什么是认知自动化，为什么重要？ (What is cognitive automation and why does it matter?)
Cognitive automation is a summarizing term for the application of machine learning technologies to automation in order to take over tasks which would otherwise require manual labor to be accomplished.
As a result, businesses can streamline their workflows beyond the scope of current automation technologies, and accomplish a next level of operational efficiency. According to a McKinsey study, businesses that adopted cognitive automation tools were able to:
因此，企业可以简化其工作流程，使其超出当前自动化技术的范围，并实现更高水平的运营效率。 根据麦肯锡的一项研究 ，采用认知自动化工具的企业能够：
- Automate approximately 50–70 percent of tasks. 使大约50％至70％的任务自动化。
- Cut down data processing time by 50 to 60 percent. 将数据处理时间减少50％到60％。
- Decrease annual labour expenditure by 20–30% 将年度人工支出减少20–30％
- Achieve triple-digit ROI 实现三位数的投资回报率
Given its potential, companies are starting to embrace this new technology in their processes. According to a 2019 global business survey by Statista, around 39 percent of respondents confirmed that they have already integrated cognitive automation at a functional level in their businesses. Also, 32 percent of respondents said they will be implementing it in some form by the end of 2020.
Thus, cognitive automation represents a leap forward in the evolutionary chain of automating processes — reason enough to dive a bit deeper into cognitive automation and how it differs from traditional process automation solutions.
认知自动化与传统自动化工具 (Cognitive automation vs traditional automation tools)
With light-speed jumps in ML/AI technologies every few months, it’s quite a challenge keeping up with the tongue-twisting terminologies itself aside from understanding the depth of technologies. To make matters worse, often these technologies are buried in larger software suites, even though all or nothing may not be the most practical answer for some businesses.
ML / AI技术每隔几个月就会以飞速的速度发展，除了理解技术的深度之外，跟上令人费解的术语本身是一个很大的挑战。 更糟糕的是，尽管对于某些企业而言， 全部或全部都不是最实际的答案，但这些技术通常都埋在更大的软件套件中。
Let’s break down how cognitive automation bridges the gaps where other approaches to automation, most notably robotic process automation (RPA) and integration tools (iPaaS) fall short.
Traditional RPA is mainly limited to automating processes (which may or may not involve structured data) that need swift, repetitive actions without much contextual analysis or dealing with contingencies. In other words, the automation of business processes provided by them is mainly limited to finishing tasks within a rigid rule set. That’s why some people refer to RPA as “click bots”, although most applications nowadays go far beyond that.
传统的RPA主要限于自动化流程(可能涉及结构化数据，也可能不涉及结构化数据)，这些流程需要快速，重复的操作，而无需进行大量的上下文分析或应急处理。 换句话说，由它们提供的业务流程的自动化主要限于在严格的规则集中完成任务。 这就是为什么有些人将RPA称为“点击bot”的原因，尽管当今大多数应用程序已经远远超过了它。
The automated processes can only function effectively as long as the decisions follow a “if/then” logic without needing any human judgment in between. However, this rigidity leads RPAs failing to retrieve meaning and process forward unstructured data.
只要决策遵循“ if / then”逻辑，而在两者之间不需要任何人为判断，则自动化流程才能有效运行。 但是，这种僵化导致RPA无法检索含义并处理向前的非结构化数据。
This is not to say that processes cannot be automated using RPA, quite the contrary: There is a variety of processes which are enhanced through it, some of the most prominent applications being found e.g. in data entry, automated help desk support, and approval routings.
In contrast, cognitive automation or intelligent process automation (IPA) can accommodate both structured and unstructured data to automate more complex processes.
It infuses a cognitive ability and can accommodate the automation of business processes utilizing large volumes of text and images. Cognitive automation, therefore, marks a radical step forward compared to traditional RPA technologies that simply copy and repeat the activity originally performed by a person step-by-step.
认知自动化的变体 (Variants of cognitive automation)
As mentioned above, cognitive automation is fueled through the use of machine learning and its subfield deep learning in particular. And without making it overly technical, we find that a basic knowledge of fundamental concepts is important to understand what can be achieved through such applications.
And while we are still far away from what scientists refer to general artificial intelligence, machines are already exceptional at a few tasks. In fact, machines are able to mimic our own abilities, often outperforming those of humans when it comes to speed and precision — proper training provided:
- Computer vision (also referred to as image processing) <> See 计算机视觉(也称为图像处理)<>请参见
- Optical character recognition (OCR) <> Read 光学字符识别(OCR)<>读取
- Natural language processing (NLP) <> Comprehend 自然语言处理(NLP)<>理解
- Sound processing <> Listen 声音处理<>听
We won’t go much deeper into the technicalities of machine learning here but if you are new to the subject and want to dive into the matter, have a look at our beginner’s guide to how machines learn.
Let’s now put some of them in context and zoom in on two examples where cognitive automation has been able to redefine processes and work content!
索赔处理 (Claims processing)
Processing claims is perhaps one of the most labor-intensive tasks faced by insurance company employees and thus poses an operational burden on the company. Many of them have achieved significant optimization of this challenge by adopting cognitive automation tools.
Through cognitive automation, it is possible to automate most of the essential routine steps involved in claims processing. These tools can port over your customer data from claims forms that have already been filled into your customer database. It can also scan, digitize, and port over customer data sourced from printed claim forms which would traditionally be read and interpreted by a real person.
通过认知自动化，可以使索赔处理中涉及的大多数基本常规步骤自动化。 这些工具可以从已经填写到客户数据库中的索赔表中移植客户数据。 它还可以扫描，数字化和移植来自打印的索赔表的客户数据，这些数据通常由真实的人读取和解释。
文件处理自动化 (Document processing automation)
As outlined above, the key to the success of cognitive automation tools lays not only in executing rules of the system but in turning unstructured files — such as documents — into structured data: By extracting relevant unstructured data from them and transforming it into a standardized format, this data can finally be integrated with the remaining systems landscape.
In addition, cognitive automation tools can understand and classify different PDF documents. This allows us to automatically trigger different actions based on the type of document received.
Example of an end-to-end cognitive workflow with a PDF classifier
These are just two examples where cognitive automation brings huge benefits. You can also check out our success stories where we discuss some of our customer cases in more detail.
这只是认知自动化带来巨大收益的两个例子。 您还可以查看我们的成功案例 ，在其中我们将更详细地讨论一些客户案例。
为您的业务评估正确的认知自动化方法 (Evaluating the right approach to cognitive automation for your business)
The concept alone is good to know but as in many cases, the proof is in the pudding. The next step is, therefore, to determine the ideal cognitive automation approach and thoroughly evaluate the chosen solution.
Besides the application at hand, we found that two important dimensions lay in (1) the budget and (2) required machine learning capabilities. This article will explain to you in detail which cognitive automation solutions are available for your company and hopefully guide you to the most suitable one according to your needs.
前进的道路 (The Path Forward)
Most businesses are only scratching the surface of cognitive automation and are yet to uncover their full potential. A cognitive automation solution may just be what it takes to revitalize resources and take operational performance to the next level.
Cognitive automation may not yet have an expansive footprint due to the underlying technology being relatively new. But it is clearly gaining momentum as it continues to deliver promising results and for what it’s worth, it adds the final missing piece to the puzzle: human-like and even superhuman intelligence.