SaaStr AI:4大供应商处理批量外联,自建AI工具攻坚Top
We Run Monaco, Agentforce and Artisan for Outbound. Why We Still Also Built Our Own “AI ABM” Tool for Our Top Accounts
直接给出了可落地的ABM执行架构:明确区分了批量自动化与高价值人工干预的边界,并提供了具体的数据整合清单与审批流程,读者可直接参考其“供应商+自建薄层”的模式优化自己的销售漏斗。
At SaaStr AI we are aggressive users of AI outbound tools. Monaco runs cold outbound. Artisan runs warm outbound. Agentforce runs our win-back campaigns, which are getting 72% open rates. Our inbound agent on Qualified has closed $2M+ in the past year. 10K, our AI VP of Marketing, now uses several of these tools more than any human on our team does.
在 SaaStr AI,我们是激进使用 AI 外联工具的用户。Monaco 运行冷外联。Artisan 运行暖外联。Agentforce 运行我们的赢回活动,打开率高达 72%。我们在 Qualified 上的入站代理在过去一年中促成了超过 200 万美元的成交。10K,我们的 AI 营销副总裁,现在使用的这些工具比团队中的任何人类都多。
We’re keeping all of them.
我们将保留所有这些工具。
We also recently built our own prospecting tool inside 10K to complement them. It writes detailed outbound pitches using all of our first-party data about an account, including the data that never syncs to Salesforce. We use it on a small set of accounts, and we use the vendors for everything else.
我们最近还在 10K 内部构建了自己的潜在客户挖掘工具来补充它们。它利用关于我们账户的所有第一方数据(包括从未同步到 Salesforce 的数据)撰写详细的外联推介文案。我们在少量账户上使用它,而在其他所有方面则使用供应商的工具。
Screenshot
截图
90% of Our Outbound Still Runs Through Vendors, and Yours Should Too
90% 的外联仍通过供应商运行,你也应该如此
For most outbound, the proven vendors are the right choice, for us and almost certainly for you.
对于大多数外联工作,经过验证的供应商是最佳选择,对我们而言是这样,对你来说也几乎肯定如此。
Outbound at scale depends on deliverability, sequencing, reply handling, meeting booking, list hygiene, and protecting your domain. Monaco, Artisan, Agentforce and Qualified have put years of engineering and thousands of customers into those problems. We have a 450K-contact database, and rebuilding deliverability infrastructure for it would be a poor use of our agents’ time and ours.
规模化外联依赖于送达率、序列编排、回复处理、会议预订、列表清理以及保护你的域名。Monaco、Artisan、Agentforce 和 Qualified 在这些问题上投入了多年的工程资源和数千名客户。我们拥有 45 万联系人的数据库,为其重建送达基础设施将是对我们代理和我们自身时间的低效利用。
If it’s a volume motion at SaaStr AI, a vendor runs it.
如果是 SaaStr AI 的规模驱动型运动,由供应商来执行。
But Most Vendors Only See A Fraction Of Our First Party Data. Mostly Just Part of Our Salesforce Data
但大多数供应商只能看到我们第一方数据的一小部分。通常只是我们 Salesforce 数据的一部分
Where most of these vendors come up short is how much of your own data they can use.
这些供应商大多不足之处在于他们能使用多少你自己的数据。
A typical AI SDR connects to your CRM and reads some of it: contacts, accounts, maybe activity history. At SaaStr, what we know about a company lives in at least six systems:
典型的 AI SDR 连接到你的 CRM 并读取其中一部分:联系人、账户,可能还有活动历史。在 SaaStr,我们对一家公司的了解至少存在于六个系统中:
- Salesforce: contracts, sponsorship history, the AE on the account, lifetime value
- Bizzabo: who attended SaaStr AI Annual, how many times, badge scans, actual lead counts
- Marketing Cloud: who on their team subscribes to our newsletters, and what they open and click
- WordPress: every article we’ve written that mentions them
- Our podcast archive: every time they’ve come up on 20VC x SaaStr or The Agents
- Momentum and Qualified: call history and chat conversations
- Salesforce:合同、赞助历史、负责该账户的 AE(客户经理)、终身价值
- Bizzabo:谁参加了 SaaStr AI 年会,参加次数,徽章扫描,实际线索数量
- Marketing Cloud:其团队中谁订阅了我们的新闻通讯,以及他们的打开和点击情况
- WordPress:我们撰写的提及他们的每一篇文章
- 我们的播客档案:他们在 20VC x SaaStr 或 The Agents 上出现的每一次记录
- Momentum 和 Qualified:通话历史和聊天对话
No outbound vendor will wire into all of that for one customer. The systems they do connect to are also getting more expensive to read from. Salesforce has announced it will meter agent API calls in Flex Credits, and 10K alone already makes ~35,000 Salesforce API calls a day. Deep CRM reads on every prospect will cost vendors more over time.
