7月11日，F(xiàn)orrester副總裁、首席分析師Craig Le Clair受邀出席2020世界人工智能大會(huì)云端峰會(huì)RPA+AI分論壇，分享RPA在全球的發(fā)展現(xiàn)狀。

Craig帶來(lái)了RPA最新趨勢(shì)報(bào)告《后疫情生產(chǎn)力時(shí)代，智能自動(dòng)化打造以人為本的企業(yè)》，并為眾多云端觀眾解讀了自動(dòng)化的關(guān)鍵趨勢(shì)。他的分享為后疫情時(shí)代，企業(yè)智能自動(dòng)化相關(guān)建設(shè)指明了思路。

01.

3 Key Trends in Intelligent Automation to move forward postpandemics

后疫情時(shí)代智能自動(dòng)化的三個(gè)主要趨勢(shì)

Human & Machine Cooperation

人機(jī)協(xié)同

As machines become moreintelligent, they replace more and more manual processes, this is generallyreferred to as the human in the loop issue. Benefit from AI technology, peoplemove towards the less deterministic type of processes, where we use machinelearning and other related AI technologies to make decisions in the process.This really alters a lot of things around the process that humans used to do.

隨著機(jī)器智能水平的不斷提高，它們?nèi)〈嗽絹?lái)越多的人工流程，這通常被稱為人機(jī)協(xié)作-“Human-Machine Cooperation”。得益于人工智能技術(shù)，人們逐漸向規(guī)則更復(fù)雜、確定性更低的流程自動(dòng)化領(lǐng)域邁進(jìn)。在這個(gè)過(guò)程中，我們使用機(jī)器學(xué)習(xí)等人工智能技術(shù)來(lái)進(jìn)行數(shù)據(jù)處理、輔助決策。這極大改變了人類很多傳統(tǒng)的工作場(chǎng)景。

Intelligent Document Extraction

智能文本提取

It's using machinelearning, which is a subset of AI, to go into documents, forms, emails, what'sgenerally referred to as unstructured information. Based on that, you have aclean set of data that's high quality, then you can apply analytics. For example,you can look at errors that might have occurred in someone filling out of form,you can check transactions that may not be correct based on analyzing the data,and you can dig customer sentiment in the text.

文本挖掘?qū)儆跈C(jī)器學(xué)習(xí)的范疇，也是人工智能的一個(gè)子集。所處理的文本通常是文檔、表格、票據(jù)、電子郵件等非結(jié)構(gòu)化的信息。基于智能文本抽取技術(shù)，可以獲取更高質(zhì)量的數(shù)據(jù)，用于后續(xù)的數(shù)據(jù)分析工作。例如，查看用戶填寫表單時(shí)可能出現(xiàn)的錯(cuò)誤，檢查可能不正確的交易，挖掘文本中的客戶情感。

Although the core ofthis area is still natural language processing, what’s interesting now is usingmachine learning to train a model to understand the meaning behind the text.There are also a lot of advancements in computer vision and spatialunderstanding. So you can understand the forms and images that are on adocument to be able to get a better context. So it's much more of aninsights-driven value now where in the past it's been more about just takingcost out of the process.

雖然這一領(lǐng)域的核心仍然是自然語(yǔ)言處理，但用機(jī)器學(xué)習(xí)的訓(xùn)練模型，來(lái)理解文本背后的含義，也是近期熱門的領(lǐng)域。隨著計(jì)算機(jī)視覺(jué)、文本分類等方面技術(shù)的逐漸成熟。機(jī)器可以理解文檔上的表單和圖像，以及上下文的邏輯。在AI技術(shù)的賦能下，智能自動(dòng)化蘊(yùn)含了具有洞察力的價(jià)值。而不僅是過(guò)去的降低運(yùn)營(yíng)成本。

Automation Strike Teams

自動(dòng)化突擊隊(duì)

It is very important fora company to take a broader view of intelligent automation. There are a numberof reasons for this.

One is that some ofthese automation introduces some new issues that need to be considered from agovernance standpoint. Second, robots are using human credentials and areoperating on very trusted and secure application areas. So you need to haveguard rails or control around the use of those credentials. Also, Intelligentautomation has involved a number of software technologies. You need teamsinternally to explain the use of the different automation technologies to thebusiness than to apply it in the right way.

對(duì)于企業(yè)的自動(dòng)化建設(shè)，用更廣闊的視野進(jìn)行整體規(guī)劃十分重要。這里有多方面的原因：

一是一些自動(dòng)化場(chǎng)景，引入了需要從組織治理角度考慮的新問(wèn)題。其次，機(jī)器人通常會(huì)使用員工的密碼憑證，進(jìn)行一些生產(chǎn)系統(tǒng)的日常操作。因此，需要關(guān)注密碼憑證的安全管控。此外，智能自動(dòng)化涉及許多技術(shù)的應(yīng)用，需要團(tuán)隊(duì)在企業(yè)內(nèi)部解釋不同自動(dòng)化技術(shù)的特點(diǎn)，以便用正確的方式應(yīng)用和落地。

02.

