AI研习社第二期马来西亚站：跨越语言与地域，Cata如何让采购AI接手真实任务？

跨国企业可以集中采购支出、统一系统与流程，却难以复制专业判断。同一品类在不同区域的供应基础、成本结构与业务约束各不相同，全球策略能否落地，仍高度依赖少数专家。

Multinational companies can consolidate spend and standardize procurement systems and processes. Extending category expertise across regions is harder. Supplier markets, cost structures and business constraints vary, while translating global strategies into local action still depends heavily on a few specialists.

业务版图持续扩大，采购能力能否突破专家人数的限制？

10月29日，由SDA与Cata联合发起的AI Learning Lab——AI研习社第二期在马来西亚槟城举行。全球运营、采购咨询与AI技术领域的嘉宾，与当地采购及供应链管理者围绕“采购AI的下一步：从回答问题到交付业务成果”展开闭门交流。

继上海首期之后，本次活动以全英文分享、产品演示与圆桌讨论，展示了包装需求如何进入成本分析、VAVE与寻源询价。当地采购管理者进一步提出：AI能否支持上游供应风险识别、供应商协议比较，以及跨工厂的信息协同？

At the event venue

How can procurement scale beyond the capacity of its expert teams?

On October 29, SDA and Cata hosted the second AI Learning Lab in Penang, Malaysia. Speakers from global operations, procurement consulting and AI joined local procurement and supply chain leaders for a closed-door session on “The Next Step of Procurement AI: From Answering Questions to Delivering Business Results.”

Following the inaugural Shanghai session, the English-language event demonstrated a workflow from packaging requirements through cost analysis, VAVE, sourcing and RFQs. Local managers then raised further questions about upstream supply risks, supplier agreement comparisons and coordination across sites.
全球运营：经验共享之后，如何支持现场判断？

AT&S制造工程高级副总裁、SDA亚太区主席Sean Wu先生，从全球制造与供应链运营的实践出发，谈到了业务扩张与人才培养之间的时间差。企业需要更多专业人员，但招聘和培养都有周期。跨工厂运营还带来另一层挑战：已有经验如何帮助其他团队处理问题，而不必每次重新排查？

Sean先生分享了AI支持知识调用、问题排查与人员培养的思路，同时强调管理层投入、员工参与和专业验证。知识进入系统之后，仍需由专业人员判断其是否适用于当前业务条件。这一讨论对采购组织的启示在于：经验沉淀需要保留判断所依据的条件，使其他团队能够理解何时适用、何时需要重新评估。

SDA Asia Pacific Chair；Senior Vice President of Manufacturing Engineering, AT&SSean Wu

Sean Wu, Senior Vice President of Manufacturing Engineering at AT&S and SDA Asia Pacific Chair, addressed the gap between business expansion and the time required to develop skilled professionals.Operating across sites creates a further challenge: how can existing expertise help other teams resolve problems without repeating the same investigations?

Sean discussed AI-assisted knowledge retrieval, troubleshooting and employee development, emphasizing management commitment, employee participation and expert validation. Once knowledge enters a system, specialists still need to assess whether it applies to the situation at hand.For procurement, this highlights the importance of retaining the conditions behind a judgment, so that other teams can understand when it applies and when it needs reassessment.

品类管理：让策略随着执行结果更新

SDA常务理事会主席Phil Xu先生，将讨论聚焦于品类策略的持续运行。他指出，品类管理需要贯通内部需求、供应市场、策略实施与绩效反馈。市场条件变化、供应商表现和实施结果，都应进入下一轮判断。策略如果不能随之更新，就难以持续指导采购行动。

Phil先生认为，成熟品类经理的培养通常需要八至十年，而个人能够持续跟踪的品类与任务有限。他提出，企业可根据品类特点，从AI辅助分析逐步走向委派明确任务，将更多市场监测、数据分析与执行跟进交由智能体承担。

专业人员则保留利益相关方协调、供应商关系、重大谈判和关键决策。评价这一分工的标准，是策略能否转化为行动，执行结果能否反馈到品类KPI与后续调整。

SDA Council Chair

Phil Xu

Phil Xu, Council Chair of SDA, focused on keeping category strategies active throughout execution.Changes in market conditions, supplier performance and implementation results should inform subsequent decisions. Without these updates, a strategy becomes less useful as a guide to procurement action.