没有外联供应商会为单个客户接入所有这些系统。他们连接的系统在读取成本上也越来越高。Salesforce 宣布将在 Flex Credits 中对代理 API 调用进行计量,仅 10K 每天就已经产生约 35,000 次 Salesforce API 调用。对每个潜在客户进行深度 CRM 读取将使供应商的成本随时间推移而增加。
The vendors write good emails from the data they can reach. For most of our outbound, that’s enough.
供应商利用他们能获取的数据撰写出色的邮件。对于我们的绝大多数外联活动来说,这就足够了。
Our Pitch Generator Pulls Salesforce History, Event Attendance and Newsletter Data
我们的 Pitch Generator(推介生成器)会提取 Salesforce 历史记录、活动出席情况和时事通讯数据
[Screenshot: 10K’s Prospecting tab with the Pitch Generator]
[截图:10K 的 Prospecting(潜在客户开发)标签页,带有 Pitch Generator]
This is the Prospecting tab inside 10K. It has an Attendee Lookup and Ticket Follow-ups, and for new sponsors we mostly use the Pitch Generator.
这是 10K 内部的 Prospecting(潜在客户开发)标签页。它包含 Attendee Lookup(参会者查询)和 Ticket Follow-ups(票务跟进),对于新赞助商,我们主要使用 Pitch Generator。
You enter a company name. 10K pulls their Salesforce history, event attendance and newsletter data, then writes a custom sponsorship pitch. The pitch can include:
你输入一家公司名称。10K 会提取其 Salesforce 历史记录、活动出席情况和时事通讯数据,然后撰写一份定制化的赞助推介方案。该推介方案可以包括:
- Who on their team has been to SaaStr AI Annual, and how many times
- Which of their executives subscribe to and read our newsletters
- Whether they’ve sponsored before, at what level, and the ROI they got
- What we’ve written and said about them on the blog and the podcasts
- 他们团队中有多少人参加过 SaaStr AI Annual,以及参加次数
- 他们的哪些高管订阅并阅读了我们的时事通讯
- 他们是否曾进行过赞助,级别如何,以及获得的 ROI(投资回报率)是多少
- 我们在博客和播客上关于他们的报道内容和言论
A typical vendor email says something like “companies like yours sponsor SaaStr to reach B2B executives.” 10K’s version can say how many people from their team came to Annual last year, which of their leaders read our newsletter, and what their leads looked like the last time they sponsored. Every one of those facts is about their company, and they can check each one.
典型的供应商邮件可能会说“像贵公司这样的企业赞助 SaaStr 是为了触达 B2B 高管”。而 10K 生成的版本则可以说明去年有多少来自他们团队的人参加了 Annual 大会,哪些领导者阅读了我们的时事通讯,以及他们上次赞助时带来的潜在客户情况。这些事实都关乎他们公司本身,且他们可以逐一核实。
We think of it as an “AI ABM” tool: account-based marketing where an agent does the account research.
我们将它视为一种“AI ABM”工具:即基于账户的营销(Account-Based Marketing),其中由智能体(agent)执行账户调研工作。
The Renewal Agent Sent 20 to 30 Custom Decks Where We Used to Send 5
Renewal Agent(续约智能体)发送了 20 到 30 份定制化演示文稿,而我们过去只发送 5 份
The prospecting tool came out of what already worked on renewals.
这个潜在客户开发工具源于在续约工作中已经行之有效的做法。
As we covered on Episode #013 of The Agents, Amelia built a renewal agent on top of 10K in about half a day. It pulls the Salesforce side (contract, history, LTV, opens, Qualified chats, Momentum calls) plus data that was never in Salesforce (WordPress, social, the podcast archive, Bizzabo lead counts), then builds a custom deck through the Gamma API.
正如我们在《The Agents》第 #013 集中所介绍的,Amelia 在半天内基于 10K 构建了一个续约智能体。它会提取 Salesforce 侧的数据(合同、历史记录、LTV、打开率、Qualified chats、Momentum 通话记录),再加上从未存入 Salesforce 的数据(WordPress、社交媒体、播客档案、Bizzabo 潜在客户计数),然后通过 Gamma API 构建定制化演示文稿。
Before agents, we built real custom decks for about five renewals, the diamond sponsors. Everyone else got a templated follow-up. With the agent, 20 to 30 custom decks went out.
