Anthropic 营销运营团队如何用 Claude Cowork 自动化报告与活动搭建

营销运营团队需要花大量时间使背后的系统与业务保持同步。虽然自动化理应属于他们的职责范围，但现实中大量工作却远非自动化：营销技术工具之间难以干净集成，报告还需手工汇总，落地页得逐一手动搭建。

在 Anthropic，营销运营团队的 Ian Chan 过去每周要花一到两天拼凑每周营销指标回顾；专注活动运营的 Annabel Custer 则要依次在 Salesforce、HubSpot、Swoogo 和邮件工具中点选，为每一个新活动进行设置。两人现在都通过在 Claude Cowork 中建立工作流，将数天的手工工作压缩到几小时。

[视频：Anthropic 营销运营团队使用 Claude Cowork 的演示]

省下来的时间改变了他们的工作重心。Ian 和 Annabel 如今更少在各系统间点来点去，而是把更多时间投入赋能、验证以及底层数据和流程上——这些正是随着公司更多同事自己拉数和驱动项目时，营销团队所依赖的基础。

自动生成每周营销指标报告

在理想世界中，Ian 为营销和领导层准备的周报中每一个指标都该存在于某个仪表盘里，他要做的只是撰写叙事。实际上，有些指标已在仪表盘中，有些尚未从数据仓库迁入，还有一些甚至还没进入数据仓库。新指标可能只存在于一条 Slack 消息或一段通话文字记录中。

在 Anthropic，业务变化速度远快于传统报告管道的跟进能力，Ian 过去每周要花一到两天追踪和校验数据。如今，Claude Cowork 承担了大部分数据搜寻工作。

每周日晚，一个定时任务会触发 Claude 去阅读前一周的回顾和最新的会议文字记录，查看 Slack 中销售团队关注的重点，查询数据仓库，然后把包含数据和若干建议聚焦领域的文件夹留下。

周一早上，Ian 打开 Claude Cowork，拉取初始报告，其中包含指标表格和建议标题或聚焦领域。

（图中为每周业务回顾摘要，使用演示数据创建，所有信息均已匿名化或为公开发表而修改。）

Ian 审阅后，一旦确认或决定叙事重点，就会让 Claude 用支撑细节和示例去展开。有时团队在响应某个销售优先级，有时则是某个产品发布。每到季度转折点，Ian 会让 Claude 以季度计划为开头，并传入季度回顾文档。

Claude 基于同样的数据和叙事生成领导层幻灯片：发生了什么变化、原因以及团队正在采取什么措施。任何后续事项都会成为 Asana 任务。

（图中为每周指标报告，使用演示数据创建，所有信息均已匿名化或为公开发表而修改，包含关键指标和建议聚焦领域。）

当数据对不上时，Claude 会标出差异而不是胡乱猜测。比如在一次销售团队重组后，营销数据与对方不再匹配。Claude 发现缺口后询问 Ian 如何处理。

这一流程基于连接各种营销平台和工具的连接器，以及 Ian 构建并持续更新的三个技能：

• 一个准备技能驱动报告组装，包括聚焦、标题以及用支撑细节展开。
• 一个校对技能将草稿中每一个数字与经过验证的来源逐一核对。
• 一个行动事项技能将后续事项转为 Asana 任务。

在每次周例会结束时，Ian 会让 Claude 总结出现了哪些应该反馈到技能中的问题，比如新的销售重组结构、他做的修正，或者他希望标题采用的新框架。就 Ian 而言，整个过程过去最多需要两天，现在最多只需两小时。

现在，Ian 相当一部分时间转向帮助营销人员构建问题、优化提示词，并解释他们从 Claude 拉取自己的数据后得到的结果。他也有了余力更深入数据层，确保 Claude 理解数据、定义和区域结构的方式与数据仓库一致。

人工验证已成为两条工作流不可或缺的一环——随着 Claude 将营销分析师过去被占用的琐碎手工任务自动化，这一转变正在加速。

自动化活动搭建与数据导入

构建营销活动背后的基础设施，传统上是营销中最手工的流程之一。每一场活动、网络研讨会或整合式营销都需要在 CRM 中设置，在运行邮件序列及其背后自动化的营销自动化平台中设置，在承载注册页和活动落地页的活动管理平台中设置。这些通常是不同的供应商，而且它们之间的集成几乎从未完整过。

在有 Claude Cowork 之前，Annabel 从专用 Slack 频道接收每个请求，然后手动完成一整套序列。现在她的整个流程几乎完全由 Claude 处理。流程从一份需求提交表单开始，请求者在此明确所需帮助的类型：活动搭建、数据导入、报名审核或审批支持。

每小时运行一次的调度技能会读取频道，挑出最紧急的请求，给工单打上标记以避免重复工作，然后将其转给 Annabel 设置的五个专家技能之一去完成所需的工作。调度技能本身不做任何活动设置，它的职责是决定接下来跑什么，将其独立出来可以让 Annabel 单独优化每个专家技能而无需触碰路由。

