HR did not wait for an AI strategy deck. ChatGPT usage data says people teams are already doing work that used to live in other departments, while leadership readiness for AI-enabled teams is still near zero. That gap is the recruiting pain, not the model itself.
OpenAI’s Work at the Frontier analysis looked at more than 800,000 work-related ChatGPT messages from U.S. users. After stripping generic tasks like writing and scheduling, 43.5% of occupation-specific messages still mapped to a different occupation than the user’s. For human resources workers, that “task crossover” rate hits 69%, behind only customer experience (77%) and design (75%).
In plain English: when HR opens ChatGPT for specialized work, about seven in ten of those prompts look like someone else’s job.
What HR is actually asking the model for
Inside those outside-HR prompts, OpenAI’s occupation map is blunt. HR users most often crossed into:
Marketing tasks (23%)
Engineering tasks (18%)
Finance tasks (16%)
That is not “HR is learning to code.” It is recruiters and people partners using the model to write employer-brand copy, scrape patterns from messy data, and stress-test headcount or budget language without opening a ticket to Marketing, Analytics, or FP&A.
OpenAI’s framing matters for TA leaders: AI lets people do the work themselves instead of handing it off. Job descriptions, leveling grids, and “who owns what” still assume handoffs. Usage does not.
Task crossover also shows up more in small businesses, where people already wear multiple hats. If your ATS workflow still assumes a clean RACI between TA, marketing, and finance, the prompt history is already ahead of your process docs.
The leadership number that makes this dangerous
Separately, ManpowerGroup Talent Solutions research with Everest Group surveyed 80 C-suite, CHRO, and senior talent acquisition leaders in the U.S. and U.K. Only
