AI 办公,最大的门槛反而不在 AI,在于一个认知错觉:你以为它都知道。
现在,腾讯想用一对新的「王炸组合」,来彻底应对这件事。

今天,企业微信与 WorkBuddy 正式打通。一个是服务使用者最多的协同办公平台,一个是国内使用者最多的办公智能体,这次联合,让两者的能力有机会在同一条工作流程中发挥作用。
企业微信上新增了 WorkBuddy 的入口,WorkBuddy会以企业微信机器人的方式浮现在聊天列表中,笔者可以直接和它对话,用 AI 来辅助工作。
和过去最大的不同是,这个机器人可以取得涵盖消息、文档、会议、网盘在内的十大能力权限,可以做到不切换App,直接用到真实的工作上下文。笔者要做的,就是直接像@ 同事一样给它派活就行。
APPSO 提前拿到了企微和 WorkBuddy 这项联合发布的能力,把一些日常的办公情形重新放到这项流程里去感受,确实有了一些「真正 AI 提效」的感受。
就像笔者每个月要在企微提交的月报,笔者现在会用 AI 总结,但一个尴尬的情况是:AI 确实帮笔者省了二颇为钟,但过程中,笔者可能要花十五分钟去做各种信息的整合、搬运,给 AI 来分析。
例如,需要先把这个月自己写的稿件从后台筛选出来,记录好对应的数据,然后再丢给 AI 帮忙分析这个月的完成情况和数据表现,再把 AI 生成的内容补充到企业微信的月报提交表格里。
而现在,企业微信和 WorkBuddy 联合发布的新版本,可以说是让 AI 更靠近了笔者的工作,笔者不需要去找 AI,AI 已经在笔者的工作方式里。
月报,一句话布置给企微里的 WorkBuddy,静待它的结果就好了。
AI 开始进入工作的地方
使用起来倒没有什么学习成本。
在企业微信侧边栏找到 WorkBuddy 的图标,领取使用资格后,就能直接在对话框里交代任务。如果需要使用更完整的工作台,也可以下载安装最新的 WorkBuddy 企业版,借助企业微信身份登录。
于是,企业微信的通讯录里,就多了一位随时可以派活的 AI 同事。
笔者可以像找普通同事一样,直接和它聊天;如果要处理更复杂的任务,也可以切换到 WorkBuddy 的工作台,挑选模型、添加连接器。
为了看看两个机型到底打通到了什么程度,笔者先让 WorkBuddy 干了件小事:拉个工作群。
不用打开通讯录,一个个寻找同事,笔者直接告诉 WorkBuddy 要邀请哪些人。它很快返回了对应的同事信息,随后弹出企业微信的建群确认框,群名和三位成员都已经列好。
确认之后,一个新工作群就建好了。
给编辑部做一个能自学的选题Agent
能从企微里面的上下文整理月报,也能建群,笔者还想看看,WorkBuddy 能不能从过去的工作记录中,总结出一些可以重复使用的经验。
就拿 APPSO 编辑部来说,笔者几乎每天都会在群里讨论选题。
哪些 AI 机型值得关注,某个模型更新值不值得写,哪些选题要跟进,哪些暂时搁置,人们都在群里留下过讨论。
这些记录里,也藏着笔者平时判断选题的一些标准。
这次不用把聊天记录导出,再上传给另一个 AI。笔者直接在企微里找到 WorkBuddy,让它整理过去这段时间笔者讨论的选题情况。
笔者接着让 WorkBuddy 把这些选题标准整理成一个 Skill,以后找选题时继续使用。
有了这套标准,就可以进一步安排每天的选题工作。
「结合本群已有选题和刚才整理的取舍意见,查找最新新闻,建议最多 10 个值得跟进的选题。说明新闻事实、报道角度、来源和时间,标出哪些是群里已有的,哪些是本轮新增。」
如果配合定时任务,这些选题还可以定期整理成早报,发送到对应的企微群里。
当然,选题最终值不值得写,还得由编辑判断。但以后做类似工作时,笔者可以继续使用这套整理好的标准,不用每次都从头向 AI 解释编辑部的习惯。
在企微,和 WorkBuddy 一起策划活动
前面两个任务主要是在整理信息,接下来,笔者想让 WorkBuddy 参与一个需要多人协作的项目。
例如筹备一场读者交流会。
活动思路放在文档里,具体安排在群里讨论,项目进度则记录在表格中。不同同事还会随时补充信息、修改流程。
笔者先在企业微信里找到 WorkBuddy,让它结合已有的项目信息,整理当下的筹备情况。
WorkBuddy 依据可以获取的项目信息,生成了一份详细的进度报告。其中涵盖已经确认的变更、任务进展,以及尚未批准的建议和待决事项。
笔者还让它进一步把这些信息整理成项目看板。
不过,光能看进度还不够。就像 WorkBuddy 回复笔者消息里提到的「需要我接着做」一样,笔者继续给它安排任务:「帮我看看哪些任务有延期风险,并提醒对应负责人及时处理。」
如果只是生成一份项目报告,后续还得靠笔者自己去催进度、通知负责人。
而这次 WorkBuddy 打通了企业微信的消息和协作能力,可以在授权范围内发送消息、创建群聊,把任务结果重新交给对应的同事。
从「读上下文」,到「产出结果」,再到「结果回到工作流」,全程没有复制粘贴。AI 干完的活可以代发消息、建群、存进微盘、推送给负责人,直接进入流程的下一环。
用 AI 减少人的摩擦,也减少软件之间的摩擦
当然,真实的企业工作比笔者测试的几个情形复杂得多。
不同部门有自己的平台,不同员工具备不同的权限,涉及业务数据和客户信息时,也不能让 AI 想读什么就读什么。
这次企业微信与 WorkBuddy 的联合更新,也考虑到了企业管理的需要。
管理员可以基于企业微信现有的组织架构,统一分配 WorkBuddy 使用资格、设置访问权限、管理用量和采购 AI 算力。
员工不必各自寻找 AI 工具、注册账号,企业也可以统一决定哪些人可以使用、哪些信息允许访问。
对于已经搭建了 CRM、工单、库存等业务平台的企业,原来的业务流程也不必推倒重来。借助相应的接口、连接器和权限规格,AI 可以逐步参与已有的工作环节。
日常的项目整理和协作,直接交给企微里的 WorkBuddy 机器人就行。遇到需要结合电脑本地文件、处理更多专业任务的情况,也可以打开 WorkBuddy 独立工作台,沿着之前的工作继续往下做。
从员工顺手使用,到企业统一管理,这次打通试图应对的,实际上是同一件事:让 AI 尽可能进入人们原来就熟悉的工作环境。
企业过去积累的数字化资产,开始有机会直接参与下一次工作,AI 会沿着这些工作轨迹,把事情继续做下去,而不再只是等待有人去查阅。
这或许就是 AI 办公要有的样子,它让笔者少解释一次背景,少搬运一份资料,少切换几个软件,让原本已经运转起来的工作继续往前走。
笔者不需要把时间花在学习用 AI 上班,要让 AI 学会怎么和笔者一起工作。
本文来源:https://www.ifanr.com/1684200
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
AI office, the biggest threshold is not in AI, but in a cognitive illusion: you think it knows everything.
Now, Tencent wants to thoroughly deal with this matter with a new pair of "Wang Fang group".
Today, enterprise WeChat and WorkBuddy officially opened up. One is the collaborative office platform with the largest number of service users, and the other is the office agent with the largest number of domestic users. This combination gives the ability of the two to play a role in the same workflow.

