Linear COO 拆解如何将 ARR 规模化至 1 亿美元并转型智能体编排系统
Scaling Linear’s GTM to $100M ARR — Cristina Cordova, COO
Linear COO Cristina Cordova 在访谈中透露,公司今年 ARR 突破 1 亿美元,拥有超 4 万付费客户且净收入留存率达 177%。目前 Linear 上超过 50% 的 Issue 由智能体创建,公司已从传统工单追踪定位转型为围绕智能体编排的产品开发系统。
访谈揭示了Linear从工单工具转向智能体协作系统时的GTM策略与定价权变迁,对探索AI原生工作流的SaaS团队有直接参考价值。
译文尚不完整,完整内容请切换到原文。
Linear 创立之初,目标只是做一款更好用的问题追踪工具。历经七年发展、融资 1.34 亿美元后,其 CEO Karri Saarinen 却发文直言:《问题追踪已死》。本期节目中,我们邀请到 Linear 的首席运营官 Cristina Cordova,聊聊公司如何完成这一转变,以及其市场拓展策略在此期间发生了哪些调整。
加入 Linear 之前,Cristina 曾在 Stripe 工作过七年半,在 Notion 掌管过平台与合作业务,还曾担任 First Round 的合伙人,随后于 2023 年加入 Linear。Linear 的几位创始人在 Coinbase、Airbnb 和 Uber 饱受传统问题追踪工具之苦,于是萌生了自己做一款的想法;后来,这款产品凭借自助模式在 Vercel、Ramp 等初创公司中迅速走红。
她刚加入时,Linear 只有约 40 名员工,而且全公司仅有一名销售。从那时起,Linear 陆续推出了 MCP 服务端、让开发者在 Linear 中构建智能体的平台,接入了 Codex、Cursor 和 GitHub Copilot,并推出了自研的 Linear Agent。一年前,Linear 上只有 3% 的工作由智能体创建,而现在这一比例已过半。今年,Linear 的年度经常性收入(ARR)突破 1 亿美元大关,付费客户超过 4 万家,净收入留存率高达 177%;今年 8 月,员工还以 25 亿美元的估值套现了 9900 万美元股票。
我们在节目中聊到了很多话题:为什么她入职三周就叫停了对大企业客户的追求,为什么 Linear 从未招聘过一名 SDR(销售开发代表),为什么他们坚决不允许大语言模型独立回复客服工单,以及短短数月 ARR 破亿美元背后所反映的 AI 预算趋势,对整个行业究竟意味着什么。
Linear 上过半工作如今由智能体创建
Linear 于 2025 年 5 月推出了自己的 MCP 服务端,Cristina 表示这在同类产品中基本算是首创。一年半后,Linear 中创建的问题里有超过一半出自智能体之手。
但这并不意味着 Linear 只想做个套了 MCP 服务端外壳的数据库。Cristina 认为,满足于现状的 SaaS 公司正在犯大错。如果产品内部的智能体工作流做不到比外部更出色,用户就根本没有理由打开你的 App。
因此,她密切关注着两条指标曲线:智能体使用量和 App 使用量。如果智能体使用量节节攀升,而 App 使用量停滞不前,说明你的产品正在沦为底层大宗商品。而在 Linear,两条曲线都在稳步上扬,因为大家仍会回到 App 中来查看和协调各项工作。
“问题追踪已死”
今年 3 月 Karri 发表《问题追踪已死》一文时,Linear 的大部分客户都深表赞同。但紧接着他们就提出了一个顺理成章的问题:那我买的到底是什么?Linear 给出的答案是:一套围绕智能体编排构建的产品开发系统,让团队告别手动提 issue、打标签、分派任务和结项等繁琐的体力活。
客户一直在倒逼他们往这个方向演进。多数前沿 AI 实验室都在用 Linear,并希望它能跑得更快;而像 Ramp、Coinbase 这些诞生于 AI 爆发前的公司,也在构建自己的内部智能体,急需一个 API 把它们接入 Linear。
Cristina 坦言,Linear 足够小,能够放手进行自我革命,这是一种幸运。相比之下,Google 面临的困境要棘手得多:一旦在每条搜索结果上方都放上 AI 生成的答案,看广告的人自然就少了。
数月 ARR 破亿背后的 AI 预算
Cristina 表示,如今推动采购提速的关键正是 AI 预算。企业特意拨出了以往并不存在的 AI 工具专属资金,任何能套进这笔预算的产品,都能避开以往激烈的预算争夺战。在某些企业,这笔预算的数额非常庞大。
变化已经显现。买家现在签的是一年期合同和付费试点,而不是三年长约,因为谁也说不准两年后自己需要什么。至于那些所谓“一夜做到1亿美元ARR”的神话,很多不过是靠几笔签给国防企业或大型律所的三四千万美元大单堆出来的。这与 Linear 的业务截然不同——截至今年 6 月,Linear 的 4 万家付费客户完全是按席位付费的。
等这一年期合同到期续约时,谁的 ARR 是真金白银,自然见分晓。
Linear 推翻了其 2025 年企业级路线图
2023 年入职仅三周,Cristina 就从 Linear 的公司目标中划掉了“进军大企业市场”。当时公司只有一名销售。她转而瞄准那些不得不替换现有工具的成长期企业,就像 Stripe 在说服成熟客户转投自己之前,很长一段时间都是靠拿下初创新业务打下江山的一样。
直到 2025 年,Linear 才正式宣布进军大企业,列出的路线图里满是为大组织准备的功能。但紧接着,智能体浪潮来袭。Linear 随即打造了 MCP 服务器、一套拥有全新 API 的智能体平台,并集成了 Codex、Cursor 和 GitHub Copilot。这一切此前全无规划,许多企业级功能也因此被搁置。
Cristina 还是决定全力冲刺。竞争对手还在观望,大企业却已开始把 Linear 列入候选名单。
不设 SDR,也没有前置部署工程师
Linear 从未招聘过一名 SDR。Cristina 认为,资质审查电话纯粹是浪费客户时间:客户想要的是产品演示,而不是被盘问预算。AE 自己负责开拓新客户,并借助工具保持沟通的个性化。
他们也不设前置部署工程师。Linear 本身应当足够易用,根本不需要外部人员接入客户系统。解决方案工程师会协助大客户销售,有时一晚上就能搭出一个可用的 ServiceNow 集成,以此向潜在客户证明其可行性。Cristina 还补充道,相比每个团队对“Bug”定义各不相同的工具,一款理念鲜明的工具能为智能体提供更干净的数据。
客户支持全部由内部团队负责,Linear 绝不让大语言模型单独回复工单。AI 负责起草回复,人工决定最终发送内容。由于回复速度太快,客户有时甚至以为这是 AI 自动发的。三年半以来,Cristina 从未接到过因客服体验不佳而升级的投诉。
所有人都想要按效果付费
Cristina 很推崇 Intercom 旗下 Fin 的模式——按 AI 成功解决的客服对话收费。但对 Linear 来说,这要难得多。把一个 Bug 自动分派给正确的团队到底值多少钱?毕竟有的公司认为这本就是产品经理的事,有的公司却不这么看。
在有人攻克这一难题之前,Linear 的原则是:产生实质成本的收费,其余一律免费。席位制依然保留;Linear Agent 的对话虽然接入前沿模型,但也完全免费。在编程会话中,用户可自选 OpenAI 或 Anthropic 的模型与运行环境,Linear 仅在官方公布的 API 价格基础上加收微薄利润。Loops 则是按 AI 额度扣费。
Linear 不会对原本免费的功能转为收费。新的 AI 功能在上线时均为免费测试版,并明确提示后续将收费;而且定价策略是先高后低,绝不中途涨价。
为什么不雇 100 个销售?
Cristina 表示,Linear 目前的人员规模大概只有同阶段、同营收规模公司的一半。如果销售团队规模超过公司其他部门的总和,研发团队如今对销售的信任也就不复存在了。
她把在 Stripe 沿用的“四成法则”带到了这里:工程、产品、设计和数据团队在全公司的人数占比至少要达到 40%。销售团队要是想招更多人,就得先协助研发招到足够的人。虽然 Linear 并没有定死这条规矩,但她入职后做的第一件事之一,就是做了一张按职能划分的人员编制表,而且至今都在密切关注这个比例。
公司有意放慢了招聘节奏。候选人必须针对实际业务问题完成一项试岗考核。她当年的试岗课题,是为 Linear 制定未来 6 到 12 个月的市场开拓计划,而这随后也成了她实际负责的工作。如今 Linear 已经有大约 170 名员工,但高管团队里依然只有两位创始人和她。
节目提要
Linear——Cristina 担任首席运营官的产品研发协同系统。
与创造者共享 Linear 的成长——8 月公司以 25 亿美元估值完成要约收购,付费客户突破 4 万家,净收入留存率(NRR)达 177%。
Issue 追踪已死——Karri 3 月撰文探讨 Linear 为何要专为智能体打造产品。
Linear MCP 服务器——2025 年 5 月上线,Cristina 称这基本是同类产品中的首创。
Linear Agent 发布——内置智能体,其聊天对话功能一直向用户免费开放。
我们如何打造 Triage Intelligence——自动给新提交的 issue 打标签、查重的功能;在 Cristina 看来,这就是“无需做成智能体”的 AI 范例。
Linear 内的代码编写功能——6 月上线,打破了 Linear 此前单一的按人头付费模式,通过 AI 点数调用 Claude Code 与 Codex。
Loops 功能发布——定期执行的智能体自动化任务,同样按 AI 点数计费。
Intercom Fin 定价机制——每解决一次对话收费 0.99 美元,这种按结果付费的模式让 Cristina 羡慕不已,可惜并不适用于 Linear。
时间戳
0:00 开场引言
0:51 对话 Linear 首席运营官 Cristina Cordova
1:39 为什么 Linear 刚到 40 人就招了 COO?
