如果明天一切都变了呢?一家加拿大公司正在利用人工智能帮助企业应对供应链的不确定性

qimuai 发布于 阅读:35 一手编译

如果明天一切都变了呢?一家加拿大公司正在利用人工智能帮助企业应对供应链的不确定性

内容来源:https://news.microsoft.com/source/canada/features/ai/kinaxis-supply-chain/

内容总结:

加拿大科技公司Kinaxis借助人工智能帮助企业应对供应链不确定性

全球冲突、关税调整、贸易中断和监管不确定性正迫使全球企业重新思考供应链管理方式。对许多企业而言,挑战不仅在于将货物从一地运往另一地,更在于实时理解世界某一地区的扰动如何波及工厂、供应商、库存、运输成本和客户需求。

总部位于加拿大的Kinaxis正试图帮助企业更快找到正确答案。其旗舰平台Maestro利用预测性人工智能、高级情景建模和代理式人工智能,帮助企业预测需求、测试可能情景并在条件变化时调整计划。

Kinaxis首席执行官拉扎特·高拉夫表示:“客户需要的是极强的适应性和敏捷性,这正是Maestro发挥关键作用的方式。”他指出,以霍尔木兹海峡冲突为例,客户的情景建模使用量较冲突前水平增长了120%以上。

该工具被全球企业广泛使用,涵盖汽车制造商、科技巨头以及能源和航运领军企业,用于管理复杂供应链、应对短缺并模拟关税及其他扰动的潜在影响。对Kinaxis而言,这反映了更广泛的全球趋势:曾经基于稳定假设运行的供应链,如今需要能够吸收持续变化的系统。

Maestro本质上是一个端到端供应链管理的编排平台。它汇集企业内部的销售订单、产品信息、产能和库存等数据,并结合天气、市场变化和突发新闻等外部信号。企业无需在独立系统中分别管理需求规划、库存、物流和生产,而是可以在一个由预测性人工智能和人工智能代理驱动的工具中评估这些相互关联的功能,帮助规划人员理解某一领域的变化如何影响其余业务,并更快响应扰动。

高拉夫表示:“我们带来的部分创新……就是真正将整个供应链网络纳入单一模型来审视。”

Kinaxis首席产品官安德鲁·贝尔表示,该工具运行于微软云平台Azure之上,通过结合大规模数据处理、高级建模和多层人工智能来应对供应链复杂性。

Kinaxis于1984年由软件工程师在渥太华创立,已从一家加拿大初创公司发展为全球供应链软件公司,服务全球数百家客户,包括《财富》500强企业。公司最初推出RapidResponse以改善制造和供应链规划,2024年推出Maestro,是其平台的云原生、人工智能驱动升级版。

时机至关重要。新冠疫情暴露了全球供应链的结构性弱点。此后企业重新设计网络以提高灵活性、多元化采购,并在某些情况下将生产转移至更接近终端市场的位置。近期地缘政治不稳定和关税不确定性进一步加剧了情景规划的需求。

微软Azure是Maestro全球云基础设施的关键组成部分。Kinaxis使用Azure Kubernetes服务和Azure Databricks运行平台、管理大量供应链数据并支持实时运营视图和预测工作负载。其他微软技术包括Azure OpenAI、Azure Cosmos DB、Azure AI内容安全和Foundry IQ,支持自然语言交互、结构化数据存储、治理和搜索等功能。

Kinaxis表示选择微软技术是因为其值得信赖、安全且能满足大型企业需求。Azure为Maestro提供大规模运行的云计算能力,Azure OpenAI帮助Kinaxis以更可控的方式将先进人工智能引入客户系统。

贝尔将Maestro描述为分层系统。基础是数据编织层,汇集内外数据以创建供应链的近实时视图。其上是智能层,应用预测性人工智能、优化模型和机器学习来帮助预判需求变化、评估约束并推荐行动。最后一层是用户界面,规划人员和分析师可使用仪表板、报告和生成式人工智能代理作为助手,提出问题、探索情景并自动化部分规划流程。

Kinaxis正致力于开发集成功能,使Maestro代理能与微软Copilot协同工作。该平台可根据任务调用不同人工智能模型,从需求预测到评估供应约束或生成建议。贝尔说:“这确实是贯穿平台和产品的不同人工智能技术的完整谱系。”

