（先搓了一版来发。估计会继续更新、优化的，后面再发V2 V3版本，更新内容会写到文尾。排版估计也会改一下。另外可能也会录视频，加点思维导图来阐述）
春秋带剑 | 公众号 chunqiudaijian
其实FDE的风口已经吹好几个月了，在我们AI这行已经应该算明日黄花了，但是我看最近似乎也没有推出什么新的职业概念，而且FDE这种职业分类也确实有其道理，所以估计还能再招一段时间（至少几个月）的人。
FDE（Forward Deployed Engineer，现场部署工程师）的核心概念是做2B端的AI服务，让企业能真正用AI的能力提高生产力或者节约成本，按项目成果交付，而不是按软件功能点、工时来交付。用AI行业常见的黑话可以说成“让AI能力在企业真正落地，实现降本增效”。
这一行还算新，各行业还都没有出现成熟、被普遍接受的智能体解决方案，所以还有做头，巨头和新人都在从0开始思考要如何把AI用起来。所以还需要人的加入。在卖方端求职，各种B2B/SaaS企业现在都在追AI的风口，都在招人；也有很多独立的新团队在做这件事。而在买方端，很多传统企业自己也想赶AI这趟潮流，实现降本增效，或者至少声称自己愿意干，显得很时髦的样子，所以可以成为FDE的目标客户，让FDE这个新概念还能再支撑一段时间。
现在买方和卖方市场都还愿意吃FDE这个概念，所以感觉做AI这行（或者想转来这行）的同志们都可以了解一下这个职业概念，尤其是已经在做企业AI应用相关的，求职、接活、汇报吹水的时候都可以参考。
也学学人家老外造概念、吹风口的水平……
本文就会简单从FDE的定义、需要学什么、如何找工作/单干接活/自己把公司的搞起来 这三个方面来分享速赶FDE风口的指南。
由于咱这行风口变得很快，所以有些内容可能会更新，我也会更新本篇文章。目前写的内容都是基于当前AI行业的情况来写的。
1. FDE是什么
FDE≈AI项目售前+部署+开发+售后+客服
FDE应该又懂AI技术，又懂行业知识，深入了解甲方的实际业务之后，能够真正将AI能力加入到甲方公司的业务链上，并且让甲方公司的人用起来，真正实现降本增效，用起来舒服，最后把尾款给你结了：既把事干好，也把人给说服。
比如甲方如果自己干的话，可能只能想到“我们要做个智能客服，接入DeepSeek来回答用户问题”，然后就让新员工在客服系统里接入了自己部署的DeepSeek，但是首先嫌慢，然后又嫌弃大模型不懂自己的业务需求，听说有知识库，就加上了知识库，但是发现大模型也查不到知识库内容。客服觉得AI从说话口气开始就不对劲，都不想用AI生成的内容，而且说不定一个不小心，AI又给客户说出了不存在的优惠政策……
于是公司觉得，要不我们请一帮专业人士，来帮助我们真正把AI用起来，实现我们对AI能力的幻想吧！
于是就有了AI落地企业的服务公司，而这些AI公司中进入甲方企业、实地了解真实需求、提出和验证解决方案的工程师，就是FDE。
FDE接到需求之后，首先要了解甲方真正的期望：包括上上下下不同人的小心思（干2B这行都这样），比如大老板是又想少招几个人工客服省钱（但如果是大公司或者立善良人设的老板，这可不能明说；而实际上也少招不了多少人，大概率只能提升效率和质量，还是需要很多活人来兜底，包括审核、沟通、宣传和展示、背锅担责这些环节，毕竟AI没有签名章，追责的时候又不能抓AI进去。比如FDE目前这么多莫名其妙的扯皮工作就不能全部交给AI），同时又希望AI客服能迅速、没有个人情绪影响地高效回复用户（这个可以明说），但是现在直接接入DeepSeek回答的效果太差了，所以想问问有没有高人（FDE，一个公众号文章上会经常出现的时髦名词）能解决问题；二老板需要显得自己在这个项目里很积极，节省的成本和提高的用户满意度都应该是自己的功劳；三老板想让自己人搞AI应用，不想让外包赚这个钱；客服部负责人觉得大事不妙，如果把自己的经验都让AI学去了，TA不就要失业了吗！新招进来的客服人员可能会很乐意直接用AI来回复，而有些老客服可能会觉得AI很烦，因为用户问AI问不明白会越问越恼火，再转给人工客服的时候就更生气了。
FDE需要在这样复杂的环境里了解公司的实际现状，在此基础上帮助公司真正用AI实现降本增效。在这个过程中可能还需要帮助公司完成基础的无纸化、数字化、云化建设，在海量的RPA中找到了少量的AI。
AI回复还会出现违规内容和虚假宣传，你就解决去吧。