5:01 她的试岗考核:为 Linear 撰写市场开拓方案
12:11 把“大企业客户”从公司目标中划掉
20:00 为什么光做 MCP 服务器背后的数据库远远不够?
23:39 在“issue 追踪已死”之后,产品该怎么卖?
34:40 架构更扁平的小团队,以及完全不设 SDR
36:14 AI 代写客服回复,人类最终把关
42:23 Linear 为什么不做现场驻场开发?
47:53 为什么有些 AI 初创公司能这么快做到 1 亿美元 ARR?
50:06 从按席位收费转向按用量计费
57:07 搁置 2025 年的大企业路线图,全力押注智能体
1:03:28 Linear 现在的护城河是什么?
1:05:54 为什么不直接招 100 个销售?
1:08:59 Stripe 的研发人员占比“四成法则”
文字实录
根据最新剪辑版本自动转录生成;人名与用词可能存在讹误。
0:00 当 Karri 写下那篇“工单追踪已死”的博文时,你怎么向客户解释?比如怎么跟他们说:“你看,我们以前的核心业务就是工单追踪,而现在自家 CEO 却刚发了这么一篇博文”?
我们业务的一个优势在于,公司体量还没大到一转型就会把自己颠覆死的地步,对吧?确实如此。比如换作 Google,想亲手颠覆搜索业务肯定难如登天,对吧?如果在每个搜索结果顶部都放一个 AI 框,可能就没人看广告了,整个商业模式都会被掀翻。而对我们来说,虽然 issue(工单/任务)这个基本单元依然存在,但我们不该再让大家纯靠手动去追踪工作进展了。没错,真正死去的正是这种手动追踪。所以在博文发布后,我们跟很多客户沟通时,大部分客户的反应其实是:“没错,手动追踪确实该淘汰了。顺便问一句,那你们现在的定位到底是什么?我花钱买的又究竟是什么?”
Cristina Cordova,你是 Linear 的 COO,能先做个自我介绍吗?
好的。我是 Cristina,加入 Linear 大概三年半了,入职时就是 COO,一直任职至今。加入 Linear 之前,我在 Notion 负责平台业务,主导了 API 的构建,同时也负责生态合作与自助式增长项目。再之前,我在 Stripe 工作了七年半,早期负责了许多商务合作与 GTM(市场拓展),后来转到产品侧,为初创团队打造了几款新产品。所以从产品、管理研发团队,一路到把控公司的商业化运营,这些领域我都深度参与过,这真是一段非常有趣的经历。
1:30 那为什么 Linear 在只有大概 40 名员工的时候就招了 COO?这种情况算少见还是挺普遍的?坦白讲,我通常觉得并不是每家公司都需要 COO。对吧?有时候有创始人跑来跟我说,“我们好像找到产品市场契合点(PMF)了,接下来该怎么把商业化这块做大?”这种时候,我往往反而会劝他们先别急着招像我这样的人。我觉得每家公司的情况都不太一样。作为创始人,你必须真正想清楚自己现阶段缺什么、招进团队的人需要具备哪些特质。很多时候在早期,你可能只是单纯需要有人去把东西卖出去,对吧?因为光靠创始人兼 CEO 一个人去谈下每一个新客户,已经超出负荷了。这种情况下,我的建议通常是:直接招销售人员,或者招个销售主管来组建团队。但如果你对市场、运营或者公司其他板块具体需要什么都还没有清晰的概念,那其实大概率没必要招像我这样的人。就拿 Linear 的几位创始人来说,他们当时极其专注于设计、工程和产品,而且三个人之间的技能互补做得非常棒。但随着公司商业化版图的扩张,面对招进来的下一个销售,他们可能并不完全清楚“这到底是不是我们要的人”,也无法准确判断销售团队的表现究竟算不算顶尖。所以,如果你审视公司的某些业务板块时,发现自己不知道“做到极致到底是什么样”,也不太清楚“该怎么达到极致”,那么到了这个阶段,招一个像我这样的人或者招一位职能线负责人,往往就会很有帮助。
3:00 明白了。那当时的业务规模大概是怎样的?方便透露当时的营收水平吗?我们当时大概……怎么说呢,确实有一些大客户,但还没有上市公司级别的客户。我记得可能没有哪家客户的席位超过 500 个左右。所以基本上我们的客群上限就是中型市场(mid-market)。而且那时候我们最出名的标签,就是“特别适合初创公司的好产品”,但当时未必能算得上是一款适合其他类型公司的优秀产品。在那个阶段,我们的产品驱动增长(PLG)模式走得很顺,绝大多数人都是通过自助注册开始使用我们的产品,然后随着自身发展逐步增购。像 Vercel、Ramp 以及很多类似的客户,其实从头到尾都没跟 Linear 的任何销售或客服沟通过。他们自己注册、直接上手,然后规模就越用越大。这当然很棒,说明产品本身就能承担大量的获客工作。但与此同时,我们也到了必须主动去谈大客户的阶段——就是那种不签正式合同连一丁点数据都不可能放进你系统里的企业客户。
4:30 比如做试点项目,调查他们团队之前用什么工具,现在这个工具怎么样、有哪些优势。对吧?所以对公司来说,要向规模更大的企业客户拓展,打法完全不同。而这基本就是我的核心目标。我加入之后,公司业务规模至少增长了 10 倍。所以……你是大约四年前加入的?三年……快三年半前吧。对。哦,好的。Linear 的几位联合创始人提过这件事,我记得你也提过:在正式加入 Linear 之前,大家都要做一次试岗项目?对。所以你也做了试岗项目?是的。好的。那具体是怎样的体验?我当时的项目其实挺直截了当的,大概就是给 Linear 做一份有明确时间周期的 GTM(市场拓展)计划,比如规划未来 6 到 12 个月的发展。对吧?我觉得这种方式的好处在于,这就是你入职后的实际工作。你拿到的不是那种虚构的、脱离实际业务的假题目,而是一个真实的项目,基本上就是我入职后要干的正事。所以如果他们不满意我做出的成果,那确实就不该录用我。对,我觉得这种方式特别好:一方面我能表达对公司改进的想法,另一方面我也能向一线员工收集反馈。当时团队里有一位销售、一位负责客户成功、几个做客户支持的,还有一两个人负责市场营销。这样我就能从不同业务视角了解情况,明确自己应该聚焦在哪里、优先级是什么。
6:00 而且,能借此确保公司内部也认可这一愿景,对我帮助很大。等我真正入职时,手上就已经有了这样一份规划蓝图,明确自己的职责和目标。能具体讲讲那份计划里都包含了什么吗?大部分内容偏向品牌和市场营销。核心思路是想方设法让 Linear 成为“不可替代的必然选择”——也就是说,你要么已经在用 Linear,要么就应该在计划迁移到 Linear。我们想争取的大多数企业客户,都应该处于这两种状态之一。因此很多工作都围绕品牌建设和故事打造展开,努力提升公司的知名度。因为 Linear 之前虽然有名,但也只是局限在特定的小圈子里,圈外其实没什么人知道。对,确实。那时我们虽然也在做营销,但唯一的动作就是发布新版本后发一封更新日志邮件。结果发现,订阅邮件的用户还不到 10%。很有意思的是,大多数人甚至连我们上线了新功能都不知道。除此以外,我们可能也就是发条推文。
7:30 好。比如是通过创始人的账号,还是通过公司账号发?然后创始人会去转发,毕竟他们自己的社交媒体影响力也挺强。但基本上也就仅此而已了。所以很多时候我们都在琢磨:要怎么用更好的方式讲一个更完整的故事?又该如何真正触达我们想触达的人群?后来发现,其实很多用户每天都在用我们的应用。那为什么不直接在他们每天都在用的地方,向他们推送产品更新的通知呢?所以我们一直在思考,有哪些好方法既能向老用户推广,又能拓展全新的客户,同时还契合品牌调性。我在这上面花了很多时间。另外很大一部分工作是扩建销售团队。很多初创公司,尤其是工程师文化浓厚、偏向产品驱动增长(PLG)的公司,往往觉得“好产品自己会说话”。但后来大家都会发现,单靠产品只能走到一定阶段。如果想跑得更快,就必须开始考虑用不同的方式去触达用户。并不是所有人都会走自助购买流程,尤其是当你要采购一套对产品开发流程至关重要的核心系统时——我们的产品正是如此。很多企业如果不先和具体的人建立起信任,是不会做决定的。我刚做入职试岗的时候,销售团队总共就一个人,那个人还不是你。所以光靠一个人,能做的事情终究非常有限。
9:00 对吧?那他们是有经验的销售吗?对,是的。他现在其实是我们的销售负责人。好,很棒。也就是掌管我们整个销售团队。他人很优秀。不过在当时,我觉得这倒不是他的问题,更多在于公司层面。主要是我们在思考:我们到底想怎样扩张这支团队?销售团队的扩张路径有很多种,你可以招大批 SDR。嗯。但我们一个 SDR 都没招。现在还可以招前置部署工程师(forward deployed engineers),这做法很流行,对吧?但我们其实不做前置部署工程。我们招的是方案工程师(solutions engineers),我觉得这两者还是有区别的。不过那是后来才加进来的。我们当时主要……那这也归在销售体系里吗?对。挺酷的。我们团队一开始的核心策略,就是招募有技术产品销售经验的 AE。因为我们清楚,如果你以前向设计师或工程师卖过东西,我们就不需要从头教你怎么跟他们打交道。向眼光极其挑剔、不想看花哨 PPT、只想直奔 Demo 演示的客户做销售,对,这和向人力运营(people ops)或类似岗位的人推销是截然不同的。所以我们专门物色具有这种对口经验的人才,确保他们能顺利入职、接受培训,并真正精通产品本身。对,因为只有这样才能建立起客户的信任。所以关键是要想清楚这家公司有何独特之处,招聘和团队扩张又会有哪些不同。另外,还要看创始人们想怎么做,以及我们能否落地执行。客户支持就是一个很有意思的例子。我面试的时候,几位创始人态度非常明确:绝不把客户支持外包。而我之前待过的每一家公司都是外包的。所以当时给我的感觉就是……