Kinaxis还采用了基于GitHub和GitHub Copilot的代理式软件开发生命周期。人工智能代理现在帮助管理开发流程中的工作,生成任务并作为拉取请求提交供人工审核,使工程师能将更少时间花在常规编码上,更多时间用于架构、判断和验证。

Kinaxis人工智能创新高级副总裁尚塔尔·比松-克罗尔解释说,Maestro的关键能力之一是需求预测。该平台帮助企业预判客户需求变化并相应调整计划。Kinaxis通过人工智能和机器学习增强了这一能力,并使用Azure OpenAI支持近期创新,包括企业需求预测模块,旨在帮助组织在复杂多变的环境中优化预测。

比松-克罗尔指出:“用于一个客户的模型……与另一个客户的模型毫无关系。”需求预测可能涉及数千个客户特定模型,随着数据和运营条件变化定期重新训练。每个实施都根据企业运营进行配置,Maestro通过独立客户环境支持客户数据隔离。

Kinaxis强调负责任的人工智能是平台设计的一部分。人工智能代理在既定权限和保障措施内运行,安全控制、身份管理、测试和评估有助于支持负责任的使用。

另一项新兴能力是代理式人工智能,正开始改变规划人员与Maestro的交互方式。Kinaxis表示客户可配置专门代理来处理部分工作流,如检测问题、生成响应方案和协调后续行动。多个专注代理可在编排代理下协同工作,在既定指令和护栏内运行,调用平台工具、评估选项并提出建议。比松-克罗尔说,这些代理充当系统内的虚拟用户,但人类规划人员仍负责最终决策。

例如,“检测代理”仅专注于识别事件并触发正确工具;“解决代理”处理异常、评估选项并生成建议。比松-克罗尔解释说:“你将检测代理和解决代理串联起来,然后由你(规划人员)按照希望它们运作的方式来编排。”

贝尔指出,最终Maestro不仅仅解决单一问题,“我们实际上是利用代理式人工智能解决跨组织横向的端到端工作流。”据Kinaxis称,今年早些时候推出的这些代理已被一小部分客户使用。

对这位高管而言,更大的意义在于企业不仅仅是试图加快速度,而是在变量不断增多的环境中做出更高质量的决策。贝尔说:“这不仅仅是加速决策和自动化决策,而是通过考虑更多要素来做出更好的决策。”

中文翻译:

如果明天一切都变了呢?一家加拿大公司正在利用人工智能帮助企业应对供应链的不确定性

全球冲突、关税变动、贸易中断和监管不确定性正迫使全球企业重新思考其供应链管理方式。

对许多企业而言,挑战不仅在于将货物从一个地方运到另一个地方,更在于实时了解世界某一地区的 disruption 如何波及工厂、供应商、库存、运输成本和客户需求。

总部位于加拿大的 Kinaxis 正试图帮助企业更快找到正确答案。其旗舰平台 Maestro 利用预测性人工智能、高级情景建模和代理式人工智能,帮助企业预测需求、测试可能的情景,并在条件变化时调整计划。

“客户需要的是极强的适应性和敏捷性。这正是 Maestro 发挥关键作用的方式,”Kinaxis 首席执行官 Razat Gaurav 表示。他指出,以持续中的霍尔木兹海峡冲突为例,客户的情景建模使用量较冲突前水平增加了120%以上。

从汽车制造商到科技巨头,再到能源和航运领军企业,全球企业都在使用这一工具来管理复杂的供应链、应对短缺,并模拟关税和其他中断的潜在影响。对 Kinaxis 而言,这背后是一个更广泛的全球转变:曾经依靠稳定假设运行的供应链,如今需要能够承受持续变化的系统。

Maestro 本质上是一个端到端供应链管理的编排平台。它汇集了企业内外部的数据,包括销售订单、产品信息、产能和库存,并将其与天气、市场变化和突发新闻等外部信号相结合。

企业无需在多个独立系统中分别管理需求计划、库存、物流和生产,而是可以在一个由预测性人工智能和人工智能代理驱动的工具中评估这些相互关联的功能,帮助计划人员了解某一领域的变化如何影响业务的其余部分,并更快地应对中断。

“我们所带来的部分创新……”Gaurav 说,“就是真正将整个供应链网络放在一个单一模型中来看待。”

