推动工业与软件自动化在工作场景中的全面协同

– Executive Summary– An Automation Experiment – The Current State of Workforce Automation– Understanding the Automation Landscape– Case Study: Amazon Vulcan – Recommendations for Business Leaders– Case Study: Academy of Advanced Manufacturing– Augmenting Automation with Skilled Technical Talent Introduction Executive Summary •TheManpowerGroup Employment Outlook Surveyresearchfound that in mid-2025,61% of global companies areincreasing their investment in task and process automation.This figure is even higher for organizations that are highlyimpacted by workforce aging (71%) and uncertainty aroundinternational trade (68%). •ManpowerGroup's 2025 Global Talent Barometerfound thatemployees aremore concerned about economic instability(34%) than being replaced by AI or other automationtechnology (19%). •The two major types of workforce automation includeindustrial automation and software automation.Industrialautomation leverages physical robots to perform manual laborin manufacturing and other production settings. Softwareautomation enhances HR and employee productivity andefficiency by delegating simpler digital tasks to smart machinepartners. •While many forms of automation are already well-established –such as business process and robotic processautomation –intelligent automation tools such as agenticAI are likely to replace generative AI as the next workforceMVP.Where they once operated with only a limited store ofinformation and required heavy oversight, today’s agentic AIbots can collaborate both with other technology systems andhuman workers with looser supervision. •Recommendations for leaders implementing new industrialor software automation includedesigning human-friendlysystems, connecting systems from end-to-end, creating aninfrastructure for human/machine team assembly, facilitatingsupervised independence, upskilling and empowering humancolleagues, and measuring success beyond raw productivity. An Automation Experiment A group of Carnegie Mellon University researchers recently set up a fake softwarecompany, TheAgentCompany,1to test how well automation in the form of AI-based agentswould fare in a real-world business scenario with no human supervision. The simulation included designated bots, such as a Chief Technology Officer and aChief HR Officer, to govern tasks such as online research, code writing and spreadsheetdevelopment.TheAgentCompany had several information and communication tools atits disposal, from an ultra-specific employee handbook to a live chat function, and the AIagents were designed to engage easily with one another. There was every reason to believe the experiment would be a success. After all, agenticAI technologies developed by Anthropic, OpenAI and others could do so much more thanexecute human instructions. TheAgentCompany bots allegedly had the ability to actindependently and make novel decisions in unfamiliar environments. However,TheAgentCompany failed.There wasn’t a single category of work in which theAI agents accomplished the majority of the tasks required, and the researchers quicklylearned that the AI agents weren’t as good at simple tasks as they thought. They oftenmisinterpreted feedback, were sidelined by minor changes or abnormalities, and lackedcommon sense when faced with a problem. The conclusion?Automation can be extremelyuseful, but it isn’t the answer to everything. "The first rule of any technology used in a business is that automation applied toan efficient operation will magnify the efficiency. The second is that automationapplied to an inefficient operation will magnify the inefficiency."– Bill Gates We actually learned this lesson a long time ago. When automation first came to the sceneduring the late 18th and early 19th centuries, it was defined as the use of mostly self-driven equipment in a system of manufacturing or other production process. At this time inhistory, known as the Industrial Revolution, many despaired that automation would meanthe end of human work. Today, we see a resurgence of that concern with the addition of intelligent, AI-basedtechnologies to traditional automation strategies, such as production forecastingand predictive maintenance. Many workers fear that mass layoffs and record-highunemployment will result from integrating machine participation in traditionally human-driven work processes. But just like in the Industrial Revolution,this has yet to happen. According toManpowerGroup’s latest Employment OutlookSurvey2of more than 40,000 employersaround the world, the story is more complex.Most employers expect to keep theirheadcount flat or hire in Q3 2025. Only 16%said they anticipate decreases.Sectorssuch as IT and Communication Servicesare poised to increase hiring volumes, andwhile industries such as financial services,real estate, healthcare and life sciences aredialing back hiring,there are multiple factorsinvolved besides automation. Most employers expectto keep their headcount flator hire in Q3 2025 Whil