10:30 对。我还以为 Stripe 是内部自己做的呢?没有,他们是外包的。哦,这样啊。挺有意思的。除非我离开之后有变动,不过他们内部确实也有人做支持。是的,但他们确实也外包。所以对我们来说,这是我们极力想要保留的原则,因为我们非常在意能否提供高水准、快速响应的服务体验;而且要明白,很长一段时间里我们根本没有销售团队,对吧?正因如此,我们深知客户如果想找真人沟通,首要触点往往就是客户支持。所以支持体验必须做到极致。对吧?直到今天,我们依然摸索出了一套方法,把支持团队做成了完全不外包的全球化团队。至于未来会怎样,我们拭目以待。这种事谁也说不准。但我们确实有自己认准的做事方式,只要对业务切实可行且合情合理,我们就会尽可能长期坚持下去。
12:00 是啊。所以你刚加入的时候……我没做过销售,但你会不会想:好,我们先从初创公司切入,不可能一上来就直接去啃大企业做销售,得一步一步来。你当时是不是也觉得得先拿下中型企业,然后再找机会进军大企业?这到底是怎么运作的?对,是这样。我刚加入 Linear 时,公司大概定下了三个目标,三个公司层面的目标。我记得其中一个就是进军大企业市场。我入职三周后公司办了次全员团建。我走到全员面前——当时公司总共也就三四十人吧——放了列有那三个目标的 PPT,当场把“进军大企业”那一项给划掉了。因为我觉得,我们销售就一个人,实打实就一个人,一个人怎么可能去开拓大企业市场?你必须先想清楚:自己到底能做什么?为了实现目标还需要摸索什么?我们手里真的有达成目标所需的资源吗?而我觉得当时这三样我们一个都不具备。我们之前能拿下那些客户,主要是因为他们自身成长为了成长期企业或中型企业,但我们当时其实没怎么主动拓展过纯新增的中型客户。所以我把目标改成了:我们要去主攻成长期企业。如果能拿下大量成长期客户,而他们原本已经有了功能类似的工具,我们得跑去说服他们把原有工具换掉、全面换成 Linear——如果我们能做到这一点,我才会有底气说:一两年后,我们可以去攻大企业了。
13:30 但当时我根本没有这个信心,那何必给团队定这种目标呢?所以我完全赞同,关键就是要小步快跑、循序渐进,确保先把眼下能服务好的企业服务到极致。这样一来,当大企业打来背书调研电话问他们“你们觉得 Linear 好用吗?体验怎么样?”时,他们给出的全都会是非常正面的评价。这反过来能帮我们拿下纯新增的客户。我之前在 Stripe 时也是差不多的逻辑。有很长一段时间,我们并没有从竞品那里挖来多少新客户,但凡是有新成立的企业,比如第一次接入支付系统的初创公司,我们基本都能全盘拿下。之后,我们才逐步开始拿下那些从竞品迁移过来的公司。这种事情急不来,只能靠时间沉淀。去销售一款必须把原有技术连根拔起、彻底替换掉的产品,和在现有系统之上卖一套光鲜亮丽的全新工具,难度完全不可同日而语。
15:00 是的。既然在这个岗位上有类似的经历,当时最关键的就是带着团队聚焦于我认为真正能解决的问题。直到 2025 年……前面说的是 2023 年,直到 2025 年,我才站到全公司面前宣布:我们今年要发力做大企业市场了。对吧?哇。也就是从那时起,我们启动了企业销售业务,组建了销售团队。是的,所以……好的,你们当时主要是做中型企业客户。所以你们基本上得让大家……因为大多数人……大多数公司都已经有工单/事务追踪工具(issue tracker)了,对吧?对。或者类似的工具……所以你们得想办法把他们拉过来。那么,你怎么去说服他们?他们可能会说:“我们已经在用某某产品了,挺顺手的。不过你来推销看看,我为什么要换?”在早期,产品必须比现有的替代方案好出一大截才行。对。我觉得光是好上 20% 或 30%,不足以让人换掉现有工具,因为这涉及迁移成本,对吧?所以你必须想清楚,怎么讲透你的产品比市场上现有的成熟方案好在哪里——那些方案能被那么多公司使用,说明本身已经运转得很好了。不过我们切入这个赛道有个有利条件,这也是 Linear 创始人选择这个方向的原因之一:他们过去在 Coinbase、Airbnb 和 Uber 工作时,全都是那些传统工具的用户。对。所以他们心里非常清楚,根本没人真正喜欢用那些老工具。
16:30 没错,对吧?当公司开始推行这些工具时,大家的反应往往是:“哎,好日子到头了,又得用这些难用的工具了。”对吧?所以从很多方面来看,他们切身体验过行业竞品带来的痛苦,清楚那是种什么感觉,这对我们帮助极大。这让他们能想出办法,做出一款真正让人有动力去管理工作的工具。按常理来说,你通常不会对这种工具有太高期待。因此,我们早期开发时把大量精力放在了设计和交互体验上。我们的目标是做出一款让人每天都愿意打开、离不开的产品。但难点在于,大家必须亲自上手试过才能体会到它的好,光看几页销售 PPT 是很难真正领会的。所以在战略上,我觉得我们做得非常漂亮的一点,就是做了一个门槛极低、任何人都能直接上手的免费产品,等用量达到一定上限后再开始收费。很多后来在公司里推广 Linear 的人,起初只是出于个人需要免费试玩,结果感受到了它有多顺手,比起竞品好用太多。我们就是这么起步的,在一段时期内效果非常好。不过现在我们销售 Linear 的方式已经大不相同了。我们通常会和企业聊聊过去这一年半到两年里行业发生的翻天覆地的变化——整个产品开发流程已经被彻底颠覆,打造产品的方式完全变了。
18:00 如果现在的世界里还全靠手动操作,那就太脱节了:发现了一个 bug,手动提交到 issue 跟踪系统里;打上标签,指派给另一个人;那个人修好后更新状态、关闭 issue,再去通知客户……整个流程里的每一步全都是纯手工操作。在我们当下这个时代,这种做法已经完全说不通了。你跟大多数公司聊,他们也都会说:“没错,我完全同意。”因此,在过去这一年半里,我们全面转向以智能体为核心,让智能体去接管工作流中那些实质上并不属于产品开发的部分。我刚才描述的那一整套链条,根本算不上是在做产品。所以,如果能把 Linear 打造成极其适合智能体工作流的工具,那么使用 Linear 的工程师、产品经理或设计师,就能把更多时间真正花在做产品上,而不是在系统里把任务搬来搬去,只为了证明自己在为公司创造价值。
19:30 确实。这非常有意思,因为早期大家主打的卖点似乎一直都是用户体验,对吧?而现在看来,人类似乎都不再直接操作界面了。我其实觉得现在很多公司都觉得,哪怕沦为一个纯数据库也无所谓,只要搭个 MCP 服务器,把所有数据开放给智能体,让智能体来跟它们交互就行,甚至觉得这是个绝佳的商业模式。但归根结底,我认为这算不上什么好模式。你必须把智能体工作流直接融入自己的工具里,并且做得比外界用来跟你系统交互的工具更好。否则,用户根本没有理由再打开你的工具,对吧?所以对我们来说,我们日常关注的核心指标——这里确实发生了巨变——比如我们在 2025 年推出了 MCP 服务器,基本上是同赛道里第一家这么做的。而现在,短短一年半左右的时间,Linear 上超过 50% 的 issue 都是由智能体创建的。
21:00 没错,现在工作的创建方式已经和以往彻底不同了。这足以说明短时间内发生了多么巨大的变化。但与此同时,如果我们看有多少用户打开应用、并在应用本身里面进行整合或创建 issue,这个数字同样在大幅增长。好的。对吧?所以有时大家会以为,‘我现在什么都靠 MCP 服务器搞定就行了。’但你其实还是会回到应用里去操作,因为你需要将全局工作可视化。而在很多情况下,聊天界面并不是可视化团队所有工作的最佳方式,它并不适合协作,因为对话基本上只是‘单人模式’。嗯。所以对我们来说,只要 MCP 的使用量和应用本身的使用量两条曲线都在齐头并进、一路向上,就证明我们的体验既适合人类,也适合智能体,对吧?我认为作为一款产品,这正是我们的目标——两端都要做到极致。智能体的出现并不意味着你可以从此不用管界面了。嗯。那应用的使用量只是按使用人数来衡量吗?因为我猜你们的客户群体本身也在增长,对吧?那你们怎么…… 创建的工作
22:30 在应用内部。要知道,我们最基本的工作单元就是 issue。如果只看创建的工作,那就是人打开我们的移动端、桌面端或 Web 端应用去新建任务,对吧?再看任务在流程中的流转,比如分配给智能体,有时是人工手动分配,有时是智能体自动处理,或者说是委派给另一个智能体。如今,我们很多时候已经直接融入了这种工作流。你可以使用我们名为 loops 的功能,它会自动把某个 issue 分配给智能体,并让智能体直接针对它开始写代码。所以在很多情况下,活儿是智能体干的,而且是 Linear 里的智能体。这正是我们希望看到的。因此我们必须综合考量多个不同因素,但总体而言,衡量活跃度主要还是看应用会话数。这就让我想起 Karri 写的那篇题为《工单追踪已死》的博文。我猜他写那篇文章时,你们手头正谈着不少合同吧。那你当时怎么向客户解释——“我们以前的全部业务都是工单追踪,结果我们 CEO 刚发了这篇博文”?你们是怎么跟客户沟通的,这种转变又是如何发生的?