The entrance of WorkBuddy has been added to enterprise WeChat. WorkBuddy will appear in the chat list in the form of enterprise WeChat bots. I can directly talk to it and use AI to assist my work.
The biggest difference from the past is that this robot can obtain ten capabilities including messages, documents, meetings, and network disks, and can use the real work context without switching apps. All I have to do is give it a job like a @ colleague.
APPsO got the joint release ability of Enterprise Micro and WorkBuddy in advance, put some daily office situations back into this process to feel, and did have some "real AI efficiency" feeling.

Just like the monthly report submitted by the author every month, the author will now use AI to summarize, but an embarrassing situation is that AI does save the author a lot of time, but in the process, the author may spend 15 minutes to do a variety of information integration, handling, to AI to analyze.
For example, you need to first screen the manuscript you wrote this month from the background, record the corresponding data, and then throw it to the AI to help analyze the completion and data performance of this month, and then supplement the content generated by the AI to the monthly report submission form of the enterprise WeChat.

Now, the new version jointly released by enterprise WeChat and WorkBuddy can be said to make AI closer to the author's work. The author does not need to find AI. AI is already in the author's work style.
Monthly report, a sentence is arranged for WorkBuddy in Enterprise and Wei, just wait for its results.
There is no learning cost to use.

Find the WorkBuddy icon in the WeChat sidebar of the enterprise, and once you have received the qualification, you can explain the task directly in the dialog box. If you need to use a more complete workbench, you can also download and install the latest WorkBuddy Enterprise Edition and log in as a business with WeChat.
As a result, there is an AI colleague in the address book of the enterprise WeChat who can be dispatched at any time.

