销售可能是企业里最"苦哈哈"的岗位之一：白天见客户，晚上写纪要，周末做PPT，月底算预测。
OpenAI最新发布的销售团队ChatGPT Work指南，瞄准的就是销售团队这些大量的案头工作。根据OpenAI披露的试点数据，使用Agent工作流的销售团队人均产能提升了40%——不是因为AI帮他们说服了客户，而是因为AI把他们从案头工作里解放了出来，让他们有更多时间见客户。
销售的五类可自动化工作
OpenAI在指南里列举了五个典型场景，每一个都是销售每天花大量时间但不直接产生收入的工作。
第一，管线简报（Pipeline Briefs）。每周一销售经理开周会前，要花半天时间整理团队管线：哪些机会在推进、哪些停滞了、哪些需要支持。现在Agent每天自动从CRM拉取最新数据，生成一份结构化的管线简报——按阶段分类、标注风险点、给出下周重点跟进建议。销售经理周一早上打开邮箱就能看，省出半天时间用来做1:1辅导。
第二，会议准备包（Meeting Prep Packets）。见客户之前要做多少准备？看客户公司最近的新闻、看之前的沟通记录、看客户所在行业的动态、准备几个有深度的问题——以前销售至少要花30-60分钟准备一个会议，还经常准备不充分。现在Agent自动生成一份完整的会议准备包：客户公司背景、近期动态、历史沟通摘要、推荐讨论话题、可能的异议及应对方案。销售在去客户的路上花10分钟过一遍就行。
第三，预测复盘（Forecast Reviews）。月底做销售预测是每个销售经理的噩梦——要跟每个销售过数字、判断commit的单子能不能close、识别风险。Agent自动从CRM历史数据中训练预测模型，给每个机会算一个close概率，然后跟销售自己报的预测做对比，标记差异大的单子重点review。预测准确率提升了，扯皮时间减少了。
第四，客户计划（Account Plans）。大客户销售需要针对每个重点客户做季度计划：客户的业务目标是什么、我们的切入点在哪里、关键决策人是谁、竞争对手在做什么。以前一份客户计划要写好几页，花一两天；现在Agent根据已有客户数据和公开信息自动生成第一版，销售只需要补充自己的判断和策略就行。
第五，停滞交易诊断（Stalled-Deal Diagnoses）。最让销售头疼的就是"停滞的单子"——客户说"再考虑考虑"然后就没下文了。Agent可以分析整个沟通历史，识别常见的停滞信号（决策人变更、预算冻结、竞争对手介入、需求变化），给出诊断和激活建议。很多时候销售自己没注意到的问题，Agent从数据里一眼就能看出来。
销售的核心永远是人和信任，AI放大但不替代
看到这里可能有人会说："销售靠的是关系和信任，这些是AI替代不了的。"说得对，但也说得不全对。
关系和信任确实是销售的核心，AI替代不了销售跟客户吃饭喝酒打高尔夫建立起来的信任。但是——建立信任只是第一步，之后还有大量的执行工作：跟进、写方案、做报价、协调内部资源、处理异议。这些执行工作占了销售60%以上的时间，但它们不直接产生信任，只是完成交易的必要环节。
"AI不会替代Top Sales，但会用AI的普通销售，一定会替代不用AI的Top Sales。"—— 一位SaaS公司销售VP
AI的价值，是把销售从这些执行工作里解放出来，让他们把更多时间花在真正能建立信任、推进交易的事情上：跟客户深入沟通、理解客户真实需求、提供有价值的解决方案。
销售这个职业不会消失，但会分化。会用AI的销售，一个人能干以前两个人的活，而且做得更好；不会用AI的销售，会发现自己越来越难跟对手竞争。这不是AI抢饭碗，这是工具升级带来的生产力代差——就像当年会用Excel的销售替代了只会用计算器的销售一样。
明天见。
Sales might be one of the most grueling roles in any company: seeing clients during the day, writing call notes at night, making slides on weekends, crunching forecasts at month-end.