目前在市场上FDE一般还是跟乙方驻场人员差不多，只是由于AI应用具有很多不可预知性，所以很难像传统软件外包开发一样甲方直接提出明确需求就招标，而最好是能进厂直接调研，跟业务人员一起工作，需要FDE在真实的工作流程基础上结合经验判断，AI能如何提升效率、节约成本。
所以FDE行业非常需要“同行业合作经验”，很看重口碑，来减少用户的怀疑，毕竟看起来调研、纯问的环节太多了，敲代码、做功能的时间反而短，就不显得很厉害。当然2B这行本来也都需要同行业经验，或者说其实大多数行业的客户都希望找有经验的服务提供商，高中生的家长找家教都希望找能带出过状元的老师。而且现在AI编程更擅长写代码而非探索需求，所以现在的软件开发行业也都会把更多时间花在明确、分析需求上而不是写代码上，就很容易让客户觉得看起来好像一直在耍嘴皮子而不是干活，客户容易觉得说话是不需要钱的！（当然这早已是纯咨询行业容易面临的困境）
此外由于合规流程的问题，很多大公司也很难签下这么一个自由度很高的合同，会在合同里提前规定好很多功能之类的内容。但实际业务可能并不是一纸合同能说全的，乙方FDE还是得展开实地调查，才能真正确定到底应该做什么功能：比如甲方可能本来想做一个AI客服系统，但是FDE可能评估认为用户与客服对话过程中会产生对售后、维修部门的需求，所以还需要将信息直接传达到他们那里去，在人工客服时代都是靠人工客服打电话完成的，那FDE还得设计信息传达流程。这就是合同没有包括的新内容了。而且AI技术本身就发展很快，随时都可能提出有效的新技术，合同上可能规划了当时觉得有效的方法，但是很快就会更新、更有效的解决方案，能更好地实现用户的需求，但是跟合同上规定的不一样——那咋搞呢？那就很难搞了。
这些都是AI本身、纯搞技术很难解决的问题，都需要沟通、协调。因此对FDE来说，软技能非常重要。
当然，FDE仍然应该对技术有整体的了解，有企业应用AI的实践经验，能敏锐找到有效的解决方案，可以迅速开发原型、演示demo。如果团队成员少，可能FDE自己还要完成开发工作。
所以FDE被期望是一个全能的工作，这也是我在本节提出的：FDE≈AI项目售前+部署+开发+售后+客服
2. 想赶FDE的风口可以准备些什么
1. 技术栈
- 软技能：适当润色，专业画饼。对非专业人士讲技术，讲得有逼格，更要展示成效
- 向上汇报能力：展示对比变化，面向汇报开发
- 对外宣发能力
- 与甲方员工沟通、了解其业务流程和需求的能力
- 与自己团队成员沟通协作、共同维系项目运营的能力
- 谈判和判断能力：需要判断一家公司的活适不适合接（自己的能力能不能hold住，相关对接人有没有责任心、能不能频繁交流（包括差旅问题）、有没有真正的权力来改变整个生产流程（阎王易见，小鬼难缠），老板有没有耐心，甲方有没有能力和意愿结尾款）（但是呢，有时活没干好，却能结项拿到钱，也是一种本事）
- 产品设计：平衡实用与逼格
想让实际员工用起来就不能过度设计，不要改变他们的工作。但想要统一、严格，就要大改，要AI-native，而且有的时候领导会觉得可以大改，统一范式，虽然实际业务人员觉得每个地区都不一样、没法统一
- 行业知识：比技术能力更重要，专业人士转AI应用其实比AI从业者进入专业更容易
当然，也都能干，其实我本人确实是干AI的转来干业务的。真的是干起来就老实了，我干着干着学会了越来越多的业务知识，虽然还是对业务知识一窍不通，但是，干就完了……
（这里其实再补充一条，就是根据我个人的经验判断，这里面每一条要求都是锦上添花，没有都可以干，注意每一条没有都可以干，先入行）
- AI应用开发能力，尤其是与行业知识关联的
- AI原理
- 相关的安全、法律意识
- 性格积极、主动，有好奇心，愿意积极探索用户需求，不能让用户的付费意愿在眼前溜走。在实操上，能够针对用户需求快速实现demo、快速改进优化
- 不眷恋机会成本，敢于推翻以前辛辛苦苦的工作，按照真正的需求来改进甚至重做；但同时又能够从过去的产品中学习，复用成熟经验，加快优化迭代进度、也减少用户替换新工具的学习成本
- 有攻击性，敢于挑战成规，对现状不容易满足；然而又要保持足够的谦逊，尊敬相关的管理和业务人员和现有流程，只是积极提出改进建议
再强调一下：想用AI优化整个业务流程，这项工作最好从业者能具有上述的性格和技术特点，但是没有也能干，任何工作都这样，只要入行了就能干，不会没关系，学就行了。
2. 能力证明
- 学历
- 准备项目
如何证明你拥有知识，最好的方式就是将知识应用于实践：
- 真实商业2B项目