In Phil’s view, developing an experienced category manager typically takes eight to ten years, while each individual can monitor only a limited portfolio. He proposed moving from AI-assisted analysis toward delegating clearly defined tasks, depending on the category, with agents taking on more market monitoring, data analysis and execution follow-up.

Professionals retain responsibility for stakeholder alignment, supplier relationships, major negotiations and key decisions. The test is whether strategy translates into action, and whether results feed back into category KPIs and future adjustments.

Cata的技术路径：把采购方法带入每一步执行

Cata创始人兼CEO Simon Nie先生，解释了模型之外的Harness执行机制：将业务上下文、品类知识、工具与过程控制组织起来，支持AI推进多步骤任务。

据其介绍，Cata围绕“理解—分析—寻源—执行—改进”构建工作流。不同任务需要不同的业务依据：成本分析需要材料、工艺与采购条件，供应商评估需要具体需求与能力信息，后续询价则需要承接已经明确的规格和商业条件。

这些信息能否在任务之间持续传递，决定了成本分析、供应商评估与后续询价能否建立在一致的业务条件上。

针对企业治理要求，Simon先生介绍了敏感数据脱敏安排，并表示客户专有数据不用于模型训练。部署方式根据企业要求评估，客户保留规则定义与关键决策权，重要行动受约定的权限和审批机制约束。

Founder and CEO, Cata

Simon Nie

Simon Nie, Founder and CEO of Cata, explained the “harness” surrounding the AI model: a framework combining business context, category knowledge, tools and process controls to support multi-step tasks.

According to Simon, Cata structures its workflows around “Understand–Analyze–Source–Execute–Improve.” Each task requires specific inputs. Cost analysis depends on materials, processes and purchasing conditions; supplier evaluation requires defined needs and capability data; subsequent RFQs must reflect the specifications and commercial requirements already established.

Carrying these inputs through the workflow helps keep cost analysis, supplier evaluation and subsequent RFQs aligned with the same business requirements.

Simon also outlined arrangements for masking sensitive information and stated that proprietary customer data is not used to train models. Deployment options are assessed against enterprise requirements. Customers retain control over rules and key decisions, with significant actions subject to agreed permissions and approvals.

现场演示：从一项包装需求，推进到寻源与询价

Cata技术专家Archy Liu先生，以一款无线鼠标的包装需求，展示了设计、成本与采购任务之间的衔接。

上传产品图片后，系统继续确认尺寸、重量、销售渠道与运输方式，形成设计方案；随后结合最小起订量（MOQ）、区域等条件开展成本分析，并提出价值分析与价值工程（VAVE）优化方向。演示进一步覆盖供应商寻源、能力比较、询价单生成、报价分析，以及质量异常处理与8D报告。

设计报告、成本分析与供应商比较，使参会者能够具体观察任务所需的输入及其输出。案例也呈现了一个关键关系：设计和采购条件构成成本分析的基础，明确后的需求再进入寻源与询价。

对品类负责人而言，这类应用值得验证的是：能否减少需求整理、成本分析和寻源准备中的重复工作，让专家将更多时间用于假设校验、方案取舍与供应商决策。

Technical Expert, Cata

Archy Liu

Archy Liu, Technical Expert at Cata, used a wireless mouse packaging requirement to demonstrate the links between design, cost analysis and procurement tasks.

After an image was uploaded, the system requested dimensions, weight, sales channels and shipping methods to develop a design proposal. It then incorporated purchasing conditions, including minimum order quantity (MOQ) and region, into cost analysis and proposed value analysis/value engineering (VAVE) opportunities. The demonstration also covered sourcing, supplier comparisons, RFQ generation, quotation analysis, quality issue handling and 8D reporting.