The Pandemic Will Create A Surge In Digital Transformation

疫情給數(shù)字化轉(zhuǎn)型帶來(lái)的機(jī)遇

Despite trends inmobility, in social media, and the digital disruption of companies like theUbers of the world and the great Chinese companies that have come into thesharing market, the progress on digital transformation has been quite slow.Unfortunately, modernization is hard. Digital transformation is hard.

盡管在移動(dòng)互聯(lián)網(wǎng)、社交媒體等領(lǐng)域，Uber等全球公司，以及美團(tuán)、滴滴等中國(guó)互聯(lián)網(wǎng)企業(yè)，在數(shù)字化驅(qū)動(dòng)業(yè)務(wù)模式創(chuàng)新等方面的勢(shì)頭表現(xiàn)良好。但傳統(tǒng)企業(yè)數(shù)字化轉(zhuǎn)型的進(jìn)展相當(dāng)緩慢。這讓我們不得不面對(duì)一個(gè)現(xiàn)實(shí)：現(xiàn)代化是艱難的，數(shù)字化轉(zhuǎn)型不易。

When we came to thispandemic point in January, February, March. Suddenly, we had to transformdigitally really fast. Everyone had to work from home. We had to conduct remotebusiness in new ways. Then we encountered some issues with our supply chainsand so forth. So under tremendous pressure, we had to innovate. The best out ofthe worst, with a surge in digital transformation, we've developed more digitalmuscles in the past two months than we've had in the last five years.

在2020年1月到3月的全球疫情高峰期，突然之間，企業(yè)快速進(jìn)行數(shù)字化轉(zhuǎn)型。每個(gè)人都不得不在家工作，企業(yè)必須以新的方式開展遠(yuǎn)程業(yè)務(wù)。隨之而來(lái)的，是在供應(yīng)鏈等領(lǐng)域衍生出的一系列問(wèn)題。在巨大的壓力下，企業(yè)不得不進(jìn)行業(yè)務(wù)創(chuàng)新。不幸中的萬(wàn)幸，疫情也從側(cè)面推動(dòng)了一部分企業(yè)的數(shù)字化轉(zhuǎn)型進(jìn)程，部分企業(yè)在過(guò)去的兩個(gè)月里構(gòu)建了比過(guò)去五年更多的數(shù)字化能力。

That's what that spikeyou see in this graph. Now the challenge for companies is to take thistransformation that's occurred under stress and to see what should remain as wemove back to a more normal work environment. This will affect theinstitutionalize progress and transformation that we made in the past fewmonths. 

上圖中閃電標(biāo)志描述的數(shù)字化轉(zhuǎn)型激增的區(qū)域?，F(xiàn)在，公司面臨的挑戰(zhàn)是如何在壓力下進(jìn)行這種轉(zhuǎn)變，并在疫情結(jié)束恢復(fù)更正常工作環(huán)境時(shí)，繼續(xù)保持?jǐn)?shù)字化轉(zhuǎn)型的勢(shì)頭。這將影響我們?cè)谶^(guò)去幾個(gè)月中取得的數(shù)字化轉(zhuǎn)型成果。

03.

Changes in the Intelligent Automation Roadmap

后疫情時(shí)代的智能自動(dòng)化路線圖

Forester proposed apost-pandemic roadmap that gives you a way to prioritize your intelligentautomation projects. This is what we're seeing our clients do, who are thecompanies and governments that we interact with consistently. If you wereworking on a large AI project, transformational project, you might push thataside. That might drop into the losing momentum zone. Because this recessionthat we're going in is going to be, by most estimates, long and painful.

Forrester提出了后疫情時(shí)代智能自動(dòng)化路線圖，為企業(yè)提供了一種確定智能自動(dòng)化項(xiàng)目?jī)?yōu)先級(jí)的方法。基于Forrester對(duì)所服務(wù)企業(yè)、政府客戶的調(diào)研和溝通。一些正在進(jìn)行中的，大型人工智能、數(shù)字化轉(zhuǎn)型項(xiàng)目，進(jìn)度會(huì)受到影響甚至停滯，會(huì)掉進(jìn)左下象限的動(dòng)量損失區(qū)。因?yàn)楦鶕?jù)大多數(shù)人的估計(jì)，我們正在經(jīng)歷的這場(chǎng)疫情引發(fā)的經(jīng)濟(jì)衰退，將是漫長(zhǎng)而痛苦的。

It's going to focus likeall previous recessions on cost reduction and cost take out. So in theacceleration zone to the upper right, RPA task automation becomes veryimportant because it has a very visible ROIfor cost reduction.  What we havementioned above, text analytics, which allows you to extract a lot of hoursshuffling paper and dealing with forms and finding errors, sentiment, fraud,and other issues. Also, remote work is a practical technology that's availabletoday.

