonic also reported that AI users with personal accounts often used it for work, with business use accounting for 64.5 percent of activity. Copilot Free is the most widely used, making up 80.2 percent of business activity followed by Claude Pro (75.5 percent) and ChatGPT Plus (71.2 percent). Thus, Harmonic suggested that companies monitor free accounts, given the input of corporate information into platforms that cannot be monitored by IT.
Thus, Harmonic said AI effectiveness should not be quantified through token and usage counts, which it said measured adoptions rather than value. Instead, use-case numbers are what link AI activity to result and highlight accelerated workflows, judgment enhancement, and cases with direct audit implications.
Harmonic examined 1,935,247 classified AI-session minutes over seven weeks until April 2026.