读完这篇才懂：AI重构企业的底层逻辑，早就藏在「算法图谱」里了

导读

This article completely changed my understanding of how AI will transform enterprises. The core point is that any company, no matter how complex its business, can be dismantled into a connected "graph of algorithms" composed of step-by-step processes. AI can accurately identify redundant, inefficient links in this graph, and even completely replace some processes, which will bring subversive changes to enterprise operation efficiency. This is not a far-off fantasy, but a trend that consulting firms are already promoting. Whether you are a business operator or an ordinary employee, this perspective can help you see the direction of change in advance and prepare for the upcoming AI era.

这篇文章彻底刷新了我对AI改造企业逻辑的认知，核心观点是：任何企业不管业务多复杂，都可以拆解成由一个个流程步骤组成、互相关联的「算法图谱」，AI能精准识别这个图谱里冗余、低效的环节，甚至完全替代部分流程，会给企业运行效率带来颠覆性的改变。这不是遥远的空想，而是咨询公司已经在推进的趋势，不管你是企业经营者还是普通员工，这个视角都能帮你提前看清变化方向，为即将到来的AI时代做好准备。

I think the reason so many people don't understand how big AI is going to be is that they don't understand that everything is an algorithm.

我发现很多人之所以没意识到AI的影响力会有多大，本质上是没搞懂一个道理：所有事务本质上都是一套算法。

Specifically, they don't realize that companies are just a collection of algorithms.

更具体来说，他们没意识到，企业本质上就是各类算法的集合体。

Let me give you an example. Suppose there's a company called Memories that focuses on photo restoration and customization services. Its business process is very simple: users upload or send physical photos to the company first, the team is responsible for high-precision scanning, uses traditional photography technology and Photoshop to repair damaged parts, then makes stylized adjustments according to user needs such as retro style, cinematic style, family style, adds matching text descriptions, and finally provides users with digital downloads or physical prints.

举个例子，假设有家叫Memories的公司主打照片修复定制服务，业务流程很简单：用户先上传或者邮寄实体照片给公司，团队负责高精度扫描，用传统摄影技术和Photoshop修复破损部分，再根据用户需求做复古、电影感、家庭风这类风格化调整，加上匹配的文字说明，最后给用户提供数字版下载或者实体打印件。

Anyone who knows a little about computer logic can see that this whole set of processes is a clear step-by-step flow, which is the algorithm we usually talk about. And each step can be further split into smaller algorithms: for example, the upload link can be split into user upload, format verification, quality detection and other sub-steps, and each sub-step can continue to be subdivided, and the bottom layer is all independent algorithm units.

稍微懂点计算机逻辑的人都能看出来，这整套流程就是清晰的步骤流，也就是我们常说的算法。而且每个步骤还可以再拆成更小的算法：比如上传环节可以拆成用户上传、格式校验、质量检测等子步骤，每个子步骤还能继续细分，往下全是独立的算法单元。

Of course, the business of an enterprise is not only the core business flow, but also includes company registration, employee recruitment, tax payment, infrastructure procurement, marketing, customer service and other links. When you split all these links layer by layer and sort out the connection relationship between each module, you will get a complete "algorithm graph" of the enterprise. The word "graph" is used here because it can clearly show the interaction logic between different modules, such as which modules the output of a certain link will be sent to, and which modules' input it needs to receive.

当然企业的业务不止核心业务流，还包含公司注册、员工招聘、缴税、基建采购、市场营销、客户服务等各类环节。当你把所有这些环节层层拆分，梳理清楚每个模块之间的连接关系，就会得到一个完整的企业「算法图谱」。这里用「图谱」这个词，是因为它能清晰展现不同模块之间的交互逻辑，比如某个环节的输出会发送给哪些模块，又需要接收哪些模块的输入。

This kind of process dismantling itself is already very valuable. If you could sort out a complete algorithm graph for an enterprise in 2022, that enterprise would have gained a huge competitive advantage, because you can clearly see which links are redundant and which processes can be optimized. But when AI is added to this system, the effect will be completely subversive: AI is not only good at performing independent discrete tasks, but also good at sorting out the connection between various links. Once the enterprise is presented in the form of an algorithm graph, almost all workflow links are waiting for AI to optimize or even completely replace.