就像以前所有的經(jīng)濟(jì)衰退一樣，企業(yè)將更專注降低成本和成本轉(zhuǎn)移。因此，在象限右上方的加速區(qū)中，RPA自動(dòng)化變得非常重要，因?yàn)樗诮档统杀痉矫婢哂蟹浅Ｃ黠@的ROI。

上面提到的智能文本提取也在這個(gè)區(qū)域，它能幫助員工節(jié)省大量的時(shí)間，來(lái)整理文檔、處理表單、發(fā)現(xiàn)錯(cuò)誤、客戶投訴、欺詐風(fēng)險(xiǎn)等問(wèn)題。此外，一些視頻會(huì)議、遠(yuǎn)程協(xié)作類的辦公產(chǎn)品和工具，也在疫情期間發(fā)揮了巨大的作用。

No one was prepared forwhat we call a systemic shock, as the pandemic. The next systemic shock mightbe climate change. So there's a greater awareness of companies in the sort ofgovernance spectrum to be prepared for these kinds of systemic risks that mayoccur. So resiliency has become a top priority item for automation. And thismight mean diversity in your supply chain so that you can quickly get new bidsout. You have the ability to have a more agile approach to sourcing variousgoods and services.

極少數(shù)人能夠準(zhǔn)備好，應(yīng)對(duì)我們所說(shuō)的系統(tǒng)性全球風(fēng)險(xiǎn)。下一個(gè)系統(tǒng)性全球風(fēng)險(xiǎn)可能是氣候變暖。我們看到，更多的公司開始關(guān)注治理領(lǐng)域的問(wèn)題，增強(qiáng)了風(fēng)險(xiǎn)管理意識(shí)，為可能發(fā)生的此類系統(tǒng)性風(fēng)險(xiǎn)做好準(zhǔn)備。因此，系統(tǒng)彈性、業(yè)務(wù)韌性等可持續(xù)發(fā)展能力，已成為數(shù)字化轉(zhuǎn)型中的重中之重。例如，提升供應(yīng)鏈的多樣性，以確保可用的材料采購(gòu)，和商品交付能力。

04.

The Impact of IntelligentAutomation on Different Workers

智能自動(dòng)化對(duì)不同類型工作者的影響

The degree of influenceof intelligent automation on different types of work is different. For example,cubicle workers are employees that may work in a contact center taking phonecalls or work in a back office doing finance and accounting. They are put inthe same category because the skills they have are very similar. So the effectof automation on their work will be similar.

智能自動(dòng)化對(duì)不同類型工作者的影響程度是不同的。例如，呼叫中心的客服員工，企業(yè)后臺(tái)從事財(cái)務(wù)工作的員工，由于他們擁有相似的工作模式，標(biāo)準(zhǔn)的工作流程，因此自動(dòng)化對(duì)他們工作的影響也是相似的。

On the other hand, forthe knowledge workers who might be a legal strategist. They are makingconnections across a wide range of complex information and data. So automationmay not be used in this field for a long time. But we also saw some innovativescenes, such as digital assistance will help in the cognitive search in thehealth industry.

另一方面，對(duì)于法律從業(yè)人員等創(chuàng)造工作者，他們通常會(huì)處理復(fù)雜的信息，并在海量數(shù)據(jù)之間建立聯(lián)系。由于工作的創(chuàng)造性水平、流程不固化等特殊性，自動(dòng)化可能在很長(zhǎng)一段時(shí)間內(nèi)，都不會(huì)應(yīng)用在這個(gè)領(lǐng)域。但我們也看到了一些創(chuàng)新的場(chǎng)景，比如數(shù)字員工助手輔助醫(yī)療行業(yè)的從業(yè)人員，進(jìn)行認(rèn)知搜索和知識(shí)發(fā)現(xiàn)。

RPA + AI most Affected Knowledge andAdministrative Workers

RPA+AI對(duì)特定職能的知識(shí)工作者和行政人員影響最深

Currently, the hottestapplication area which makes the real effect is in the operating field.Intelligent automation is very suitable for the function-specific knowledgeworkers, coordinators, administrative workers. This is the target area whereyou can do a lot of work with machines. A lot of automation technology, AItechnology is changing the traditional process in this area.

