一个小企业主用AI独立起草营销文案、审阅合同、做基础财务分析。一个销售用AI分析客户数据集——以前这是数据分析师的活。一个市场人员在没有开发人员帮助的情况下自己排查网站问题。
这些不是未来场景，是正在发生的现实。OpenAI经济研究团队发布的"Work at the Frontier"系列首篇报告，通过分析80多万条美国ChatGPT用户的工作消息，发现了一个被称为"任务跨界"（task crossover）的现象：43.5%的职业特定AI使用，涉及的是另一个职业的任务。
谁在跨界：客服、设计师、HR冲在最前面
"任务跨界"不是均匀分布的。剔除通用任务（写作、总结、排程等几乎所有岗位都会做的事）后，有些岗位的跨界率高得惊人：
客服人员77%的职业特定消息涉及其他岗位任务——他们在用AI做营销文案、数据分析甚至基础财务计算。设计师75%在跨界——除了设计本身，他们还在做工程排错、营销素材制作、财务预算。HR69%在跨界，法务56%，营销53%。
这些岗位的共同特点是什么？它们都是信息密集型、需要跨部门协作的岗位。以前这些岗位的人遇到"不是我的活"时，需要发消息、开会、等其他部门的人来处理。现在他们直接把任务丢给AI，自己就搞定了。
哪些任务被"借"得最多？财务计算和技术故障排查出现在所有其他七个职业群体最常见的三大跨界任务中。营销类任务的传播范围也很广——制作营销素材出现在五个其他职业群体中。
任务的"进出口"：谁在借，谁在被借
跨界不是单向的。报告区分了两种角色：任务"进口方"（从其他岗位借任务来做的人）和任务"出口方"（自己的任务被其他岗位的人做了的人）。
设计师是典型的纯进口方：35.2%的设计师消息涉及其他岗位的任务，但设计任务只占其他岗位人员消息的1.7%。设计师在大量借用别人的活，但自己的活很少被别人借走——可能因为设计需要审美判断力，AI在这方面还不够强。
工程是典型的纯出口方：只有18.5%的工程师消息涉及其他岗位，但工程任务占其他岗位人员消息的7.4%。工程师自己不怎么跨界，但全公司都在借用工程能力——AI让非技术人员也能排查技术问题、写简单脚本、理解技术架构。
营销是双向冠军：24.3%的营销消息跨界（进口），同时营销任务占其他岗位消息的8.9%（出口，全样本最高）。营销人既在做别人的活，别人也在做营销的活。
还有一个有趣的发现：小企业的跨界比例高于大企业。2-5人工作区的用户跨界率18.9%，而100人以上工作区只有16.3%。这很合理——大公司有专门的团队和流程，员工遇到问题可以交给专家；小企业里谁碰到问题谁解决，AI给了他们这个能力。AI作为"通才工具"，在专业资源稀缺的地方最有价值。
先改变"谁做什么"，再改变岗位本身
这份报告最重要的洞察不是"AI会取代工作"那种陈词滥调，而是一个更微妙的过程：AI对劳动市场的影响，首先体现为岗位任务的重组，而不是岗位的消失。
职位描述还没改、岗位名称还没变、组织架构图还没画——但活已经不是原来那么分了。一个客服在做数据分析，一个销售在做文案，一个市场在修网站。等HR和管理层意识到这些变化的时候，新的分工模式已经在日常使用中成型了。AI使用数据就像一个早期预警系统，让我们能在传统劳动市场统计数据捕捉到变化之前，就看到工作正在怎么变。
这对企业和个人都意味着什么？对企业来说，岗位边界正在变得模糊，招聘时过度强调"专业对口"可能会错过最能利用AI的跨界人才。对个人来说，你的竞争力不再仅仅是你本职岗位的技能有多深，而是你能用AI跨界做多少"不是你的活"。
未来最有价值的人，可能不是最专业的人，而是最会用AI"越界"的人。
明天见。
A small-business owner independently uses AI to draft marketing copy, review contracts, and do basic financial analysis. A salesperson uses AI to analyze customer datasets — previously a data analyst's job. A marketer troubleshoots website issues without waiting for a developer.
These aren't future scenarios — they're happening now. OpenAI Economic Research's inaugural "Work at the Frontier" report, analyzing over 800,000 work messages from U.S. ChatGPT users, documents a phenomenon called "task crossover": 43.5% of occupation-specific AI use involves tasks associated with a different occupation.