这种流程拆解本身就已经很有价值，如果2022年就能给某家企业梳理出完整的算法图谱，那家企业已经拿到了巨大的竞争优势，因为你能清晰看到哪些环节是冗余的、哪些流程可以优化。但当AI加入这个体系后，效果会完全是颠覆性的：AI既擅长执行独立的离散任务，又擅长梳理各个环节之间的联系，一旦企业以算法图谱的形式呈现，几乎所有工作流环节都等着AI去优化、甚至完全替代。

Actually, most enterprises are far less efficient than everyone thinks, with a lot of redundant processes and waste of resources. Consulting firms such as Accenture, KPMG and McKinsey have already targeted this pain point, and they will soon promote AI-based transformation solutions to enterprise management: first, fully survey all workflows of the enterprise through automated tools and manual interviews, distinguish which links are already automated and which are still manually performed, then find out wasteful processes, redundant teams and ineffective departments, and finally merge or even eliminate unnecessary links. The final result of this transformation must be that the enterprise becomes more compact, the operation cost is lower, and the number of employees required to maintain operation will be greatly reduced.

其实现在大部分企业的运行效率远没有大家想的高，有大量流程冗余和资源浪费。埃森哲、毕马威、麦肯锡这类咨询公司已经盯上了这个痛点，他们很快就会给企业管理层推广基于AI的改造方案：先通过自动化工具和人工访谈，全面调研企业所有工作流，区分哪些环节已经自动化、哪些还是人工执行，接着找出存在浪费的流程、冗余的团队、效能不佳的部门，最后合并甚至直接砍掉不必要的环节。这套改造的最终结果，一定是企业变得更紧凑，运营成本更低，维持运转需要的员工数量会大幅减少。

And this kind of optimization will not be a one-time action. After AI is implanted into the enterprise system, it will continuously analyze all links of the enterprise: for the marketing department, it will ask how many people are involved in a certain link, why the idea generation needs to wait for a whole month, why it takes so long from idea to official launch of the campaign, how much time is spent on copywriting, and who is responsible for sending a large number of promotion emails. The same logic applies to customer service, human resources, recruitment and other departments, and almost no department can avoid being analyzed and optimized by AI.

而且这种优化不会是一次性动作，AI植入企业系统后，会持续不断地分析企业的所有环节：针对市场部，会追问某个环节涉及多少人力，为什么创意生成要等一整个月，从想法到正式上线活动为什么要花这么久，文案撰写要耗费多少时间，大量推广邮件是谁负责发送。同样的逻辑放到客服、人力、招聘等部门都成立，几乎没有部门能逃过AI的分析和优化。

Many people will say that my company's business is very special, the process is more complex, and it is far more difficult than the simple photo processing case mentioned above. But in fact, this only means that the algorithm graph of the enterprise is larger. It may be very difficult for humans to sort out such a complex system at one time, but it is not a problem for AI at all. No matter how special the product and business model are, the operation of the enterprise is a step-by-step pipeline, and AI is best at sorting out the logical relationship between these steps and presenting it clearly to decision-makers.

很多人会说，我所在的公司业务很特殊，流程更复杂，远不止上面提到的简单照片处理案例。但实际上这只意味着企业的算法图谱更庞大，人类想要一次性梳理清楚这么复杂的系统可能很困难，但对AI来说完全不是问题。不管产品和商业模式多特殊，企业运行本质上都是一套分步骤的流水线，AI最擅长的就是梳理清楚这些步骤之间的逻辑关系，清晰地呈现给决策者。

I don't write these contents to scare everyone, or to make everyone resist AI. On the contrary, this change has both advantages and disadvantages: on the one hand, the efficiency of enterprises will be greatly improved, the overall productivity will rise, and the threshold for entrepreneurship will be much lower than before, and many people who could not start a business in the past can also try to build their own business; on the other hand, a large number of jobs that are easy to automate will disappear, which will bring a lot of employment pressure. Instead of being anxious and depressed, it is better to recognize this trend in advance, think about what parts of your enterprise's algorithm graph are easy to be replaced by AI, and make preparations before the change comes.

我写这些内容不是为了吓唬大家，也不是要让大家抵制AI。相反，这个变化有弊也有利：一方面企业效率会大幅提升，整体生产力上升，创业门槛会比之前低很多，过去没条件创业的人也可以尝试搭建自己的业务；另一方面大量容易被自动化的岗位会消失，会带来不少就业压力。与其焦虑沮丧，不如提前认清这个趋势，想清楚你所在企业的算法图谱里，哪些部分容易被AI替代，在变化来临之前做好准备。

来源：https://danielmiessler.com/blog/companies-graph-of-algorithms