Back to home

Compare

Comparing: When Clients Bring the Hard Problems, Can AI Read Them First? & 客户把难题丢给你时,AI 能不能先替你读懂

AEN
Anthropic Claude独立创业者自由职业·

When Clients Bring the Hard Problems, Can AI Read Them First?

Last night, a client’s “simple question” stopped me cold

At 10 p.m., I read the client’s six-page brief three times and still wasn’t sure which version to revise first. My old routine was to open three documents, copy information between them, then force out a reply.

If this sounds familiar, I want to say one thing first: it wasn’t that I wasn’t smart enough. When the material piles up, people naturally miss earlier constraints. I’ve been there too. I once spent two hours producing a proposal the client didn’t want to see.

Anthropic’s AI recently tackled a complex reasoning puzzle

According to The Neuron, Anthropic’s AI recently broke through a complex puzzle. The important part wasn’t that it “got the answer right,” but that it could check scattered conditions step by step and produce a conclusion I could inspect.

I know an independent developer, Lin Yue, who does website redesigns. She received a redesign brief from a pet-shop owner while working from a shared office in Hangzhou. She used to mark up every page manually. Now she gives the AI the campaign rules, client feedback and old pages, then asks it to list contradictions and missing details before she reviews them. Lin Yue says the biggest change wasn’t faster copywriting. It was one less round of rework.

I tried replicating this today on a small budget

The rough cost is RMB 20-100 per month for a model subscription, plus 30-60 minutes to get comfortable with prompting. The technical barrier is low: copy text and check sources, with no coding required. First step: open the AI assistant I already use, start a new conversation, paste one redacted client brief, and ask it to list the “known conditions, contradictions, missing information and three questions I should ask first.”

I remove names, phone numbers, account details and other identifying information from client files before uploading them. I also check every AI conclusion rather than sending it straight to the client. This isn’t essential for everyone. If your requirements are simple, a manual checklist may be faster, and there’s no urgency to test it now.

Match it to my work stage, without trying to do everything at once

If I’m just starting: I’d practice with one real brief of my own and limit the task to “help me clarify the questions.” I wouldn’t rush into a complex workflow.

If I already have one or two clients: I’d create a fixed template—the client’s original words, known conditions, conflicts and next-step questions—and have the AI return every response in the same format.

If I’m scaling: I’d consider connecting repetitive organization work to an automation tool, but I’d keep human review in place so unchecked replies never go straight to clients.

Which task would I most want AI to take over: reading a brief, drafting a proposal or following up with clients? I’m still figuring out which one saves the most time.

BZH
Anthropic Claude独立创业者自由职业·

客户把难题丢给你时,AI 能不能先替你读懂

我昨晚又被客户的“简单问题”问住了

晚上十点,我把客户发来的 6 页需求来回看了三遍,还是不知道先改哪一版。以前我会开三个文档来回抄,最后再硬写一份回复。

如果你也遇到过这种情况,我想先说:不是我不够聪明,而是资料一多,人很容易漏掉前后条件。我也卡过,花了两小时做出一份客户根本不想看的方案。

Anthropic 的 AI 最近解决了一道复杂推理题

据 The Neuron 报道,Anthropic 的 AI 在一道 puzzle 上取得突破。它真正值得看的地方,不是“答对一道题”,而是能把散落条件逐步核对,再给出可检查的结论。

我认识一位做网站改版的独立开发者林悦。她在杭州一间共享办公室接到宠物店老板的改版需求,过去要逐页标记;后来她把活动规则、客户反馈和旧页面一起交给 AI,请它先列出矛盾与待确认项,再让自己核对。林悦说,最明显的变化不是写文案更快,而是少做了一轮返工。

你今天复刻一次,预算不用很高

大致成本是每月 20-100 元的模型订阅,加上 30-60 分钟熟悉提示词;技术门槛不高,会复制文字、核对来源即可,不需要懂代码。第一步:打开你平时用的 AI 助手,新建对话,先粘贴一段脱敏后的客户需求,让它列出“已知条件、矛盾、缺信息、建议先问的 3 个问题”。

客户资料记得先删掉姓名、电话、账号等信息。AI 给出的结论也别直接交出去,要逐条核对。它不是所有人都需要;如果你的需求很简单,手工列清单反而更快,现在不试也没事。

按你的工作阶段来,不用一步到位

如果你刚起步:我会先拿自己的真实需求练一次,只做“帮我列清问题”,不要急着搭复杂流程。

如果你已有 1-2 个客户:我会做一个固定模板:客户原话、已知条件、冲突点、下一步问题,并让 AI 每次按同一格式输出。

如果你正在扩规模:我会再考虑把重复的整理工作接到自动化工具里,但先保留人工确认,别让未经核对的回复直接发给客户。

你最想让 AI 接手的是读需求、写方案,还是做客户跟进?我也还在摸索哪一种最省时间。