在引入智能体之前,我们为大约五个续约客户(钻石级赞助商)构建了真正的定制化演示文稿。其他人则收到模板化的跟进内容。有了智能体后,我们发送了 20 到 30 份定制化演示文稿。
The silver sponsors, with the smallest checks and historically our lowest renewal rate, replied at a higher rate than the diamonds. A $25K sponsorship is a bigger decision for a smaller company than $300K is for Google Cloud, and a custom deck showed them we’d tracked their results as closely as a diamond’s.
银级赞助商虽然支票金额最小,且历史上续约率最低,但他们的回复率却高于钻石级赞助商。对于一家小型公司而言,2.5 万美元的赞助决策远比 Google Cloud 面临的 30 万美元决策更为重大,而定制化演示文稿向他们表明,我们对其成果的关注程度不亚于对钻石级赞助商的关注。
The prospecting tool applies the same approach to new logos.
潜在客户开发工具将相同的方法应用于新客户(new logos)。
Two Rules From the Renewal Agent We Kept for Prospecting
我们从 Renewal Agent 中保留下来用于潜在客户开发的两条规则
The agent proposes a narrative, and a human approves it before anything gets built. For one silver renewal, the agent proposed “you’re a silver, upgrade to gold.” That company had come out of stealth right before the event and grown a lot since, so Amelia changed the pitch to three options, including a media-plus-content tier. Fixing the story before generation took a few minutes. Editing a finished deck would have taken much longer.
代理提出一个叙事方案,人类在构建任何内容之前对其进行审批。在一次银级续费中,代理提出了“你是银级,升级为金级”的方案。该公司在活动前夕刚刚结束隐身模式并实现了大幅增长,因此阿米莉娅将提案改为三个选项,包括媒体加内容层级。在生成前修正故事只需几分钟。编辑已完成的演示文稿则会花费更长的时间。
The first email is short, and the deep version goes in the follow-up. On renewals, a first email without the deck got better responses than leading with it. Replies to that first email also told us what to put in the pitch. The agent sends the first touch, and a human sends the detailed follow-up.
第一封邮件简短,深度版本放在跟进邮件中。在续费方面,不带演示文稿的第一封邮件比直接附上演示文稿获得了更好的回复。对那封第一封邮件的回复也告诉了我们提案中应包含的内容。代理发送首次触达,人类发送详细的跟进内容。
Our Setup: Four Vendors for Volume, One Internal Tool for Named Accounts
我们的设置:四个供应商用于批量处理,一个内部工具用于重点账户
This is how we’ve set it up, and we’d make the same call again:
这就是我们目前的设置方式,我们会再次做出同样的决定:
- We buy the volume layer. Four AI SDR vendors run the bulk of our outbound. We haven’t rebuilt deliverability or sequencing, and we don’t plan to.
- We built a thin layer on our own data for top accounts: event attendance, newsletter engagement, content mentions, call history. The renewal agent took Amelia half a day. The prospecting tool lives in the same app.
- A human approves the story on every named account before the agent writes the pitch.
- 我们购买批量层。四个 AI SDR 供应商负责我们大部分的外展工作。我们没有重建送达率或序列,也没有计划这样做。
- 我们在自有数据上为顶级账户构建了一个薄层:活动参与、新闻通讯互动、内容提及、通话历史。续费的代理花了阿米莉娅半天时间。潜在客户挖掘工具位于同一个应用程序中。
- 在代理撰写提案之前,人类会审批每个重点账户的故事。
"We couldn't get that from third-party services," says Jason Lemkin.
杰森·莱姆金说:“我们无法从第三方服务中获得这些数据。”
SaaStr wanted hyper-personalized outreach: the perfect email, the perfect deck.
SaaStr 希望进行高度个性化的外展:完美的邮件,完美的演示文稿。
So they built it themselves. "It just can't, today, it can't be bought."
所以他们自己构建了它。“目前,它无法被购买。”
17,000 conversations, 600 meetings, and a 2.1x increase… pic.twitter.com/AjpbUQKEzw
17,000 次对话,600 次会议,增长 2.1 倍… pic.twitter.com/AjpbUQKEzw
— SaaStr.ai (@saastr) September 29, 2026
— SaaStr.ai (@saastr) 2026年9月29日
更进一步:量化金融体系
看懂新闻只是起点——沿量化金融路径,把它变成能交付的工程能力