Kinaxis 首席产品官 Andrew Bell 表示,该工具运行在微软云平台 Azure 上,旨在通过结合大规模数据处理、高级建模和多层人工智能来应对供应链的复杂性。

Kinaxis 于1984年由软件工程师在渥太华创立,已从一家加拿大初创公司成长为全球供应链软件公司,服务于全球数百家客户,其中包括《财富》500强企业。该公司最初推出 RapidResponse,用于改善制造和供应链规划运营。2024年,它推出了 Maestro,这是其平台的云原生、人工智能驱动进化版。

时机至关重要。新冠疫情暴露了全球供应链的结构性弱点。此后,企业重新设计网络以提高灵活性、多元化采购来源,并在某些情况下将生产转移到更接近终端市场的地方。最近,地缘政治不稳定和关税不确定性进一步加剧了情景规划的需求。

微软 Azure 是 Maestro 全球云基础设施的关键组成部分。Kinaxis 使用 Azure Kubernetes Service(AKS)和 Azure Databricks 来运行平台、管理大量供应链数据,并支持实时运营视图和预测工作负载。其他微软技术,包括 Azure OpenAI、Azure Cosmos DB、Azure AI Content Safety 和 Foundry IQ,则支持自然语言交互、结构化数据存储、治理和搜索等能力。

Kinaxis 表示,选择微软技术是因为其值得信赖、安全,并能够支持大型企业的需求。Azure 为 Maestro 提供了大规模运营所需的云计算能力,而 Azure OpenAI 则帮助 Kinaxis 以更可控的方式将先进人工智能引入客户系统。

混合决策力量

Bell 将 Maestro 描述为一个分层系统。底层是数据编织层,将内部和外部数据汇集在一起,创建近乎实时的供应链视图。

在其之上是智能层,应用预测性人工智能、优化模型和机器学习来帮助预判需求变化、评估约束条件并推荐行动。

最后一层是用户界面,计划人员和分析师可以使用仪表盘、报告和生成式人工智能代理作为助手,提出问题、探索情景并自动化部分规划流程。

Kinaxis 正致力于开发集成功能,使 Maestro 代理能够与微软 Copilot 协同工作,后者是面向工作场景的人工智能助手。

该平台可以根据任务类型调用不同的人工智能模型,从预测需求到评估供应约束,再到生成建议。“这确实是贯穿平台和产品的全方位不同人工智能技术的应用,”Bell 说。

Kinaxis 还采用了基于 GitHub 和 GitHub Copilot(微软的人工智能编程助手)的代理式软件开发生命周期。人工智能代理现在帮助管理开发过程中的工作,生成任务并作为拉取请求提交供人工审核。这使工程师能够将更少时间花在常规编码工作上,更多时间用于架构、判断和验证。

Kinaxis 人工智能创新高级副总裁 Chantal Bisson-Krol 解释说,Maestro 的关键能力之一是需求预测。该平台帮助企业预判客户需求变化并相应调整计划。

Kinaxis 利用人工智能和机器学习增强了这一能力,并使用 Azure OpenAI 支持近期创新,包括 Enterprise Demand Forecasting——一个旨在帮助组织在复杂多变环境中优化预测的模块。

“用于一个客户的模型……与另一个客户的模型毫无关系,”Bisson-Krol 指出。

需求预测可能涉及数千个客户特定模型,这些模型会随着数据和运营条件的变化定期重新训练。每个实施都根据企业的运营情况进行配置,Maestro 通过独立的客户环境支持客户数据隔离。

Kinaxis 强调负责任的人工智能是平台设计的一部分。人工智能代理在既定权限和保障措施内运行,安全控制、身份管理、测试和评估有助于支持负责任的使用。

另一项新兴能力是代理式人工智能,它正开始改变计划人员与 Maestro 的交互方式。

Kinaxis 表示,客户可以配置专门的代理来处理工作流的部分环节,例如检测问题、生成响应方案和协调后续行动。多个专注于特定任务的代理可以在一个编排代理的协调下协同工作,在既定指令和护栏内运行,调用平台工具、评估选项并提出建议。Bisson-Krol 说,这些代理在系统中充当虚拟用户,但人类计划人员仍然负责最终决策。