- 学术项目
- 个人AI项目
既要有技术指标，更要有技术指标之后带来的业务、管理上真正提升的效果，最好既有量化数据（从4小时降到8分钟），又有客户满意度这种抽象数据（客户选择续约）。
（删除了考证相关内容）
3. 怎么干
1. 怎么入行
- 在招聘网站（包括在内容平台发的招聘广告）直接找工作
- 接私活
对小公司来说差旅费非常贵，干起来就老实了
- 在自己的公司里应用上：当然这建议你就是老板或者CTO，小兵就不要搞事了
工作流程：
- 最好先有自己对行业的理解和通用服务产品形态
- 吹水、宣传（加入社群、参加活动、打广告、喝酒私聊）、谈判/招标→签单
真正的FDE合同里会规定AI真正用起来的业务、经济层面的目标指标，包括定量和定性的，这也是想让AI在企业中真正用起来必须得给相关负责人的压力。如果只给技术指标，很容易只对技术指标优化，而不是做好使用体验；如果只写了功能，很可能只是做了一些僵硬的功能，而不能让员工能真正舒服用起来。
还有些公司连AI想应用于什么业务场景都没想好，只想要个哆啦A梦来解决一切，那不可能，对这种用户建议先评估一下交易成功率，有赚头的可以先陪他们一起挖掘挖掘可能有效的业务场景、AI提效的方向。如果老板还在纯做大梦，那我唯一的建议就是：RUN！！！
Sierra的智能体工程负责人默勒的标准是：只接两类问题——真的难的，和真的有业务影响的。
还有，不要接无论如何都不提供真实数据、只让你出产品的。这种情况只建议一种乙方干，就是研究型机构出论文，毕竟是学术研究嘛。不过我觉得本文的大多数读者不是这种情况，但凡是真想应用于业务实践的，都肯定得有真实的业务数据才行，真实业务数据里会看出更多真正的矛盾和规律。当然甲方对数据保密性有一定要求是合理的，比如演示、宣传和大规模开发的时候可以用脱敏、模拟数据，但是不能完全不给所有人看真实数据啊！
项目周期最好也不要整太长了。
（还有，要注意确定甲方的财务情况。你干起来就知道了）
- 通过访谈、查资料、参与真实工作（Palantir称之为影子工作法）等等方式，研究明白用户的真实业务流程（可能需要纯内网甚至无网工作，那就要本地化、容器化部署），以及其中可以优化的方向，数据来源和类型（必要的时候可能得从数据电子化、打通各个部门数据源、数据质量校验和旧数据清洗（黑话“数据治理”）开始干起），各层级各岗位上员工们的真实预期目标。
- 在此基础上完成自己的业务目标——整出AI优化工作流程的智能体demo，催用户测试，获得反馈（用户在看到产品之前，可能不知道他们自己想要什么。在看到产品之后，可能就知道自己不想要什么了）
干的时候要服从甲方的需求，但不要屈从于甲方的奇思妙想，要保留合理需求，拒绝离谱意见。要在合理的范围内让用户真正获得自己想要的东西，而不是跟着用户临时的想法，比如说看到什么新技术就想上什么新技术，领导容易关注哪些部分，就只优化哪些部分。长期来说也是为他们好。
- 改进循环
与团队协作优化更新整体产品，继续测试，继续改。
甲方的短期目标和组织结构都随时可能变。
尽量减少后期维护成本，比如培训用户、更新需求什么的。理论上最好甲方公司以后要更新什么需求，自己问问AI就能改；甲方公司进入新员工想要使用工具，都不用学，好用到打开就会用。
部署的AI工具记得要建立持续、可更新的评估方案，以及时发现和解决数据、模型漂移的问题。
注意安全问题，包括数据隐私脱敏、AI输出合规性、正确性的问题。
- 结项后与整个产品团队对自己的工作流程进行总结和复盘，思考能从中学到什么，将这些经验和知识继续应用到长期的服务工作中。最好能建立出可以灵活调整的统一产品甚至平台，实现规范化++，定制化--。根据特定甲方只需要做简单适配，而不用每一家单独开发一整套工具（做SaaS服务如果给每家都做定制化的话，跟外包有什么区别，但是现在很多FDE真的在给每家都做定制化，没有自己的产品，这跟外包有什么区别，做AI的外包就叫FDE是吧……）
↑但需要注意，这是项目结束后需要研究的事情，在项目里干的时候，最重要的还是服务好当前的客户，让他们自己把AI用起来，而不是考虑你自己以后的发展和平台功能的更新迭代。
4. 其它吐槽
- FDE这个词语的来源是Palantir公司，Palantir公司也被认为是FDE这行的祖师爷（至少是之一），人家一开始干就能接到美军和空客的单，要我说虽然他们提出新概念、解决甲方问题的能力确实很强，但是销售的能力更是不知道哪里去了……