24:00 是的。我觉得我们业务的一大优势在于,体量还没大到无法自我革新的地步。要是换作 Google,想干掉自己的搜索业务恐怕极难。如果在每个搜索结果顶部都放一个 AI 对话框,可能就没人去看广告了,整个商业根基都会被颠覆。我认为这就是很多前 AI 时代的工具都会遭遇的“创新者的窘境”,我们当然也是一家诞生于前 AI 时代的公司。但对我们来说,幸运的是,既然看到世界正在朝这个方向发展,与其死守一个几年后可能就不复存在的时代,不如主动拥抱它,立志为新时代打造最好的产品。如果我们刚才聊到的那种每一步都需要人工介入的工作流程不再是未来的常态——你看现在的 Linear,虽然 issue 这个最小单元依然存在,但我们未来的世界不应该再需要人们手动去跟踪工作了。所以说,真正死掉的是“手动追踪”。那篇博文发布后,我们跟各家客户沟通时,绝大多数客户的反应其实是:“没错,手动追踪确实死了。不过顺便问一句,你们现在到底是什么定位?我买的究竟是什么?”
25:30 所以它现在更偏向于定位成一个“产品开发系统”。我们希望大家能专注于做出更好的产品,而我们打造的系统能让你省去绝大多数繁琐的手动工作。这也意味着我们在开创新品类。虽然我们在很多方面依然属于原有品类,但团队已经把绝大部分资源都投向了这个新品类——比如智能体编排,让人们不再需要在系统里一步一步手动推进任务,无论在 Linear 内部还是外部,都能以更高度智能化的方式协同运作。所以对于这种转变,或者说那篇博文,你当时是被打了个措手不及吗?比如 Karri 写的时候……还是说?不,不,我们全员想法都是完全一致的。那是因为你们看到客户有这样的需求,还是你们想推着客户往这个方向走?我觉得确实有一部分客户有这样的需求。举个例子,绝大多数前沿 AI 实验室都在用 Linear,顶尖 AI 公司的大多数也已经是我们的客户了。如果看这批客户群——也就是在 Linear 之后成立的新一代公司,他们很多都明确表示:“这就是我们要走的方向,我们希望 Linear 也能朝这个方向发展,否则它就不适合我们了。” 与此同时,我们也看到一些在 AI 爆发之前就存在的公司,比如 Ramp 或 Coinbase,它们早就成立了,但也决定投入内部团队开发内部智能体来协同办公。所以我们当时就觉得,其实需要提供一套 API,让大家能在 Linear 内部构建智能体。
27:00 所以,如果你的公司虽然在 AI 浪潮前就成立、但依然走在 AI 技术的最前沿,正在打造专属智能体来协助管理内部的各类工作或代码,并且希望对相关专有数据拥有完全所有权,我们希望能打造出这样的 API,让你可以把 Linear 与这些工具深度打通。在当时,没有任何竞品拥有类似的系统或 API 套件。所以,我们再次率先开创了许多与 Linear 交互的全新方式、智能体工作流形态,并且把智能体直接做进了 Linear 内部。这一切对我们来说都非常新颖,有些客户确实向我们提出了需求,但我必须承认,很多客户并没有提。不过我可以告诉你,现在很多客户都非常庆幸我们拥有这些功能,因为并不是每家公司都恰好身处产品开发模式变革的最前沿。
28:30 我想确认的其实是这个:我猜既然你们想往企业级市场走,那很多客户应该都是那种偏传统的公司,对吧?是的。他们原本并不是那种在推着我们往前走的公司。对。但现在正是这批公司主动找上门来,说:“我们现在很多做事情的方式都变了,我们用来打造产品的系统,迭代速度也该跟上编码智能体这类工具的发展节奏。”嗯。要是你觉得自己的底层工具链在眼下这个新环境里没怎么变过、也没跟上进化,那大概就该考虑换一套工具了,对吧?那当你们开始做这些新功能时,怎么教客户上手呢?你们具体是怎么做的?对,是这样。首先我觉得,我们一直很强调让销售和客户成功这些团队吃透产品。我们非常注重让销售和客户成功团队互相分享经验心得。比如我们很多通话都会用 Gong 录下来,客户也经常会屏幕共享,给我们展示他们的工作流。我们就能借此学习,发现“这家公司是这么搞的,很有意思”,说不定还能把这套经验介绍给别的客户。对吧?所以一方面是向客户学习——如果你的客户本身就是一线开发者,你就更应该多想想怎么从他们身上学东西,而不是单方面想着怎么教他们。对吧?
30:00 有 AI 实验室当客户确实挺好的。没错,太对了。然后你再把学到的这些东西打包整理,教给那些还没走到这一步的用户,对吧?比如之前有客户提过:“要是有个智能体能帮我把所有事情都分流梳理好,那该多爽啊。”因为大家都不想手动去整理需求池(backlog),对吧?这种工作现在看实在太原始了。于是我们就做了一个叫“分流智能”(triage intelligence)的功能,不管是 Bug 报告还是功能需求,只要进入分流流程,就会被自动打标签、分配责任人并放入对应团队的待办队列。人工完全不需要介入,所有事情就自动办好了。这样你根本用不着自己去搭一个智能体,我们直接内置搞定。对。所以说,不是什么事情都非得做成一个智能体,直接做进产品里就行。这既让我们懂得怎么去教下一家客户把事情做得更好,也帮我们把这些经验沉淀进了产品本身。
31:30 是的,这样一来,我们团队的人卖起来就容易多了。那这是怎么做到的?比如,你们是怎么教销售人员使用工程工具的?Linear 的妙处就在于,任何人都能轻松上手。比如像 GitHub Issues 之类的工具,作为非工程师,我一进 GitHub 就会觉得:哎呀,我会不会把什么搞砸了?我根本不属于这里,对吧?但 Linear 的优势恰恰在于打破了这种隔阂。我们努力打造一款不仅适合工程师、所有人也都能顺畅使用、设计极佳的产品。对吧?所以现在,越来越多的企业开始全员普及 Linear。像销售、运营团队,或者任何想了解、想参与产品路线图规划的人,都在用我们的产品,而且用起来毫无门槛。这就让事情变得容易得多,因为作为一个系统或工具,它完全不让人感到畏惧。我们的销售人员也逐渐体会到了这一点。而且正如你所料,我们在公司内部的大量工作都是在 Linear 上进行的,所以基本上只要进公司,就必须得学会怎么用它。此外,很大程度上还得归功于出色的员工培训,我们在赋能(enablement)上投入了大量精力。销售要想获得销售 Linear 的认证资格,能单独跟客户通电话,实际上必须完成一系列课程。所以我们对销售代表的产品理解深度要求极高。其中最复杂的工程工作流,可能要数我们的代码审查产品以及 GitHub 集成了。这些往往比较复杂,因为你得配置 GitHub,还要搭建一个 GitHub 演示工作区和一个 Linear 演示工作区,确保它们能协同工作,这部分确实会稍微复杂一些。
33:00 但那主要是因为它依赖外部工具以及相关的集成工作流。而真正让人兴奋的是,现在我们团队里的销售人员甚至可以直接给网站提交文案修改;我团队里的市场人员也会主动提出:“我们想调整这部分数据的存储方式,以便更清楚地了解这个收入运营系统的运作情况。”类似这样的事情,现在每个人都能参与进来。我们团队的工程师只需要审查一下修改,然后点击批准就行。我觉得这也让 Linear 这款产品以及我们所做的事情对他们来说变得更加容易触及。不过,这确实需要一个过程。
34:30 没错,任何事情都是这样。那你们的团队架构有变化吗?因为现在像不同团队都会直接去更新网站文案,放以前可能根本不会这样。而且你有观察到客户的团队结构在变吗?比如因为他们使用 Linear 的方式,或者单纯因为基于智能体的开发模式普及了?确实。大家现在讨论很多的,其实就是团队如何变得更精简、更扁平。我感觉见过很多公司一开始都想保持这种运作模式,但规模做大之后就慢慢变味了。不过幸运的是,当行业里掀起这些变革时,Linear 本来就是这种扁平模式,而且一直保持了下来。这其中很大一部分当然得益于 AI。举个例子,别的团队做法可能不同,但我们从来没有设过 SDR(销售开发代表)。因为我们一直认为,销售人员自己就应该会写推销邮件,完全没理由不会。那么 SDR 平时一般干什么?通常要么是主动联络潜在新客户、把他们拉进销售管线,要么就是对找上门来的线索做初步资质审核。所以初次沟通电话一般是他们打。但这种初筛电话对客户来说体验很差,因为对方一上来就会盘问:“你们团队多大?”“你们大概能出多少预算?”我们的客户极其讨厌这种电话,心想:“你老查我户口干嘛?我来是看产品演示的。”
36:00 就是说啊,客户会觉得:“我凭什么浪费 30 分钟打电话,就为了排期见一个能真正给我做演示的人?”所以我们直接取消了这个岗位,一方面是因为客户体验不好,另一方面也是因为完全可以搭建系统、工具和自动化流程,让客户经理(AE)能更轻松地主动联络客户,写出更有针对性、更有价值的内容,从而真正促成有效互动。这算我们和其他公司的一个不同之处吧。另外,我们也会去审视业务里哪些环节可以用更少的人,哪些可以通过工具来优化。比如我们的客户体验(CX)团队就是个典型例子:我们有哪些工具可以用来自动起草回复,应对客户发来的咨询?那用大语言模型给客户写回复,不会生成那种 AI 垃圾内容吗?你们怎么解决这个问题的?CX 团队里的人就是最终的质量把关人。所以我们不会让 AI 直接群发。很多其他公司会这么做,这也无可厚非,各家有各家的选择。但确实很多公司会要求:凡是进到客服团队的问题,百分之百先让大模型去挡一层。
37:30 没错。只有在客户愤怒地回信时,人工客服才会介入,说一句“实在抱歉,我来给您一个真正的答复”,对吧?我们根本不认同这种模式。我更倾向于让 AI 根据掌握的信息先针对具体问题起草一份回答,再由人工来评估这个回答是否合适。但很多时候情况往往是:本来三句话就能讲明白的事,AI 却洋洋洒洒生成了三整段,结果反而没提供什么真正有用的信息。我们收到过很多关于客服的反馈,用户的反应基本分两类:要么问“这是 AI 回复的吗?你们响应太快了”——其实只是我们团队动作迅速,大家因为太快就误以为是 AI;要么就是感叹“哇,这是我很长时间以来体验最好的一次客服互动”。对市面上其他工具和产品来说,这其实挺让人唏嘘的。但对我们而言,这正是我们对团队赞赏有加的地方,因为它树立了良好的品牌形象。有时是在客户刚接触产品的初期,有时则是用了好几年的老客户第一次来咨询,这都会让他们确信自己当初的选择是正确的。所以我认为工具可以用,它们终究只是工具。但我绝不赞同盲目使用,以至于损害品牌形象,甚至让团队觉得自己可有可无、不够专业,因为大家把本该由人掌握的知识全都甩给了工具。