例如,她说,一个“检测代理”只专注于识别事件并触发正确的工具。一个“解决代理”则处理异常、评估选项并生成建议。

“你把检测代理和解决代理拼接在一起,然后你(计划人员)按照你想要它们运作的方式来编排它们,”Bisson-Krol 解释道。

最终,Bell 指出,Maestro 不仅仅是在解决单一问题,“我们实际上是在利用代理式人工智能解决跨组织横向的端到端工作流。”据 Kinaxis 称,今年早些时候推出的这些代理已被一小部分客户使用。

对这位高管而言,更重要的一点是,企业不仅仅是试图加快速度。他们试图在变量数量不断增长的环境中做出更高质量的决策。

“这不仅仅是加速决策和自动化决策。这是通过考虑更多要素来做出更好的决策,”Bell 说。


所有图片由 Kinaxis 提供。

Juan Montes 撰写关于人工智能和数字创新如何重塑拉丁美洲和加拿大各行各业及决策方式的文章。他的报道涵盖从跨国企业为高管部署人工智能代理,到公立学校教师在课堂上采用技术等各类故事。他出生于马德里,曾在西班牙和危地马拉担任记者,并曾任《华尔街日报》驻墨西哥、中美洲和加勒比地区的外国记者。您可以通过 LinkedIn 与他联系。

Gustavo Lo Valvo 是一位专注于新型叙事形式的编辑设计师。此前,他曾在阿根廷《号角报》担任设计总监,领导视觉架构以及印刷和数字平台新闻叙事的创新。您可以通过 LinkedIn 与他联系。

英文来源:

What if everything changes tomorrow? A Canadian company is using AI to help businesses navigate supply chain uncertainty
Global conflicts, shifting tariffs, trade disruptions and regulatory uncertainty are forcing companies worldwide to rethink how they manage supply chains.
For many, the challenge is not only moving goods from one place to another but understanding, in real time, how a disruption in one part of the world could ripple through factories, suppliers, inventory, transportation costs and customer demand.
Canada-based Kinaxis is trying to help companies find the right answers faster. Its flagship platform, Maestro, uses predictive AI, advanced scenario modeling and agentic AI to help businesses forecast demand, test possible scenarios and adjust plans when conditions change.
“What customers need is to be very adaptable and very agile. And that’s the way Maestro plays a very critical role,” says Razat Gaurav, Kinaxis’ CEO. With the ongoing Strait of Hormuz conflict, for instance, customers have increased scenario modeling by more than 120% from pre-conflict levels, he notes.
The tool is used by global companies, from automakers and technology giants to energy and shipping leaders, to manage complex supply chains, respond to shortages and model the potential impact of tariffs and other disruptions. For Kinaxis, there’s a broader global shift: supply chains that once ran on stable assumptions now need systems that can absorb constant change.
Maestro is essentially an orchestration platform for end-to-end supply chain management. It brings together data from across a company, including sales orders, product information, production capacity and inventory, and combines it with external signals such as weather, market shifts and breaking news.
Rather than managing demand planning, inventory, logistics and production in separate systems, companies can evaluate these interconnected functions in one tool, powered by predictive AI and AI agents, helping planners understand how changes in one area affect the rest of the business and respond more quickly to disruptions.
“Part of the innovation we’ve brought…” says Gaurav, “is to really look at the entire supply chain network in one single model.”
The tool, which runs on Azure, Microsoft’s cloud platform, is built to tackle the complexity of supply chains by combining large-scale data processing, advanced modeling and multiple layers of AI, says Andrew Bell, Chief Product Officer at Kinaxis.
Founded in Ottawa in 1984 by software engineers, Kinaxis has grown from a Canadian startup into a global supply chain software company serving hundreds of customers worldwide, including Fortune 500 companies. The company first introduced RapidResponse to improve manufacturing and supply chain planning across operations. In 2024, it launched Maestro, a cloud-native, AI-driven evolution of its platform.
The timing has been significant. The COVID-19 pandemic exposed structural weaknesses in global supply chains. Companies have since redesigned networks to increase flexibility, diversify sourcing and, in some cases, move production closer to end markets. More recently, geopolitical instability and tariff uncertainty have intensified the need for scenario planning.
Microsoft Azure is a critical part of Maestro’s global cloud infrastructure. Kinaxis uses Azure Kubernetes Service (AKS) and Azure Databricks to run the platform, manage large volumes of supply chain data and support real-time operational views and forecasting workloads. Other Microsoft technologies, including Azure OpenAI, Azure Cosmos DB, Azure AI Content Safety and Foundry IQ, support capabilities such as natural language interactions, structured data storage, governance and search.
Kinaxis says it chose Microsoft technology because it is trusted, secure and able to support the needs of large enterprises. Azure gives Maestro the cloud computing power to operate at scale, while Azure OpenAI helps Kinaxis bring advanced AI into customer systems in a more controlled way.
A hybrid decision-making force
Bell describes Maestro as a layered system. The foundation is a data fabric that brings together internal and external data to create a near real-time view of the supply chain.
Above that, an intelligence layer applies predictive AI, optimization models and machine learning to help anticipate demand shifts, evaluate constraints and recommend actions.
The final layer is the user interface, where planners and analysts can use dashboards, reports and generative AI agents as their assistants to ask questions, explore scenarios and automate parts of the planning process.
Kinaxis is working toward developing integrations that would allow Maestro agents to work with Microsoft Copilot, the AI assistant for work.
The platform can draw on different AI models depending on the task, from forecasting demand to evaluating supply constraints or generating recommendations. “It’s really the full spectrum of different AI technologies applied throughout the platform and the products,” says Bell.
Kinaxis has also adopted an agentic software development lifecycle using GitHub and GitHub Copilot, Microsoft’s AI-powered coding assistant. AI agents now help manage work across the development process, generating tasks and submitting them as pull requests for human review. This allows engineers to spend less time on routine coding work and more time on architecture, judgment and validation.
One of Maestro’s key capabilities is demand forecasting, explains Chantal Bisson-Krol, Kinaxis’ Senior Vice President for AI Innovation. The platform helps companies anticipate changes in customer demand and adjust plans accordingly.
Kinaxis has enhanced that capability with AI and machine learning, and it uses Azure OpenAI to support recent innovations, including Enterprise Demand Forecasting, a module designed to help organizations refine their forecasts in complex, changing environments.
“The models used for one customer… have nothing to do with the models of another customer,” Bisson-Krol notes.
Demand forecasting can involve thousands of customer-specific models that are retrained regularly as data and operating conditions change. Each implementation is configured to reflect a company’s operations and Maestro supports customer data separation through separate customer environments.
Kinaxis emphasizes responsible AI as part of the platform’s design. AI agents operate within established permissions and safeguards, security controls, identity management, testing and evaluations help support responsible use.
Another emerging capability is agentic AI, which is beginning to change how planners interact with Maestro.
Kinaxis says customers can configure specialized agents to handle parts of a workflow, such as detecting issues, generating response scenarios and coordinating follow-up actions. Multiple focused agents can work together under an orchestrator agent, operating within defined instructions and guardrails to call platform tools, evaluate options and propose recommendations. The agents act as virtual users inside the system, says Bisson-Krol, but human planners remain responsible for final decisions.
A “detector agent,” for instance, focuses only on identifying events and triggering the right tools, she says. A “resolver agent” handles exceptions, evaluates options and generates recommendations.
“You stitch together your detector agent and then your resolver agent, and you (the planner) orchestrate them in the way that you want them to operate,” Bisson-Krol explains.
In the end, Bell points out, Maestro is not just solving singular problems “but we’re actually solving end-to-end workflows that go horizontal across an organization using agentic AI.” According to Kinaxis, the agents, introduced earlier this year, are already being used by a small group of customers.
For the executive, the larger point is that companies are not simply trying to move faster. They are trying to make higher-quality decisions in environments where the number of variables keeps growing.
“It’s not just about accelerating decisions and automating decisions. It’s about making better decisions by considering more elements,” Bell says.


All photos by Kinaxis.
Juan Montes writes about how AI and digital innovation are reshaping industries and decision‑making across Latin America and Canada. His reporting spans stories from multinational companies deploying AI agents for executives to public‑school teachers adopting technology in classrooms. Born in Madrid, he worked as a journalist in Spain and Guatemala and was a foreign correspondent for the Wall Street Journal in Mexico, Central America and the Caribbean. You can contact him on LinkedIn.
Gustavo Lo Valvo is an editorial designer specializing in new storytelling formats. Previously, he served as design director at the Argentine newspaper Clarín, where he led visual architecture and innovation in journalistic storytelling across both print and digital platforms. You can contact him on LinkedIn.

微软AI最新进展

文章目录


    扫描二维码,在手机上阅读