OpenAI's newly released ChatGPT Work playbook for sales teams targets exactly this mountain of desk work. According to pilot data disclosed by OpenAI, sales teams using agent workflows saw 40% higher productivity per rep — not because AI helped them persuade clients, but because AI freed them from desk work so they could spend more time in front of customers.
Five Automatable Sales Tasks
The guide lists five typical scenarios, each of them work that takes up huge amounts of sales time without directly generating revenue.
First: Pipeline briefs. Before weekly forecast calls on Monday, sales managers spend half a day compiling team pipelines: which opportunities are advancing, which are stuck, which need support. Now agents automatically pull latest CRM data daily and generate structured pipeline briefs — sorted by stage, flagged risks, recommended focus areas for the week. Sales managers open it in their inbox Monday morning, freeing half a day for 1:1 coaching.
Second: Meeting prep packets. How much prep goes into a client meeting? Reading recent client news, reviewing past communications, tracking industry developments, preparing insightful questions — previously reps spent 30-60 minutes prepping per meeting, often inadequately. Now agents auto-generate a complete prep packet: client background, recent developments, historical communication summary, recommended discussion topics, potential objections and responses. The rep skims it in 10 minutes on the way to the client.
Third: Forecast reviews. Month-end forecasting is every sales manager's nightmare — going through numbers with every rep, judging whether committed deals will actually close, identifying risks. Agents train prediction models on historical CRM data, assign a close probability to each opportunity, compare against reps' self-reported forecasts, and flag deals with large discrepancies for focused review. Forecast accuracy goes up, arguments go down.
Fourth: Account plans. Enterprise sales reps need quarterly plans for each key account: client business goals, entry points, key decision-makers, competitor activity. Previously an account plan took pages and a day or two to write; now agents auto-generate a first draft from existing client data and public information, and reps just add their own judgment and strategy.
Fifth: Stalled-deal diagnoses. Reps' biggest headache: "stalled deals" — the client says "let me think about it" and goes dark. Agents analyze the full communication history, identify common stalling signals (decision-maker changes, budget freezes, competitor entry, requirement shifts), and deliver diagnoses and reactivation recommendations. Often, problems the rep didn't notice jump out from the data.
Sales Is Always About People and Trust — AI Amplifies, Doesn't Replace
At this point someone will say: "Sales is about relationships and trust — things AI can't replace." That's true, but incomplete.
Relationships and trust are indeed the core of sales; AI can't replace the trust built over dinners, drinks, and golf rounds. But — building trust is just step one. After that comes massive execution work: follow-ups, proposal writing, quoting, coordinating internal resources, handling objections. This execution work takes 60%+ of a rep's time but doesn't directly build trust — it's just the necessary mechanics of closing a deal.
"AI won't replace top salespeople — but average salespeople who use AI will definitely replace top salespeople who don't."— A SaaS company VP of Sales
AI's value is freeing reps from that execution work so they can spend more time on things that actually build trust and advance deals: deep conversations with clients, understanding real needs, delivering valuable solutions.
The sales profession won't disappear — but it will bifurcate. Reps who use AI can do the work of two, better; reps who don't will find themselves increasingly unable to compete. This isn't AI taking jobs — it's a productivity generation gap driven by tool upgrades, just like Excel-using sales reps replaced calculator-only reps back in the day.
See you tomorrow.
"AI不会替代Top Sales，但会用AI的普通销售，一定会替代不用AI的Top Sales。"
—— 一位SaaS公司销售VP
"AI won't replace top salespeople — but average salespeople who use AI will definitely replace top salespeople who don't."
— A SaaS company VP of Sales
ChatGPT Work · sales AI · pipeline management · meeting prep · sales forecasting · stalled-deal diagnosis · 40% sales productivity gain · AI commercialization
Sources · 信源 Sources
本文基于 Dawn Vision 认知引擎处理的 9 个源信号生成，经编辑部人工审核。素材来源：OpenAI Blog。
This article was generated from 9 source signals processed by the Dawn Vision cognitive engine, with editorial review. Source: OpenAI Blog.