39:00 是的。那……你刚才提到了客户体验团队的精简?倒不是裁员。但我认为团队规模没有像没有这些工具时那样大幅扩充。所以我觉得某些工具确实很有帮助,让我们不必单纯随着业务的快速增长而线性扩大团队规模。在其他业务环节也是如此,比如市场营销等部门:我们会为了撰写客户案例去采访客户,然后用 AI 工具转录访谈录音、提炼出最吸引人的亮点,但最终的故事依然由人工撰写。关键就在于决定在哪些环节去用它。我们并不喜欢很多 AI 生成文本的语感,所以我们的做法是让它帮我们梳理与客户两小时访谈中的精彩片段。但在很多层面上,为了呈现 Linear 所追求的质感——也就是那种量身定制、精工细作的感觉——我们在某些事情上绝不会假手于这些工具。
40:30 是啊,至少在它们变得更好之前是这样。所以……我最近注意到 Astra 的写作水平相当不错。对。但不管怎么说,是的。售前工程师是如何融入你们团队的……或者说融入销售团队的?对。我们内部把这个岗位叫做解决方案工程师。不好意思。其实本质上是一回事。他们就像是我们大客户经理(AE)的搭档,也会支持增长业务线中规模偏大的客户。所以当客户提出非常深入的工作流问题时,他们会花很多时间和销售一起配合。或者客户会想:“我们系统通过非常定制化的 API 接了那边某个工具,这种情况下怎么迁移到 Linear 呢?”对吧?这就需要解答更深入的技术问题。对。或者:“我们要怎么把三万人从传统工具迁移到 Linear?”对吧?这时候解决方案工程师就能带着技术思维切入。他们所有人都能……而且跟他们对接的人大概率也是工程师吧?对,几乎全都是。所以对方会问非常具体细节的问题。有时候我们团队的人甚至会搭出完整的端到端演示:“你想看我们怎么接入 ServiceNow?我昨晚刚好做了一个 ServiceNow 集成,现在展示给你看。”这样一来,客户面临的各种系统迁移就变得格外轻松,因为他们会发现:“哇,原来这些事情可以做得这么快,迁移其实没那么难。”毕竟这边的工程师一天之内就能搞出一个快速导入工具,对吧?
42:00 没错,所以有时候我们是在一路辅助客户完成迁移。但我们通常不会像前置部署工程师那样,直接获取客户系统的访问权限进去帮他们落地部署 Linear。我觉得这很大程度上是一种产品理念。我们始终坚信,Linear 应该足够简单,任何人都能轻松集成。有些竞品甚至衍生出专职管理员岗位,那些人在公司里的全职工作就是维护那套工具,成了专门伺候工具的人。对于那些支持近乎无限度定制的产品来说,这种模式或许行得通。但 Linear 不是那种产品。我们更倾向于做一款目标明确、专为特定场景打造的工具。所以在 Linear 里,做事的方式非常明确,而且我认为这也能让智能体更容易理解。很多人从其他工具转到 Linear,在原来的工具里,“Bug”这个标签在一个团队眼里是一个意思,在另一个团队眼里又是另一个意思。允许各种团队以完全不同的方式工作固然可以,但这意味着数据会变得极难辨析。因此我们发现,在我们的系统里,正因为有了这些规范,智能体在读取、理解和协作时都轻松得多。所以我觉得,这种产品特性,再加上有技术人员配合销售……通常我们的销售对接的是客户内部认可我们的人,也就是内部支持者,他们会说:“Linear 棒极了,但我说服不了另外某个人。”
43:30 “能帮我一把吗?”比如去找 IT、采购,或者其他平时根本不用这款工具的人。对吧?他们可能完全体会不到痛点。而我们的销售人员基本上就是这个人的后盾,会说:“好,那我怎么帮你在内部推销?”工程师通常不习惯在公司内部向其他人安利工具,对吧?这项技能需要慢慢摸索,我们常常能帮上忙。我们发现,销售人员处于一个很有意思的位置——借助其他人来帮 Linear 做推广。我很想知道,像这种情况通常要花多久……关于客户最终完成购买需要多长时间,你有没有一个经验法则?比如当他们必须去找有决策权的人审批的时候?对,也就是内部推动者(champion)。不好意思打断了。
45:00 确实。有时候一个决策最初可能只是源自一个人,这种情况我们见过。接着他们搞定了 CEO,CEO 直接自上而下拍板:“我们这周末就全迁到 Linear。要么适应,要么走人。”决策风格非常专断。而这种决策可能一个月就能敲定。这通常发生在小公司……不过,我们其实在一家 1000 人的公司也碰到过这种情况。所以说,如果遇上这种 CEO……我个人倒不属于这一类。如果我觉得团队该用某个工具,我会说:“不如大家评估一下,自己做决定?”但其他公司的运作方式大不相同。你必须考虑特定组织的内部政治,而任何组织在某种形式上都带有政治色彩。接着,你就要想办法把决策过程引导到对你的业务有利的方向。但我们也见过相反的情况:有一家公司大概谈了两年甚至两年半。并不是说他们一直停留在销售管线里,而是他们先来接触,接着说“哎,我们没能说服 CTO”。过了九个月,换了另一个人来找我们,说“我觉得这次能说服 CTO 了”。最终往往要花漫长的时间才能让他们通过审批、拍板决定。有时候是因为预算问题,比如“我们今年的预算已经定死了,如果要追加开支,那得等到 2028 年再做决策了”。有些公司确实就是这么运作的。
46:30 Are you saying that like they're now operating like a lot faster because of the way like things are changing so fast now or not really still? Maybe like these more legacy companies? Yeah. I don't think they operate any faster, honestly. At least with decisions like this. Yeah. I would say for a lot of these companies, the things that... The thing that's helping them move faster is that like AI tools have effectively been like unblocked. Right? So if they believe that... The engineering team in particular believes that like this tool is going to make them better and it fits in like the AI budget. Yeah. For some companies, you know, that is maybe not infinite, but it is very high. Yeah. Whatever that budget is. And it was like, you know, add on budget that kind of had to come out of nowhere. Right? I think that's going to change. Yeah. It's starting to change for companies. But in a lot of ways, you if can like fit your company into that AI budget, then you are unlocking kind of budget that previously didn't exist. Right? And so that makes it far easier, I think, for companies to make a decision because like, oh, there is budget. We found... We've magically found budget that didn't previously exist before. Yeah. Know? Okay. So you think like a lot of this like AI companies are, you know, getting to a $100,000,000 of revenue within like three months? That's because of this basically. It's a lot of companies that previously didn't have budget for like experimenting with new tools. They're just kinda like, oh my god, our company's gonna die if we don't try any of these tools. So like, let's just experiment. And then, you know, at some point maybe a year from now, two years from now, that's gonna stop. Like they're experimenting right now and then, you know, they're gonna figure out what actually works, what doesn't.
