what 23,000 workers told Gallup they use AI for every day

Gallup asked 23,717 U.S. workers what AI helps them with at work. The top three answers: drafting written content, summarizing information, generating ideas. Not agents. Not autonomous pipelines. Three verbs, same as always — just faster.

Five surveys from late 2025 through early 2026 all arrive at the same place. Here’s the breakdown by who you are.

If you write things

MIT’s study of 453 professionals found ChatGPT cut task time 40% and improved output quality 18% on routine business writing — cover letters, delicate emails, cost-benefit analyses. Two years on, Gallup confirms it’s the most common daily AI task in America: draft, summarize, brainstorm, repeat.

Tools people actually use: ChatGPT, Claude, Grammarly, Microsoft Copilot.

If you write code

GitHub Copilot hit 20 million all-time users in July 2025. GitHub’s own controlled study (95 developers) found Copilot users finished tasks 55% faster. Stack Overflow’s 2025 survey of 49,000+ developers put AI tool adoption at 84%.

LangChain’s State of Agent Engineering survey (1,340 professionals, December 2025) found coding assistants are the most-used daily agent type by a significant margin — respondents kept citing Claude Code, Cursor, Copilot, Windsurf, and Amazon Q. Research agents came second.

Tools: Cursor, Claude Code, GitHub Copilot, Windsurf.

If you do research

Research and deep-research agents are the #2 daily use case in the LangChain survey. In practice this means asking ChatGPT or Claude to find, synthesize, and explain — not “deploying an agent.” ChatGPT commanded 59.47% of US AI chatbot web traffic in December 2025, though that lead has been narrowing fast as Gemini and Claude have grown.

Tools: ChatGPT, Perplexity, Claude, Gemini.

If you build workflows

n8n leads the agentic workflow category among builders, per Lenny’s Newsletter’s survey of ~1,750 tech workers. The realistic use case: deterministic automations with LLM nodes — not agents making open-ended decisions, but workflows that route, summarize, and trigger on cue.

Tools: n8n, Zapier, Make.com.

Where enterprise “AI agents” actually live

In production deployments, customer service is the single largest use case — 26.5% of all enterprise agent deployments per LangChain, with research and data analysis at 24.4%. PwC’s survey of 308 executives found data analysis is the AI task they trust most: 38% ranked it their highest-confidence area for handing off to agents.

When companies say “we deployed AI agents,” they mostly mean customer service bots and data pipelines. That’s a different animal from what an individual does at their desk.

The takeaway

Nobody says “I used my research agent today.” They say “I asked ChatGPT.” Across five major surveys and tens of thousands of workers, daily AI use collapses into three verbs: write faster, code faster, find things faster.

The agentic future is here. It’s a chatbot handling support tickets and a coding assistant writing your boilerplate. Make your peace with that, then go save the 40%.