I can chat with it directly like a regular colleague, or switch to WorkBuddy's workbench to select models and add connectors for more complex tasks.
In order to see how far the two models have come, I first let WorkBuddy do a small thing: pull a work group.
Instead of opening the address book and searching for colleagues one by one, I directly told WorkBuddy who to invite. It quickly returned the corresponding colleague information, and then popped up the group building confirmation box of the enterprise WeChat. The group name and three members have been listed.

Once confirmed, a new workgroup will be established.
I would like to see if WorkBuddy can summarize some of the experiences that can be reused from the past work records.
In the case of the Appso editorial department, the author discusses the topic in the group almost every day.

People have left discussions in the group about which AI models are worthy of attention, which model update values are not worth writing, which topics to follow up on, and which are temporarily on hold.
In these records, there are also some criteria for the author's usual judgment on the topic.

No need to export the chat log this time and upload it to another AI. The author found WorkBuddy directly in the micro-enterprise, and let it sort out the topics discussed by the author in the past.
The author then asked WorkBuddy to organize these topic selection criteria into a Skill, and continue to use it when looking for questions in the future.
With this set of criteria, you can further organize your daily topic selection.

"Find the latest news and suggest up to 10 questions worth following up on by combining the existing topics in this group with the trade-offs just compiled. Explain the news facts, reporting angles, sources, and timing, indicating which are already in the group and which are new in this round.”
If you cooperate with the timing task, these topics can also be regularly organized into morning reports and sent to the corresponding enterprise and micro-enterprise groups.

Of course, the final value of the topic selection is not worth writing, and it must be judged by the editor. However, when doing similar work in the future, I can continue to use this set of well-organized standards without having to explain the editorial department's habits to AI from scratch every time.
The first two tasks are mainly to collate information. Next, I want to involve WorkBuddy in a project that requires multiple people to collaborate.
For example, preparing a reader exchange meeting.

The idea of the activity is placed in the document and arranged for discussion in the group, and the progress of the project is recorded in the table. Different colleagues will also add information and modify the process at any time.
The author first found WorkBuddy in the corporate WeChat, and let it combine the existing project information to organize the current preparations.

WorkBuddy generates a detailed progress report based on available project information. It covers confirmed changes, task progress, as well as recommendations and pending issues that have not yet been approved.
I also asked it to further organize this information into project boards.
However, the light energy is not enough to see the progress. Just like WorkBuddy replied to the "need me to continue" mentioned in the author's message, the author continued to schedule tasks for it: "Help me see which tasks are at risk of being delayed, and remind the corresponding person in charge to handle them in a timely manner."

If only a project report is generated, it will be up to the author to remind the progress and notify the person in charge.
This time, WorkBuddy has opened up the messaging and collaboration capabilities of enterprise WeChat, which can send messages within the scope of authorization, create group chats, and return the task results to the corresponding colleagues.
No copy and paste from 'Read Context' to 'Output Result' to 'Result Back to Workflow'. The work done by the AI can send messages, build a group, save it on the microdisk, push it to the person in charge, and directly enter the next part of the process.

Of course, real business work is much more complicated than the several scenarios I tested.
Different departments have their own platforms, different employees have different permissions, and when it comes to business data and customer information, they can't let AI read what they want to read.

This joint update of enterprise WeChat and WorkBuddy also takes into account the needs of enterprise management.
Based on the existing organizational structure of enterprise WeChat, administrators can uniformly allocate WorkBuddy usage qualifications, set access rights, manage usage, and purchase AI computing power.
Employees don't have to search for AI tools, sign up for accounts, and businesses can decide who can use what information to access.

For enterprises that have already built business platforms such as CRM, work orders, inventory, etc., the original business process does not have to be pushed back. With the corresponding interface, connector, and permission specifications, AI can gradually participate in the existing work.
Organize and collaborate on daily projects, and hand them directly to the WorkBuddy robot in Enterprise and WeiLi. If you need to combine your computer's local files to handle more professional tasks, you can also open the WorkBuddy standalone workbench and continue with your previous work.

From the easy use of employees to the unified management of the enterprise, what we are trying to deal with this time is actually the same thing: let AI enter the work environment that people are familiar with.
The digital assets that companies have accumulated in the past are beginning to have the opportunity to directly participate in the next work, and AI will follow these work trajectories and continue to do things, rather than just waiting for someone to look up.
This may be what the AI office should look like. It allows the author to explain the background one less time, carry one less material, and switch a few less software, so that the work that has already been running continues to move forward.

I don't need to spend time learning to work with AI, I want AI to learn how to work with me.
This article is from: https://www.ifanr.com/1684200 For copyright issues, please write to: 13951431913#139.com or submit a review on the right.