目前，RPA真正發(fā)揮作用的應(yīng)用領(lǐng)域主要集中在運(yùn)營(yíng)領(lǐng)域。智能自動(dòng)化非常適合于特定職能的知識(shí)工作者、協(xié)調(diào)員、行政工作者。這些流程標(biāo)準(zhǔn)、操作規(guī)范的場(chǎng)景是RPA主要的應(yīng)用領(lǐng)域。很多自動(dòng)化技術(shù)、AI技術(shù)正在改變企業(yè)的傳統(tǒng)流程。

05.

Five Levels of Human & Machine Cooperation

人機(jī)協(xié)作的五個(gè)層次

There are differentlevels of human in the loop related to intelligent automation. Level five iswhere you're using the most advanced AI and the machine, such as a self-drivingcar. To the contrary, level zero is where human is doing everything.

在與智能自動(dòng)化相關(guān)人機(jī)協(xié)作中，會(huì)根據(jù)技術(shù)復(fù)雜度和自動(dòng)化模式的不同分為5個(gè)層次。第五層是應(yīng)用最先進(jìn)的人工智能技術(shù)實(shí)現(xiàn)機(jī)器的自主運(yùn)動(dòng)，比如自動(dòng)駕駛汽車。相反，第零層是描述人類日常工作中沒(méi)有自動(dòng)化驅(qū)動(dòng)的場(chǎng)景。

In the middle of them, we have automationtechnology developed to different stages. Level one is the area of workflow,where you're using software to design a process, generally moving from task totask, which is a deterministic pattern.

在這中間，隨著自動(dòng)化技術(shù)發(fā)展的不同階段，又有進(jìn)一步的細(xì)分。第一層是工作流領(lǐng)域，我們使用BPM軟件來(lái)設(shè)計(jì)流程，連接不同的工作節(jié)點(diǎn)，處理一些確定性流程的自動(dòng)化。

Level 2 Human Drives Machine Actions

第二層：人類驅(qū)動(dòng)機(jī)器完成任務(wù)

Level two is what a lot of RPA is doing rightnow, where we have built some digital workers or digital assistance. And thehuman has some level of interaction with the robot. The human in a contactcenter can tell the robot to update all these addresses. So there's a level ofvery positive automation and productivity there.

第二層是很多RPA正在做的事情，企業(yè)已經(jīng)構(gòu)建了一些數(shù)字員工或數(shù)字助理。人類與機(jī)器人之間有某種程度的互動(dòng)。比如在客服中心，人可以使用機(jī)器人批量更新客戶的地址。以此來(lái)降低員工的信息系統(tǒng)負(fù)擔(dān)，解放生產(chǎn)力，投入更有價(jià)值的工作。

Level 3 Human Completes Task with the Help ofMachine

第三層：人類在機(jī)器的幫助下完成任務(wù)

Level 3 is where aseries of AI technologies are combined with RPA to give people stronger dataprocessing capabilities. These AI components, such as NLP usually use machinelearning to provide a more flexible extraction of data. In the old days, youhad to know exactly based on a template where a particular field was and takethe data out perfectly. But now you can really understand the content in thedocument, in general with machine learning, where the data is, what it lookslike, and use the training of the data to be more and more precise on yourextraction.

第三層是一系列AI技術(shù)與RPA相結(jié)合，賦予人們更強(qiáng)的數(shù)據(jù)處理能力。一些AI組件，例如NLP，通常使用機(jī)器學(xué)習(xí)來(lái)提供更靈活的數(shù)據(jù)提取。在過(guò)去，我們必須根據(jù)模板準(zhǔn)確地定位字段的位置，隨后才能取出數(shù)據(jù)。但現(xiàn)在，機(jī)器學(xué)習(xí)可以理解文檔中的內(nèi)容，識(shí)別文字、數(shù)據(jù)在哪個(gè)區(qū)域，并利用訓(xùn)練機(jī)制，使機(jī)器的識(shí)別和提取越來(lái)越精確。

We can foresee whatlevels four and five do where the AI is making all the decisions. So you mayhave explained ability issues, transparency issues. No one knows how thedecision was made. You need a perfect algorithm. So that the car doesn't driveinto a stone wall. You need perfect data. So you're not doing a biasedassessment. Those issues in level four and five become less of a concern inlevels three and level two.

我們可以預(yù)見在第四層和第五層，人工智能將更多的參與決策工作。我們也會(huì)遇到一些智能自動(dòng)化能力的黑盒問(wèn)題，沒(méi)有人知道決策是如何做出的。我們需要更完美的算法和智能技術(shù)，以保證無(wú)人駕駛汽車不會(huì)因錯(cuò)誤識(shí)別引發(fā)的事故。第四層和第五層中的所面臨問(wèn)題，在第二層和第三級(jí)中，并不是那么引起關(guān)切。

06.

Hurdle to Scale for Enterprise-Level RPA

企業(yè)級(jí)RPA規(guī)?；卣沟淖枇?/strong>