Who's Crossing Over: Customer Service, Designers, HR Lead the Pack
Task crossover isn't evenly distributed. After filtering out generic tasks (writing, summarizing, scheduling — things almost every role does), some job categories show staggeringly high crossover rates:
Customer experience workers: 77% of occupation-specific messages involve tasks from other jobs — they're using AI for marketing copy, data analysis, even basic financial calculations. Designers: 75% cross over — beyond design itself, they're doing engineering troubleshooting, marketing asset creation, and budget planning. HR: 69% cross over. Legal: 56%. Marketing: 53%.
What do these roles have in common? They're information-intensive roles requiring cross-functional collaboration. Previously, when people in these roles encountered "not my job" tasks, they needed to send messages, schedule meetings, wait for other departments. Now they just hand the task to AI and handle it themselves.
Which tasks get "borrowed" most? Financial calculations and tech troubleshooting appear among the top three crossover tasks across all seven other occupation groups. Marketing tasks also travel widely — creating marketing materials appears across five other occupational groups.
Task Imports and Exports: Who Borrows, Who Gets Borrowed From
Crossover isn't one-directional. The report distinguishes between two roles: task "importers" (people borrowing tasks from other jobs) and task "exporters" (people whose own tasks get done by others).
Designers are pure importers: 35.2% of designer messages involve other-occupation tasks, but design tasks account for only 1.7% of other workers' messages. Designers are borrowing heavily from other roles, but their own work rarely gets borrowed — likely because design requires aesthetic judgment where AI isn't yet strong enough.
Engineering is a pure exporter: only 18.5% of engineering messages involve other fields, but engineering tasks appear in 7.4% of non-engineers' messages. Engineers don't cross over much themselves, but the whole company is borrowing engineering capability — AI lets non-technical people troubleshoot tech issues, write simple scripts, understand technical architecture.
Marketing is the two-way champion: 24.3% of marketing messages cross over (importing), while marketing tasks make up 8.9% of other workers' messages (exporting — highest in the sample). Marketers are doing other people's jobs, and other people are doing marketing jobs.
Another interesting finding: small businesses show higher crossover than large ones. Users in 2-5 seat workspaces show 18.9% crossover, compared to just 16.3% in 100+ seat workspaces. This makes sense — large companies have specialized teams and processes; when an employee encounters a problem, they hand it to a specialist. In small businesses, whoever encounters the problem solves it, and AI gives them that ability. AI as a "generalist tool" is most valuable where specialist resources are scarce.
First Changing 'Who Does What,' Then Changing Jobs Themselves
The report's most important insight isn't the tired cliché that "AI will replace jobs" — it's a more nuanced process: AI's impact on labor markets is manifesting first as task reorganization, not job disappearance.
Job descriptions haven't been rewritten, titles haven't changed, org charts haven't been redrawn — but the work is already being divided differently. A customer service rep is doing data analysis, a salesperson is writing copy, a marketer is fixing websites. By the time HR and management notice these changes, the new division of labor has already taken shape in daily practice. AI usage data is like an early warning system, letting us see how work is changing before traditional labor market statistics can catch up.
What does this mean for companies and individuals? For companies, job boundaries are blurring, and over-emphasizing "specialist match" in hiring may miss the crossover talent who can leverage AI most effectively. For individuals, your competitiveness is no longer just about how deep your primary skills are — it's about how many "not your job" tasks you can handle with AI.
The most valuable people in the future might not be the most specialized — they might be the ones best at using AI to cross boundaries.
See you tomorrow.
AI对工作的改变，不是先炒掉谁，而是先让每个人都能做更多"不是自己的活"。岗位边界模糊之后，组织图才会跟着变。
—— Dawn Vision编辑部
AI doesn't change work by firing people first — it starts by letting everyone do more 'not their job' tasks. After job boundaries blur, the org chart follows.
—— The Dawn Vision Editorial Desk
OpenAI · 工作研究 · task crossover · 任务跨界 · 岗位重组 · AI就业 · ChatGPT · 劳动市场 · 43.5% · 小企业
OpenAI · work research · task crossover · job reorganization · AI employment · ChatGPT · labor market · 43.5% · small business
Sources · 信源 Sources
本文基于 Dawn Vision 认知引擎处理的 6 个源信号生成，经编辑部人工审核。素材来源：OpenAI官方博客、今日头条全球AI简报。
This article was generated by the Dawn Vision cognitive engine processing 6 source signals, with human editorial review. Sources: OpenAI Blog, Jinri Toutiao Global AI Briefing.