48:00 I think I think a lot of companies are experimenting. And, you know, you're signing like one year deals maybe with companies or you're signing like pilot agreements with companies that have costs associated with them still. But you're not signing like a three year deal with a company because you're like, I don't know if I'm gonna like this. Yeah. You know, in two years, like, things are changing so rapidly. Is it normal for a company to sign a three year deal? If you're trusted Yeah. Like, we we sometimes sign very long term deals for that reason. It has trade offs. Right? Like, we can't move up our prices Oh, yeah. In that three year period of time unless you have, like, step ups built in to the agreement. But you know that you have that committed revenue. Right? So, you know, like... And for the business, the same is true. They feel like maybe they're getting a discount Yeah. For, you know, three years that they wouldn't have otherwise been getting. But but I would say like to the example you you had, I mean, I think a lot of those businesses are are very concentrated in the customers that we have. Right? So it's very different, I think, to be like a business like Linear that is, you know, 40,000 paying customers. Right? And like, historically, up until June was like an entirely seat based business. Right? Much harder to, like, grow revenue to, like, from zero to a 100 Yeah. Million overnight. Right?
49:30 And then, you know, you see companies where, I don't know, you sell to, the defense industry or you sell to, like, gigantic law firms or, you know, something like that. Yeah. And, you know, you sign a few deals that are, like, 30 to $40,000,000 in ARR. And then, okay. Great. We got to a 100,000,000 ARR. Yeah. Right? So, you know, certain kinds of businesses do have, like, more of that profile. So you're usage based now? We have some features that are usage based in terms of how we charge for them. Yeah. So how do you how do you convince your customers that that's that's a good idea to go from, like, seat based to usage based? Yeah. Is that hard to convince them? Yeah. I mean, I I would say, like, a lot of companies that are in, you know, the SaaS world Yeah. Are are certainly moving to usage based pricing. I think everyone would love for there to be like... Almost like the Intercom Fin model that existed for a long period of time where... Which was like, oh, if AI responds to your support ticket, like, you know, that costs a dollar. Yeah. And and you're like, is every support ticket that gets replied to, like, worth a dollar? Yeah. Like, okay. Great. Like, right? So you're you're pricing on, like, the outcome that you drive. Right? Think I it's a lot harder for other types of businesses to be like, okay. So what is automatically triaging this bug to the right team worth?
51:00 Yeah. Right? Maybe to every company, that's a little different. Maybe you believe, oh, that's the PM's job. Like, they they should be, like, you know, triaging bugs into the right queue, right, to the right person. And, you know, maybe another company says, well, I'd want my PM to spend their time talking to customers and, you know, building the product roadmap and doing other things instead. So I think there can sometimes be a debate in functions that are not very clearly outcome oriented, how you charge for usage based pricing. Mhmm. So I would love for there to be a world where, like, there's some kind of, like, outcome based model that makes sense for our world. I don't know that there is. So, you know, for now, we're mostly charging, you know, based off of like costs that we incur to run our services. Mhmm. And in some cases, those costs can be, you know, pretty high. Yeah. If you're running like a coding session in Linear or if you're doing things like, you know, run through my entire backlog and tell me what I should work on next. You know? You could have a very large backlog of a lot of things. Right? So, you know, those kinds of even just like chat questions to the agent can be like more expensive. We've decided to make certain things free. So like, you know, Linear agent chat conversations free, and we think we can continue to drive down the costs for that over time. You guys are using like open source models for that?
52:30 Right now, we're using Frontier. Okay. Even for the chat. Even for the chat. Okay. Okay. Yeah. Yeah. And, you know, TBD what the future holds there. And then beyond that, when you have a coding session in Linear, you can choose between like right now, OpenAI or Anthropic, models and harnesses. And the user ultimately gets to decide there and we charge basically by saying, like, here's what we were charged from the API. Here's the published rate and we charge like a little bit on top of that. Mhmm. And that's it. So making it very clear and transparent Yeah. For most users. And then for products like Loops, we have an AI credits model as well for that feature. So, you know, for everything else, it's a little different. The vast majority of our AI features are just like baked into Linear and you don't have to pay extra for them. So... How do you, like, how do you go about going to your customer and saying, hey, previously you guys were doing seat based pricing and now we're gonna do this? Yeah. I mean, I think it helps that, like, the seat based pricing isn't going away. Right? Like, so you're still paying for the core service, and we are not charging for things that were previously otherwise free. Oh, I see. Right? So it's it's very different, I think, when you go to a customer and you're like, So, like, what was free is now paid and usage based. Are, like, new features, basically. Yeah. Exactly. So sometimes we'll have a feature that's in a beta period, and we don't exactly know what it's going to cost or how to charge for it with usage based pricing because like, you know, you can, for example, run a very expensive Linear agent query and you could run a very inexpensive one. Right?
54:00 Yeah. So charging based purely off of like, you know, $1 per per query, like, wouldn't make sense in in that world. Right? So sometimes you just need to let customers, like, play with a new product to determine, like, what it's going to cost Mhmm. And what the best pricing model is. So typically, what we'll do is we'll have, a feature that's in beta, and we'll say, hey, this is an AI feature in beta. You know, we expect that we will charge for this in the future. But right now, it's free. Mhmm. Like, feel free to use it. See if you get value out of it. Give us feedback. And then we're... When we're ready to charge for it, we will let you know and give you notice. And I think that's worked generally, like, pretty well for us so far. And, you know, just trying to make sure that we're very clear with our customers. Sometimes, you know, we can have a six month beta period and then five months later, you're like, oh, I totally forgot that this was free. And in beta, I'm gonna be charged for it, and I'm disappointed. Right? It's hard... You know, there's nothing we could have done other than, you know, I don't know, warning banners in our product like, you are using a free feature that will soon be paid. There's only so much you can do about that. But I think just trying to, like, be open and honest with your customers that, we don't know what the business model is for this. Yeah. Like, we know we're going to charge for it because it's expense... It's an expensive product to serve to you. But we are going to try and, be as transparent as possible. And, you know, we hope you trust us. Right? And hopefully, like, every time one of these things happens, you are building more trust over time. I'll also say we tend to bias towards making things like more expensive at the start and then decreasing costs rather than like making things really cheap and then being like, oops. Like, actually, we can't solve... Like, serve this product for that cheap. Right? So just being like conservative in that way and where we are, like, able to lower cost for customers so that we get more adoption, we will.
55:30 Yeah. So you mentioned earlier that, like, when you came in, you had this, like, six, twelve month, like, go to market plan. Did you basically follow that go to market plan, like, to a tee, or did you have to change it along the way? And then how did that change back in 2025 when you guys started building the new agents features? Yeah. Yeah. I think I probably did every... I mean, I haven't looked at it in a while. Yeah. But I think I probably did everything, you know, for the most part, at least certainly at this point. Maybe I didn't do it all in the six to twelve month period that I thought I would. But a lot of it was like team building, hiring, you know, starting to think about launches, like product launches in a much more robust way. Like, you don't just, like, ship something and, like, move on to the next feature. You have to, like, continuously educate and and market products over time. So we think a lot of those things we've, like, adopted into, like, new make motions and ways of doing things for the team. And, you know, we certainly... And we went from, like, no sales team almost to to it being a very large percentage of the company now. So so, you know, I think a lot of those things have, like, ultimately, you know, shipped and and gotten out the door. And what was the second part of your question? It was like, how did you have to change that go to market strategy back in twenty twenty twenty twenty five when, like, the agents features came out?
57:00 Yeah. I think a lot of it was just, you know, sometimes, I think from a go to market perspective, you have a you have a long term road map. Right? And, you know, a lot of those things at the time were like enterprise features and other things to kind of help us go up market. And that was also... 2025 was a year where we said we were going after enterprise. Right? And then we saw everything that was happening in this space, and we're like, oh, we've gotta build an MCP server. Oh, we've gotta build an agent's platform with a whole new set of APIs. So anyone can build an agent into Linear. Oh, we've gotta build partnerships with like Codex and Cursor, and GitHub Copilot. So all of these agents are in Linear. And we did all of those things, but none of those things were planned. Like, we did not know we were going to do any of those things at the beginning of 2025. So that doesn't mean we did all of that and everything else that we planned for. It meant that, like, a bunch of the enterprise stuff that we planned for, you know, didn't get done. Yeah. So Meaning, like, there were, like, certain customers you were trying to go after? Or Like certain features that we wanted to build for, you know, workspaces that were of a certain size that would, like, you know, help us scale to like larger and larger organizations. And, you know, again, a lot of those things didn't get done. But I looked at it and I said, well, do we wanna be like behind all of this new stuff or do we wanna be like really racing ahead of where we think we should be? I think we chose to say like, no. We're gonna race ahead. And I think I also believed that if we did that and we did it well, that it would also win us new enterprise customers, just not in the way that we thought we would win them.
58:30 Yeah. Right? It wasn't gonna be about, you know, building, like, sub teams and multilayer, like, work streams, like roadmaps that work across multiple different teams, which you might want if you're a large organization. Right? No. It was more about, you know, building these things that make Linear work really well with agents. And, you know, because we built that and and our competitors, you know, waited a long time to do the same thing, I think we got ahead in a way that we wouldn't have otherwise. And a lot of enterprises put Linear on the list to be considered that wouldn't have otherwise. If you could go back to when you first started at Linear, knowing what you know now, like, what what would you tell yourself? Or what do you wish you knew? I feel like I probably go back and be like, you know, calm down. It'll all be fun. Like, it's probably what I would say. And what I mean by that, what that's, like, really alluding to is when I joined Linear in 2023, the startup market was not doing well. Right? Had all these dbt just came out. You had all these companies, yeah, in, like, 2021 and 2022 that raised these, like, gigantic funding rounds and were doing really well, you know, as we were kind of coming out of COVID. And then in 2023, like, it was the great, like, startup culling. Like, companies couldn't really get funded. A lot of companies were, like, cutting costs, like dramatically, trying to consolidate spend. So in in that world, it, you know, it was like much harder to grow, especially when in 2023, we were focused on startups. And then like growth stage companies a little bit. Right? So, you know, I think in a lot of ways, sometimes you enter into a space in like a market that's hard.