（调度技能截图，为公开发表做过编辑以展示演示数据与信息，其职责是读取频道并将请求转交给五个专家技能之一去完成所需工作。）

对于最复杂的请求类型——活动搭建，一个活动搭建技能端到端处理完整序列：CRM 活动创建、含工作流和名单的营销自动化活动、活动平台设置、邮件起草、落地页生成，以及它们之间的所有集成。

（活动搭建技能截图，摘录并编辑以供发表，它编排了两次 Slack 更新：一次是 Claude 接到请求时，一次是落地页准备好等待请求者审核并由审计接管时。）

搭建完成后，它将任务转交给一个新的审计代理。审计代理从零上下文开始，在真实落地页上提交测试注册，在 Gmail 中打开确认邮件，如果一切正常则将 Asana 任务标记为完成。Annabel 在每个结果发布前都会进行审核。

这个工作流同样基于连接器，并配合多个 Annabel 构建并随着发现新极端情况而不断更新的技能：

• 调度技能读取需求频道，将每个请求路由给下面对应的专家技能。
• 活动搭建技能执行跨平台的端到端搭建。
• 网络研讨会落地页创建技能为网络研讨会生成落地页。
• 审计技能由一个单独的、全新 Claude 实例运行，它在任务标记完成前验证活动搭建技能的输出。
• 报名审核技能处理注册流程进行中的变更。
• 审批支持技能处理活动审批并按计划时间点发送相应邮件。
• 数据导入技能清洗名单并处理参会者数据。

她还保持着一个独立的“管家”代理。当某次运行出错时，她就打开管家，让它查看发生了什么并提出如何调整的建议。所有值得保留的发现都会被反馈到相应的技能中。

虽然这些自动化工作流将显著节省 Annabel 的时间，但她构建它们的首要动机是工作质量。随着营销团队扩大，营销人员若顺手从现有模板克隆活动页面，可能产生缺陷——比如确认邮件显示错误的城市名，或者落地页链接失效。借助 Claude Cowork，她得以在大规模下保持每次搭建的一致性。

当 Claude 接过活动运营中重复性的部分，Annabel 便能聚焦到更具战略性的项目上，比如赋能，以及为获得更好的洞察而自动化或优化流程和活动架构。

给营销运营团队上手 Claude Cowork 的建议

• 把重复的纠正变成技能。 当你发现自己不止一次地纠正 Claude 同一件事时，这条反馈就该放进技能里。而且你不需要亲手构建技能，Claude 可以替你完成。
• 先建一个校对技能。 校对技能会检查 Claude 放入报告中的每一个数字是否都能追溯到经过验证的来源。
• 让 Claude 自我反思。 Claude 阅读指令的方式与人类书写的方式不同，因此新工作流最初运行几次后，可以问问它在理解指令时有什么难点。Annabel 会将浮现出来的问题反馈回技能，作为她不断更新技能的广泛实践的一部分。
• 善用定时任务。 每个周日晚或每小时自动运行的工作，就是没人需要记住去做的工作。

Marketing operations teams spend a meaningful portion of their time keeping the systems behind marketing programs in step with the business. While automation sits firmly in their purview, a lot of the work is anything but: martech tools don’t integrate cleanly with each other, reports are consolidated manually, landing pages get spun up one at a time.

Ian Chan, on the marketing operations team at Anthropic, used to spend one to two days a week pulling together the weekly marketing metrics review. Annabel Custer, who focuses on campaign operations, used to set up each new event by clicking through Salesforce, HubSpot, Swoogo, and email tools in sequence. Both have now compressed days of manual work into hours by setting up workflows in Claude Cowork.

VIDEO

The recovered hours have shifted the shape of their work. Ian and Annabel now spend less time clicking through systems and more time on enablement, validation, and the underlying data and processes the marketing team relies on as more people across the company pull their own numbers and drive their own programs.

Generating the weekly marketing metrics report

In a perfect world, every metric in the weekly report Ian prepares for marketing and leadership would live in a dashboard and his job would be to simply put together the narrative. In practice, some metrics are in the dashboard already, while others haven’t yet made it there from the data warehouse, and others haven’t been piped into the warehouse yet. New ones might exist only in a Slack message or a call transcript.

At Anthropic, the business moves faster than a traditional reporting pipeline can keep up with and Ian used to spend a day to two days every week tracking down data and validating it. Claude Cowork now handles most of that data hunt.

A scheduled task runs every Sunday evening, prompting Claude to read the previous week's review and the latest meeting transcript, check Slack for what the sales team is focused on, query the warehouse, and leave a folder with the numbers and a few suggested focus areas.

On Monday morning, Ian opens Claude Cowork and pulls the initial report, which contains the metrics tables and suggested headlines, or areas of focus.

The weekly business review summary is shown here created with demo data and all information anonymized or modified for publication.

Ian reviews them and once he’s confirmed or decided where to focus the narrative, he tells Claude to expand on them with supporting details and examples. Some weeks the team is responding to a sales priority, and others—to a product launch. At the quarter turn, Ian tells Claude to lead with quarterly plans and feeds in the quarterly review doc.