这行就是99%的销售加上1%的技术，但是这1%的技术不可或缺。
- Palantir甚至还能免费开干，有效才收钱，这雄厚的现金流，凡是见过2B创业公司的人都应该知道这含金量。
- 出差，累得要死，到底谁在喜欢出差，我又要去出差了……
5. 参考资料
（删除了相关内容）
6. 关于作者
（删除了相关内容）
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
(Stamp a copy first. It is expected that it will continue to be updated and optimized, followed by a V2 V3 version, which will be updated at the end of the text. The layout is expected to change. And maybe a video and a mind map.
WeChat Official Account cunqiudaijian
In fact, the FDE has been blowing for months, and it should have been in our AI business, but it seems to me that no new career concept has been introduced recently, and that the FDE job classification does make sense, so it may be possible to recruit people for another period (at least a few months).
The core concept of FDE (Forward Deployed Engineer, Field Deployment Engineer) is to provide 2B-end AI services that enable enterprises to truly use AI capabilities to increase productivity or save costs and deliver on project results rather than on software functionality points and hours. It can be said in the language common to the AI industry that “let AI's capacity actually land on the business and achieve efficiency gains”.
This business is new, and there are no mature, generally accepted AI agent solutions in any industry, so there is a head start, and giants and newcomers are beginning to think about how to use AI. So people are needed to join. At the seller ' s end of the job search, B2B/SaaS enterprises are now pursuing the AI front and recruiting; there are also many new independent teams doing this. On the buyer side, many traditional firms themselves want to catch up with AI, achieve efficiency gains, or at least claim that they are willing to do it, which is fashionable, so they can be FDE's target customers, so that the new FDE concept can be sustained for some time.