The reports gave participants concrete inputs and outputs to examine. Design and purchasing conditions provided the basis for cost analysis, while clarified requirements informed sourcing and RFQs.

For category leaders, the opportunity to test is whether this approach can reduce repetitive preparation and analysis, allowing specialists to spend more time validating assumptions, assessing trade-offs and making supplier decisions.

企业圆桌：有了市场信息，是否就能预警供应风险？

在Simon先生主持的圆桌中，Phil先生、Sean先生与某世界500强外企采购经理Koay Ai Ling，将讨论推进到具体业务需求。

结合原材料供应紧张的情况，Koay Ai Ling提出了提前识别风险、支持预防措施的需求。Sean先生分享了跨工厂产能、供应承诺与物料分配中的管理问题。Phil先生指出，识别风险来源，需要了解一级、二级乃至更上游的供应关系，并结合品类分析。

这使讨论进入了关键条件：市场变化需要对应到企业对上游供应商与物料的依赖，才能支持影响判断与行动安排。Simon先生回应，专业品类应用需要结合客户知识与业务数据进一步完善。

供应商协议比较，则提供了范围更清晰的切入点。Koay Ai Ling希望利用AI识别条款差异，提高审核效率。嘉宾们讨论了采购与法务的分工：AI支持初步比较和问题整理，复杂或高价值合同仍由专业人员审核。嘉宾也指出，管理层支持之外，团队是否愿意提供专业知识、复核结果并持续反馈，会影响AI能否进入日常工作。

圆桌讨论将试点条件进一步明确：供应预警依赖上游信息与品类知识，合同比较需要清晰的审核分工。企业需要围绕具体任务，确定数据、输出标准与行动责任。

Roundtable sessionIn a panel moderated by Simon, Phil and Sean joined Koay Ai Ling, Purchasing Manager of a Fortune 500 multinational company, to examine specific business needs.

Drawing on raw material constraints, Koay Ai Ling asked whether AI could identify risks earlier and support preventive action. Sean discussed capacity, supply commitments and material allocation across sites. Phil emphasized the need to understand tier-one, tier-two and further upstream supplier relationships through category-specific analysis.

Market developments must be linked to the company’s dependencies on suppliers and materials before they can inform action. Simon noted that specialized applications require further development using customer expertise and business data.

Supplier agreement comparison offered a more clearly scoped starting point. Koay Ai Ling expressed interest in identifying clause differences to improve review efficiency. Panelists discussed AI-assisted comparisons and issue identification, with specialist review retained for complex or high-value contracts.Panelists also noted that, beyond management support, adoption depends on teams contributing expertise, reviewing outputs and providing ongoing feedback.

The discussion clarified the conditions for a pilot: supply risk monitoring depends on upstream visibility and category knowledge, while contract comparison requires clear review responsibilities. Each use case needs defined data inputs, output criteria and accountability.

从个人经验，到跨区域的组织能力

从上海到槟城，采购AI面对不同的语言、供应环境与企业要求，也回应着同一个管理问题：如何让少数专家积累的品类经验，持续支持更多团队？

Cata展示了连接品类知识、分析与执行的应用路径。更长远的价值，在于让成本分析留下依据、供应异常留下经验、寻源决策反馈实施结果，使采购组织在每一次任务中积累能力，并在人员与业务变化中延续专业判断。

从上海到槟城，研习社走过的是地理距离；采购组织需要跨越的，是个人经验与组织能力之间的距离。

Across Shanghai and Penang, procurement AI encounters different lan

guages, supply markets and enterprise requirements. Yet the management challenge remains the same: how can category expertise concentrated in a few specialists support more teams?

Cata demonstrated how category knowledge, analysis and execution can connect. The longer-term opportunity is to retain the reasoning behind cost analyses, capture lessons from supply issues and feed sourcing outcomes into future decisions—building capability with each task and preserving expertise as teams and businesses evolve.

The Learning Lab’s journey from Shanghai to Penang crossed borders. Procurement’s next step is to turn individual expertise into lasting capability across teams and regions.