1:00:00 Right? Luckily, it was hard for everybody else too. So, you know, it wasn't like, you know, it was, you know, disproportionate in a way. But, you know, it did, like, affect the business. And, you know, you're trying to figure out ways to grow without that natural growth where you would have otherwise gotten. Mhmm. And then I think similarly with, like, all of the agentic development and the way that things have changed. Like, you you have to have a similar approach of saying, like, like trust in, like, engineering and product and design to build the products that they would wanna see Yeah. In the world. And we will build a business model around whatever that is. Yeah. Right? And it's not our job to exactly predict the future. And I think now we like, you know, we have a product roadmap that goes out max like six months because, you know, who knows how things are going to change. Right? And I think that really helps to say, okay. We're gonna create a story for, like, how the world is right now.
1:01:30 And we can expect that that story is gonna last for, like, six months from like a marketing perspective. And then it's probably going to evolve and change. And we shouldn't be afraid of being like, oh, no. We had this billboard that was like supposed to go up that had this other message, and maybe that message isn't right anymore. And it's like, yeah. You gotta just like move and change. And, you know, it's all gonna be okay. Right? But, like, you just don't have the same time to like plan a message, plan a strategy, like execute it perfectly as you used to, like, in the old world. So I think this new world, like, ultimately rewards speed and operating in a very different way. So, like, stay close to your customers. Focus would be like what I would go back. Do you think this is like a temporary moment in time, or this is just the way that it's going to be now? Like, things are just gonna continue changing, like, super, super fast. Yeah. I think it's the way that it's going to be. Yeah. I mean, I said this about the summer of twenty twenty five, where normally, like, in Linear's business, summer slows down because your customers go on vacation. Yeah. They don't buy things. They don't enter into, like, new procurement processes. Right? And and so I was expecting, like, what it was like in 2024 to be like it wasn't 2025. And I was like, this is not Yeah. This is not that. We are we are going full throttle. It used to be that companies, like, wouldn't launch products or, like, have conferences in the summer or, like, do any of these things. And now it's like, no. I'm... We're just gonna get up on, like, a stream and, like, launch something right now with, like, a week's notice.
1:03:00 Yeah. Right? And it used to matter that you, like, shouldn't launch on Fridays and shouldn't launch on Mondays. Now, who cares? You can launch on Saturday. Doesn't really matter. So just like all of those rules, I I would say, like, have gone out the window because there's so much launching and so much happening. And I don't see that changing for a while until there's like a consolidation moment in AI. Yeah. Also, like, everyone always talks about what's your moat? What's your moat like as a business? What would you say maybe the moat of Linear was like when you first started out and maybe like what it is now? Probably great design, great UX. Yeah. I would say now that it actually probably was, like, great team. And now it... I'm like, great team. Yeah. I would say speed maybe, would you say? Yeah. In a way. Just I mean, I think in a in a... It's hard because I think for a... Like, we care so much about quality. Yeah. We are often the company that will, like, delay a launch to get to, like, a certain level of quality. So we are not the flat... Fastest, but it also means we're not, like, the sloppiest Yeah. In a way. So we're hitting that fine line, basically. Right. But, like, compared to, like, legacy companies. Like, yeah, we're fast. Right? But, like, you know, it's all a spectrum. Yeah. Right? So I think in a lot of ways, like, for who we're competing against, you know, there's, like, the right trade off between, like, quality and speed. And for every company, that's different. Right? Yeah. I think we're higher, like, on the quality end, but still try to be fast and, like, cut scope, like, where it makes sense. Yeah. But I think, like, I think a lot of the moats in business, like, have kind of, like, crumbled away. Yeah. And the reality is that you just need to have a team that, like, has a really good sense, a really good pulse of like what's coming, what's changing, and you are investing in your team and your people to be able to figure it out. Yeah. Because, you know, yesterday, this model is the best. Tomorrow, this other model is the best. Like, a company you hadn't heard of, like, two months ago is now raising at $10,000,000,000. Yeah. Right. Like, so everything can change. Yeah.
1:04:30 Which means, like, we should all be very paranoid in a lot of ways because like, you know, the business that you had is not necessarily going to be, you know, the business that you have in the future. But I think what people are investing in is like a team and a way of doing things that is the future no matter what the future holds. Yeah. I I always find that question to startups is like dumb. It's like you're a startup because you're trying to beat incumbents. Like they have maybe a moat, like looking back, but if you like you don't have a moat when you just start a business. Right? It doesn't make any sense. Right. Anyway, earlier you also mentioned that like you guys were kinda like going stepwise where you were, you know, selling the startups and then you were like, oh, we can't just immediately go sell to, like, giant enterprises. We only had, like, one salesperson. Why not then just hire, you know, a 100 salespeople and just go straight for those?
1:06:00 It just... It wouldn't be the Linear philosophy or way of doing things. Yeah. Right? I think we've always operated with the, like, do more with less. Like, we would be a company with, like like, if you look at most companies that are stage and revenue, they have teams that are almost double the size in terms of headcount. And I think we are able to do a lot more with less. That also means that we can be a lot more nimble than other companies because you're not moving this, like, gigantic ship whenever you need to change direction. But I don't think we would have done well with, like, having a sales team that was bigger than, like, the entire company. You know? So, you know, I think at the beginning, I think... And this was true when I was at Stripe too. Like, there was this impression of salespeople that I think for a lot of tech companies can be pretty negative. Right? It's like, oh, like, who are these people? These, like, car salesman Yeah. Coming in to, like, sell our product. I never thought about it that way. We don't think about it that way at Linear, but that was like maybe the impression of salespeople maybe of like, you know, tech in the twenty twelves. You know? I view it very differently because I do believe that like sales has a lot of power to help inflect a business that requires human interaction Yeah. And persuasion. And I I would say that, like, for us, you know, doing it small and steady built the kind of trust that was required in different parts of the organization.
1:07:30 So like engineering really trusts sales. Like when our sales team is like having trouble in a particular segment and, you know, gets on calls with, like, PMs on our team and things like that. Like, there's a lot of trust in that salesperson really understanding the need of the customer, what they're asking for, sometimes even helping solution when necessary. I don't think he would have done that. Like, we would have been able to do that particularly well by hiring a ton of people really quickly. And again, a lot of startups take that approach, but I think they have probably more of like a hire fast, fire fast Yeah. Type of mentality. And I think with hiring, you know, we do work trials. Like, what signal does that send to you? That signal sends to you. Like, we like to, take our time and really understand if someone's a good fit. We want to get them into the, problem scope that Linear operates in and test to see, like, that their great experience somewhere else is going to translate well into our company. You know? So I think it's like a philosophy difference and, you know, every company does things differently. And I would say a lot of the great companies, you know, in the world do things their own way for a reason. Yeah. Do you guys take inspiration from any other companies?
1:09:00 I mean, I I spent seven and a half years at Stripe. And certainly, there are things I take inspiration from. I'll give an example. What... Because we were talking about hiring. You know, at a certain point, we were hiring, like, business people at Stripe faster than we were hiring, like, engineers and product folks and designers. And, you know, the founders were like, this has to come to a stop. Right? And it's like, well, you know, what kind of, you know, ratio would you like? Right? So you then take a step back and you're like, yeah. What what do we want this company to be? If we want this company to be like an engineering and product driven company, probably at least like 40% of the company should be like engineering product design, data, etcetera. And so we basically were like, okay. That's the, like, watermark. And, like, we can't It was just like an arbitrary number. Like It was 40%. 40%. And it was like, we can't go under that number. So if sales wants to hire more people, have sales help us hire engineers so that sales can then hire more people. I see. Right? And I think, like, yes, it's an arbitrary number, but I think it's important for you to understand, like, how do you want your company to be constructed? Right? Like, there are a lot of companies like... And Uber that are, like, heavy on operations, but that's because they're moving like humans and things. Right? We're a very different kind of company. We're like a software business. Right? And for us, we don't necessarily have to have, like, an operations component of the business that is, like, that huge because we're not dealing with those kinds situations. So like understand your business really well, understand what you want to be, and then figure out what model makes sense for you. So like, we don't have like that hard and fast rule at Linear of like 40%, but like I I watch that number. I like created a headcount spreadsheet. Was, like, one of the first things I did. Like, put everyone into their categories of, like, what function they're in, like, calculated the percentage, and, like, you know, I watch it.
1:10:30 And I'm, like, we really need to, like, step up our game on, like, engineering recruiting. It is always going to be harder to hire engineers than it's going to be to hire salespeople. Yeah. And so, like, what are we going to do to like make that easier for us? So a lot of it is just like the philosophy and what you want to be. And knowing that I've been at companies that like, you know, wanna do things the same way, but also, you know, for example, don't wanna outsource support or, you know, wanna treat different aspects of the product in a different way. And I think that's the challenge. There are a lot of things that like I get to bring with me from like my previous experience. And then at the same time, there are a lot of things that I wanna do in a new way that I haven't seen anyone else do. Yeah. I was gonna ask that next. Like, were there like certain things that you specifically did not wanna bring over from like Stripe or Notion? Maybe one example. I think I think Stripe did so well as a company that I think in a lot of ways when, you know, you're a 140, $160,000,000,000 company and you're like on that rise, sometimes you believe you can always hire better than what you have today. So you have a leader in a given function that's doing well. And sometimes your first reaction to that could be like, well, it's because the business is doing well, not because that person necessarily is that great. Yeah.