Now that both the buyer and the seller market are willing to eat the FDE concept, it is felt that the comrades who are doing the AI (or who want to move to the industry) can learn about the concept of the profession, especially when it comes to doing business AI applications, looking for jobs, taking jobs, reporting water.
And learn from the old foreign concept, the level of the blowjob...
This paper will simply share guidance on fast-tracking of the FDE from the definition of the FDE, what needs to be learned, how to find a job/work alone/get a company together.
Because we've become fast, there are elements that may be updated, and I will also update this article. This is based on the current situation in the AI industry.
1. What is a FDE?
Before the FDE≈AI project + deployment + development + after sale + customer service
The FDE should be able to learn about AI technology and industry, and to learn more about the actual business of A, to truly add AI capacity to the business chain of A, and to put it to use by A, to actually achieve efficiency gains, to use it comfortably, and finally to close the down payment to you: to do good and convince.
For example, if the A side did it on its own, it might have thought, “We're going to do a smart customer service, get access to DeepSeek to answer user questions” and then get new employees to access their own DeepSeek in the passenger service system, but first it's too slow and then it's too late to understand LLM's business needs, and it's added to the knowledge base, but it turns out that LLM can't find the knowledge base. It felt like AI was wrong from the beginning, didn't want to use what it generated, and maybe, by accident, AI told the client about the non-existent preferential policy...
So the company felt that if we didn't hire a bunch of professionals to help us really use AI and realize our vision of AI's power!
This led to the establishment of a service company for the AI landed enterprise, and the engineers of those companies who entered the A company, learned about real needs in the field, proposed and validated solutions were FDE.
The FDE needs to understand the real expectations of the A: including the caution of different people up and down (both in the 2B line), such as the big boss who wants to hire a few less manual customers to save money (but it cannot be said if he is a big company or a good man’s boss); and the fact that he can't afford to hire more people than a few, probably to improve efficiency and quality, or whether he needs a lot of living people to do the bottom line, including vetting, communication, publicity, and accountability. After all, AI has no signature and can’t catch AI when it comes to accountability. For example, the FDE can't give all of this inexplicable work to AI, and hopes that the AI client service will respond to the user quickly and efficiently without personal emotional influence (this can be said), but the direct access to DeepSeek's answer is too bad, so I'm asking if there's a problem that can be solved (FDE, a fashionable term that often appears in a WeChat Official Account article); the second boss needs to appear active in the project, saves costs and improves user satisfaction; the third boss wants his own people to use the AI, and does not want to outsource the money; the head of the client service feels bad, and if he leaves his experience to AI, the TA will lose his job? New recruits may be happy to use AI to respond directly, and some of the old ones may find AI annoying, because users are angry when they ask AI if they don't understand the problem and turn it over to the manual.
FDE needs to learn about the actual state of affairs of the company in such a complex environment as to help the company achieve real efficiency gains with AI. In this process, it may also be necessary to help companies complete basic paperless, digitized and clouded construction, with a small number of AIs found in big RPAs.
The AI.'s response is also accompanied by irregularities and false propaganda, so you can settle it.
At present, FDEs in the market are generally similar to those on site B, except that because of the unpredictable nature of AI applications, it is difficult for A to bid directly on the basis of a clear demand from the traditional software outsourcing development, preferably by going to the factory and working with business personnel, requiring FDEs to combine experience on the basis of a genuine workflow process and how AI can improve efficiency and save costs.
The FDE industry is therefore in great need of “collaborative experience” to reduce user suspicion. After all, there seems to be too much research, too many questions and too little time to knock on codes and do functions. Of course, the 2B business would have required industry experience, or in fact most industry clients would have preferred experienced service providers, and the parents of high school students would have preferred teachers who could have brought them out. And now that AI programming is better at writing codes than exploring needs, it's easy for the software development industry to spend more time on identifying, analysing needs rather than writing codes, making it look like it's been playing around instead of working, making it easy for clients to feel that they don't need money to talk!