Claude generates the leadership slide from the same data and narrative: what changed, why, and what the teams are doing about it. Any follow-ups become Asana tasks.

The weekly metrics report, shown here with demo data and all information anonymized or modified for publication, contains key metrics and suggested areas of focus.

When the numbers don’t line up, Claude flags the mismatch instead of guessing. After a reorg on the sales team, for example, marketing's reporting no longer matched theirs. Claude flagged the gap and asked Ian how to handle it.

The process runs on connectors to the marketing platforms and tools the team uses, and three skills that Ian has built and updates continually:

• A prep skill drives the report assembly, including focus, headlines, and expansion with supporting detail.
• A proofreading skill checks every number in the draft against a verified source.
• An action-items skill turns follow-ups into Asana tasks.

At the end of each weekly session, Ian asks Claude to summarize what came up that should go back into the skills. The new sales reorg structure, for example, the corrections he made, or a new way he wanted the headlines framed. In Ian’s case, the entire process, which used to take up to two days of work, takes up to two hours.

Now, a meaningful share of Ian’s time has moved to helping marketers frame their questions, refine their prompts, and interpret what they get back when they pull their own numbers from Claude. He also has bandwidth to go deeper into the data layer, making sure Claude interprets the numbers, definitions, and regional structures the same way as the data warehouse.

Human validation has become an integral part of both workstreams—a shift that’s accelerating as Claude automates the mundane manual tasks that have traditionally taken up much of marketing analysts’ time.

Automating event builds and data imports

Setting up the infrastructure behind marketing campaigns has traditionally been one of the most manual processes in marketing. Every event, webinar, or integrated campaign needs to be set up in the CRM, in the marketing automation platform that runs the email sequences and the automation behind them, and in the event management platform that hosts the registration page and the event landing page. Each of these is typically a different vendor, and the integrations between them are rarely complete.

Before Claude Cowork, Annabel picked up every request from a dedicated Slack channel and worked through the sequence manually. Her new setup is almost entirely handled by Claude. It starts with an intake form where requesters specify the type of help they need: event build, data import, apply-to-attend, or approval support.

Once an hour, a dispatcher skill reads the channel, picks the most urgent request, stamps the ticket so the work doesn't get duplicated, and hands it off to one of five specialist skills that Annabel has set up to do the required work. It doesn’t do any event setup itself; its job is to decide what runs next, and keeping it separate lets Annabel refine each specialist skill on its own without touching the routing.

The dispatcher skill, edited for publication to show demo data and information, reads the channel and hand off requests to one of five specialist skills that do the required work.

For an event build, which is the most complex request type, an event-build skill handles the full sequence end to end: CRM campaign creation, marketing automation campaign with workflows and lists, event platform setup, email drafting, landing page generation, and all of the integrations between them.

The event-build skill (excerpted and edited for publication) scripts two Slack updates: when Claude picks up the request, and when the landing page is ready for the requester's review and the audit takes over.

When the build is done, it hands off to a new agent for audit. The audit agent starts with no prior context, submits a test registration on the live landing page, opens the confirmation email in Gmail, and marks the Asana task complete if everything looks right. Annabel reviews each result before it ships.

This workflow runs on connectors to the marketing platforms and tools Annabel works with, plus a number of skills she's built and updates as she finds new edge cases:

• A dispatcher skill reads the intake channel and routes each request to the right specialist skill below.
• An event-build skill drives the end-to-end setup across platforms.
• A webinar-landing-page creation skill spins up landing pages for webinars.
• An audit skill , run by a separate fresh Claude instance, verifies the event-build skill's output before the task is marked complete.
• An apply-to-attend skill handles in-flight changes to the registration flow:
• An approval-support skill handles event approvals and sends the appropriate emails at a scheduled cadence.
• A data-import skill scrubs lists and processes attendee data.

She also keeps a separate "manager" agent open. When a run misfires, she opens the manager and asks it to look at what happened and propose what to adjust. Anything worth keeping goes back into the relevant skill.

While these automated workflows will become significant time savers in Annabel’s day, her primary motivation to build them was quality of work. As the marketing team scales, marketers cloning event pages from whatever template happens to be nearby can produce bugs, such as confirmation emails surfacing the wrong city name or broken landing pages. With Claude Cowork, she gets consistency across builds, at scale.

As Claude takes on the repetitive parts of campaign operations, Annabel can focus on more strategic projects, like enablement, and automating or optimizing processes and campaign architecture for better insights.

Advice for Marketing Ops teams on getting started with Claude Cowork

• Turn repeated corrections into skills. When you find yourself correcting Claude on the same thing more than once, that feedback belongs in a skill. You don’t need to build skills, either: Claude can do that for you.

• Build a proofreading skill first. The proofreading skill checks that every number Claude puts in a report traces back to a verified source.

• Ask Claude to reflect. Claude reads instructions differently than a human writes them, so after the first runs of a new workflow, ask what was difficult about the instructions. Annabel feeds what surfaces back into the skill as part of her broader practice of constantly updating skills.

• Lean on scheduled tasks. Work that runs on its own every Sunday night or every hour is work no one has to remember to do.

‍