1:12:00 And it's It's like, oh, I'm at Google. I helped increase revenue by, you know, $20,000,000,000. Exactly. But what did you really do? Right? And and so sometimes it goes really well. Is it, you know, because of that person or is it because the business was going really well? And then if the company is doing so well that you can attract really high quality talent. Right? So someone like two levels above that person, and then a year later, you're able to attract a person who's two levels above that person. And... Right? You're kind of always going to be seeking, like, someone better than who you haven't seen because you can. And I think, like, for a lot of companies, like, being able to say, like like, I've said this to our founders. I'm like, we can get someone better. Like, we're a good company. Company that people want to work for and then are excited about. Like, we don't have to settle. Right? And I think that's, like, great. And then at the same time, sometimes you also have to, like, really recognize, like, the great people that you have. And I think for Stripe, it took a longer time to, like, come around to that. Like, Stripe's been having a lot of, like, boomerangs. Like, you know, people left the company because maybe they, you know, weren't on the rise, you know, in the way that they wanted to be. And then, you know, came back and are doing like amazingly well. And I think sometimes I would like to look back on my time at Linear and say like the right people were, like, promoted and excelled. And we didn't hire above them unless the company needed to. Right? Unless we needed to get to that a different stage. We needed something that that person didn't have. And not just because, like, we weren't willing to let them try to, like, be better than they are, to, like, scale at that next stage. Because, you know, in running startups, this is true of founders. Like, you don't... Like, almost no founder who has a lot of success has, like, done it exactly before. Right? And that's probably going to be true of a lot of the early team. Like, they're leaders at the company, but they've never scaled a support team to like 200 people or 500 people or whatever. That doesn't mean that they can't. Right? Just like it doesn't mean that that founder can't be a great founder at a thousand people and also a great founder at like 50 people. Mhmm. And so, like, I just fundamentally believe that like people can grow and scale, and that you should like reward that over time.
1:13:30 And, you know, I have some time... Like different parts of of the company, like, I'm like, I don't know if they always believe that. I I think... And I think a lot has changed since then, for what it's worth. But maybe that's one thing. On support, that would be the other thing. I think it's incredibly difficult to scale support to that degree. And I think that we emphasize a higher level of quality than I've seen in vast majority of support organizations I've operated in. And so it's something that we care about. We're willing to pay more for it. And we believe it's like the right thing to do by our customers. And I love that I've literally never in three and a half years gotten an escalation because someone had a poor experience with our support team. So that's like a company that like I see and I love to be part of. Yeah. Know, and that's not true for every other company I've worked for.
1:15:00 So, okay, you were at Stripe, then you went to Notion for a little bit, and then you became an investor. Yes. Why did you make that transition to becoming an investor? Yeah. When I was at Notion, I enjoyed my job, but the parts that I enjoyed were almost the parts that I wasn't really hired to do. So I I joined around maybe 40 people or so, and we were, you know, we were scaling. We didn't have like a single recruiter. We didn't have a head of people. We didn't have anyone working on growth. So, you know, what did I do? I, like, raised my hand and was like, I'll help hire our head of people. Like, I'll help hire our first recruiter. I'll find our first product manager and, like, hire them. And I did all of those things, and then I kind of, like, hired my way out of, like, jobs that I had. Right? And then got to a point where I was like, wait. I'm, like, back at the job. I was, like, actually hired to do. Yeah. Like, I don't really Which was what? Like, head of platform and partnerships. Okay. And I was like, I'm not really excited about this job anymore. Like, I was having more fun doing all that other stuff Yeah. That I wasn't really hired to do. But it's just like, it's company building. Mhmm. Right?
1:16:30 And at the time, I was just like, I I I think I need a new role at like a different company. And I don't know what that role exactly looks like. Right? And so I thought that venture might be a great way to kind of get the breadth that I was looking for, really, by spending time with, like, a bunch of different kinds of companies in different spaces. And, you know, I think for me, it's it's it's very different being an investor than, like, coming into a company and actually doing the work. Mhmm. It's really, like, different to, like, advise a company like, oh, you should go spend more time with your customers. And you, you know, should think about this business model or, you know And you were doing early stage? Or Early stage. Yeah. And a lot of the early stage stuff is like you're investing like pre product market fit. And like, I don't know anything about your market. Like, you sell to like, I don't know, procurement leaders or like construction workers or like like, you know, you have to be in that to really understand. And I actually felt like I was more helpful when companies had that product market fit. If I think about every single company I've ever joined, it like just was finding like product market fit, and then I joined. And then I can be a lot more helpful and like build out the business. But, you know, I think for pre product market fit, a lot of it's just like investing in the right people and like getting out of the way. Yeah. And being supportive and helpful with intros and things like that. But what I... Like, I actually really love execution. Yeah. Being inside a company and, like, helping get stuff done. And I think I'm kind of known as a person who will, like, run through walls to get things done even, you know, if that might rub people the wrong way in certain larger organizations. But, you know, for a startup, it works really well. Yeah. And so that's Are worked for
1:18:00 there certain things that you noticed, like being an operator made you a better investor at? And then also vice versa, like what did you learn from being an investor that made you better at, you know, your current role at Yeah. At Linear? I think I was speaking from, like, a place of, like, knowledge and truth. And I think a lot of investors have, like, never worked at a startup before. And there's nothing wrong with that. You can be an amazing... They're amazing investors who've never worked at a startup before, and it doesn't matter. Like, you invest in the right people. And, again, let them do their thing. And, you know, as long as you invest in the right people, that strategy works. Right? But, you know, I think also doing VC is sales. Like, you are selling money. And every other investor, like, has money to sell too. Right? So you're kind of selling the same product in some ways, but you are trying to sell yourself. Right? And, like, you are trying to differentiate yourself as a product beyond the money to that founder. And I found it much easier as someone who had, like, been operating at some great companies to be able to go to a founder and say, like, I can credibly help your business. Yeah. What would maybe other investors could.
1:19:30 What would be, like, the the pitch for an investor who's never worked at a company? It'd be, like, oh, I have a lot of intros. I've already invested in a lot of like really cool companies. Is that like kinda their selling point? I think a lot of it is like sometimes like pattern matching. Right? Like I've been on the boards of these great companies and I've seen success from zero to, you know, a 100 or whatever. It may have been several times over. And, you know, maybe I can tell you when it's time to do x or when it's time to do y because I've pattern matched my way here. I, you know, in a lot of cases, you work with certain investors and they have, like, a very large Rolodex. They know lots of great people at great companies and Yeah. You know, can introduce you to people who are more knowledgeable than they are in certain areas. So, you know, often investors will come to me and be like, can you help this founder who's team up building a sales team Yeah. You know, from zero? And, well, I can tell you about building it from one. You know? And, you know, because they know me, you know, that's like an access point that helps the founder. Right? So I think for every investor, it's like it's like something different. But, you know, what I take back now as, like, an investor, you know, going back into working at a company is, like, just having more of that understanding when I'm, like, in a board meeting, you know, of, like, what are the investors thinking? Like, oh, right. I I was an investor. Like, I I have a sense of, like, what they're thinking and what data they wanna see and what are the metrics that they care about. And if we're thinking about raising, like, a new round, like, what are my expectations for their help? If we need more customers, what are my expectations for, like, their introductions? And I have a better sense of, like, what I should be getting out of that person as an investor or board member or something like that. And I think that's, yeah, again, like helpful to the company in a lot of ways.
1:21:00 Yeah. I guess you're COO, so you have to, like, interface with investors a lot. Yeah. Yeah. Yeah. And, you know, we we work with some great investors and, you know, are very helpful and have referred people. But, know, in a lot of situations, like, closed mouths don't get fed. So if you don't ask for things, they're not gonna give them to you. Yeah. So... So Linear doesn't have any other c suite other than the founders in you? Yes. Why is that? It's a good question. I mean, I think, again, we we operated this way at Stripe for a very long time. So again, I take a bias towards it. But I think it's been true actually, for a lot of companies. Like, you might have, like, one executive who's, like, hired in on the business side who who has this title or something similar. And I would say at the beginning, a a lot of it is just, you know, my role is kind of a catchall for like a really small company for like lots of different functions and things over time. But I expect that to change. I don't think that's going to be like the way it is forever. You know, if we build out like the right team, you know, the right leadership in the right places, like people will earn that level of title. And I think for us right now, because we still feel so small, like we're about a 170 people, in a lot of ways, wanna remain pretty flat rather than, like, adding a lot of layers when we are still pretty small.
1:22:30 Yeah. And, I think that's been helpful for us so far. But that doesn't mean that we haven't, like, promoted people and given people more responsibility, more ownership, more scope. But we tend to focus on things like, you know, head of titles for, like, wide areas of ownership. Yeah. Yeah. I mean, that's all I had. Is there, like, any... Anything else that, you know, you wanted to talk about? No. No. This was a great conversation. I feel like we went into a lot of different different areas Cool. Of of of a business. So thanks. You had great questions. Yeah. Thank you.
Thanks for reading
来源:Latent Space · latent.space