In addition, because of the compliance process, it is difficult for many large companies to enter into such a highly liberal contract, which would provide for many functions in advance. However, the actual business may not be complete by a paper contract, and B-FE would still have to carry out a field survey to determine what exactly should be done: for example, A might have wanted to do an AI passenger service system, but the FDE might have assessed that there would be a demand for after-sales, maintenance services during the user-client dialogue, so that the information would have to be delivered directly to them, which was done by means of a manual passenger service. This is the new element not included in the contract. Moreover, AI technology itself is fast-growing and it is possible at any time to come up with effective new technologies, and the contract may have planned the methods that were felt to be effective at the time, but it will soon be updated and more effective solutions that better meet users' needs, but unlike the contract provision — how is that possible? That's hard.
These are problems that AI itself, purely technical, is difficult to solve and require communication and coordination. Soft skills are therefore very important for the FDE.
Of course, the FDE should still have a holistic understanding of technology, practical experience in the application of AI by enterprises, be keen to find effective solutions, and develop prototypes and demos quickly. If the team is small, the FDE itself may have to do the development work.
So the FDE is expected to be an all-powerful job, and that's what I'm proposing in this section: before the FDE≈AI project is sold, deployment, development, after sale, customer service.
2. What can FDE be prepared for?
1. Technology warehouse
- Soft skills: proper colour, professional drawing cake. It's very technical, and it's more effective.
- Reporting capability upwards: to demonstrate comparative changes, for reporting development
- Public outreach.
- The ability to communicate with and understand business processes and needs of staff on the ground
- The ability to communicate with members of their own teams and to jointly sustain project operations
- The ability to negotiate and judge: a company needs to be judged to be unsuited (whether it can hold itself, whether the counterpart is responsible, whether it can communicate frequently (including on travel), whether it has real power to change the entire production process (in the case of children and children), whether the boss is patient, whether the A side has the capacity and willingness to end it) (but sometimes it is a matter of doing good work, but getting paid for it)
- Product design: balancing practical and coercive
It is impossible to over-design the actual staff and not change their jobs. But if it's to be uniform and rigorous, it's to be radical, it's to be AI-native, and sometimes the leadership thinks it's possible to make a big difference, it's to be uniform, although the operational staff feel that each region is different and it's not possible to harmonize.
- Industry knowledge: more important than technical skills, it is easier for professionals to switch to AI than for AI practitioners.
Of course, I can do it. I actually do AI myself. I've learned more and more about business, but I'm done...
Here's another one, which I've learned from my own experience, every one of these requirements is a flower.
- AI application development capacity, especially in relation to industry knowledge
- AI principles.
- Relevant security, legal awareness
- Active, proactive, curious and willing to actively explore the needs of users, who should not be allowed to slip away from paid. Physically, fast-track demo and fast-improvement improvements tailored to user needs
- Unprejudiced opportunity costs, dared to reverse hard work, improve or even re-do it according to real needs; at the same time, however, they can learn from past products, revert to mature experiences, accelerate the optimization of iterative progress and reduce the learning costs of users to replace new tools
- Be aggressive, dare to challenge the rules and are not easy to satisfy; however, be modest enough to respect the relevant managerial and operational staff and existing processes, only to actively propose improvements
It should also be emphasized that, in order to optimize the entire business process with AI, it is best for practitioners to have the characteristics and technical characteristics described above, but not for any job, so that they can do as long as they enter the profession, it doesn't matter, they can learn.
2. Certification of competence
- Education
- Prepare the project.
The best way to prove your knowledge is to apply it in practice:
- Real Business 2B
- Academic projects
- Personal AI projects
Both technical indicators and, more importantly, real operational and managerial improvements that follow them, preferably data from quantization (down from four hours to eight minutes) and abstract data of client satisfaction (client choice renewal).
(Deleting relevant test)
3. How
1. What? Enter
- Direct job placement on recruitment websites, including job advertisements on content platforms
- Private work.
It's very expensive for a small company.
- Use it in your own company: Of course, you're the boss or the CTO.
workflow:
- Better have an understanding of the industry and a general service product.
- Water blowing, advocacy (to join the community, to participate in events, to advertise, to drink, to talk privately), negotiation/bid signing
The real FDE contract provides for targets, both quantitative and qualitative, at the operational and economic levels for which AI is actually used, and it is also necessary to put pressure on the responsible persons to use AI for real use in the enterprise. If only technical indicators were given, it would be easy to optimize them rather than experience their use; if only functional, it would probably have been done with some rigidity rather than making them truly comfortable for staff.
There are also companies that do not even think about what kind of business scenario AI is going to use, but rather want a Doraemon to solve everything. That is impossible. It is not possible for such users to suggest an assessment of the trade success rate, and there is something to do with them in order to find out where it might work. If the boss is still dreaming, my only advice is: Run!
The standard for the AI agent project manager in Sierra is that only two types of questions are taken up — really difficult, and really business-related.
And don't take no real data, no matter what. This situation suggests only one type of " B ", i.e. research institutions, after all academic research. However, I do not think this is the case for most readers of this paper, but any person who really wants to apply it to business practice must have real business data, which will reveal more real contradictions and patterns. There is, of course, a reasonable requirement for data confidentiality on the A side, such as de-sensitization and simulation of data for demonstration, publicity and large-scale development, but it cannot be completely denied to all.
It would be preferable not to have a long project cycle.
(In addition, care should be taken to establish the financial position of party A. You'll see.
- Studying the true business processes of users (which may require netting or even netless work, which may require localization, containerization and deployment) and the direction in which they can be optimized, by means of interviews, data searches, participation in real work (known as the shadow work method by Palantir), etc., and by which data sources and types (which may, if necessary, begin with the electronicization of data, access to various sectoral data sources, data quality verification and cleansing of old data (blackword “data governance”) are the real expectations of staff at all levels.
- Fulfilling its own business objectives on this basis - complete the AAI agentdemo that optimizes the workflow process, prompts users to test and obtain feedback (users may not know what they want before they see the product). When you see the product, you know what you don't want.
In doing so, they must be subject to the needs of the A, but not to the tacts of the A, but to the need to preserve the legitimate needs and to reject the absurd ideas. To give users real access to what they want, within a reasonable range, rather than following the user's ad hoc thinking, for example, of seeing what new technologies are, thinking of what new technologies are, and of what is easy for leadership to focus on, what is best. It's good for them in the long run.
- Improve the cycle.
Work with the team to optimize the updating of the overall product, continue to be tested and continue to change.
The short-term objectives and organizational structure of both sides are subject to change.
Minimize post-production maintenance costs, such as training users, updating needs and so forth. Theoretically, it would be preferable for A company to update its needs later and ask AI itself what it wants to change; and for A company to enter new employees to use tools without learning to use them, it would be useful to open them.
The AI tools deployed are mindful of the need to establish a continuous and updated assessment programme to identify and address data and model drift issues in a timely manner.
Pay attention to security issues, including data privacy desensitization, AI output compliance, correctness.
- After the completion of the project, carry out with the entire product team a review and recapitulation of its workflow programme to reflect on what can be learned and continue to apply this experience and knowledge to long-term service work. It would be desirable to create uniform products or even platforms that can be adapted flexibly, to regularize and customize. It's easy to fit according to a given A, not to develop a single set of tools for each family. (What difference does SaaS make if it customizes every household?) But now that many FDEs are customizing every family without their own product, what difference does it make to outsourcing?
It needs to be noted, however, that this is something to be studied at the end of the project, and it is more important to serve current clients when they do the project, so that they can use AI on their own, rather than considering your own future development and updating of platform functions.
4. Other complaints
- The term FDE comes from Palantir, which is also considered to be one of the great masters (at least one of them) of the FDE, and who can get a list from the American army and air travelers at the beginning, and I would say that, while they do have a strong ability to come up with new concepts and solve the problem on the one hand, they do not know where to sell.
This is 99 per cent sales plus 1 per cent technology, but this 1 per cent technology is indispensable.
- Palantir can even work free of charge to collect money, which is a huge cash flow that anyone who has seen a 2B start-up company should know about.
- On a business trip, I'm so tired, I'm going on a business trip...
References
(Deleted)
6. On authors
(Deleted)
Auto-fetched and translated by AI Atlas. All rights belong to the original author.