【访谈】对话 Kyligence 韩卿：一个创业者的「+大模型」中场故事

年初至今OpenAI一路高歌猛进，加入这股AGI浪潮是一个「外行人」都不难做出的论断。但当一名创业者在充满挑战的环境下，真的需要做出「+大模型」的关键决策，不是说一下这么简单。

数据分析与指标平台公司Kyligence（跬智信息）就面临这样的处境。但在国内企服领域里，他们是「+大模型」实践落地上跑得最快的公司之一。 CXO UNION-CXO联盟（cxounion.cn）

Kyligence的产品能够从底层维护一套以数据驱动决策的标准化能力，让客户基于数据及指标，像搭乐高一样搭建自己的管理流程、实践。客户可以把自己的管理方法论和Kyligence的指标平台结合，构建自己的决策指挥体系。Kyligence创始人韩卿认为，这样的灵活性符合当下中国企业管理的现状。

「而AI的到来，会加速管理方法论在中国走向成熟。」韩卿表示，因为AI的理解、推理能力，能够让指标平台这套指挥管理系统运行得更好。同时，压缩了「世界知识」的AI，可以辅助管理者做出更好的选择和决策，提升其管理能力。

某种意义上，大模型与数据分析有着天然的结合点——指标。但即使大模型已经火透半边天，作为创始人的韩卿，迈出第一步其实也并不容易。

7月14日，Kyligence推出的Kyligence Copilot（AI数智助理）产品，瞄准利用大模型的理解和交互能力来帮助企业更自然、灵活、准确地用数据做决策。比如，通过AI对指标的自动分析和建议来提升管理半径；通过自然语言交互的方式赋能所有业务人员使用数据，通过Agent代替人类执行复杂任务等等。

韩卿告诉极客公园，早在今年春节甚至更早之前，他们就关注到了ChatGPT的进展，公司内部自发成立了不同的兴趣小组做一些研究和Demo。今年5月，一名「大胆」的同事指着鼻子质疑「外面AI这么火，公司到底行不行，有没有搞AI」的时候，韩卿秀出当时还很早期的原型，并在2个月后的用户大会上，现场Live Demo了他们的AI新品Copilot，并提出以AI变革组织运营及管理理念，赢得各方的认可和关注

Kyligence的故事是一个鲜明的例子：在今天的企业服务赛道，创新和适应外界变化，才是生存和成功的关键。近日，有记者和Kyligence联合创始人兼CEO韩卿进行了一场近2个小时的对话，韩卿以Kyligence的内部视角向我们讲述了过去近一年「+大模型」的历程，并分享了一系列Kyligence在推动大模型落地过程中遇到的挑战和收获的认知。

以下是此话对话的全文，由极客公园整理。 CXO UNION-CXO联盟（cxounion.cn）

一、Kylin+Intelligence，Kyligence的使命是做一只聪明的神兽

Q1：大模型可能是史上最快形成共识的一次科技变革，从Kyligence的角度，为什么决定+AI？

韩卿：我们做AI这件事情，不是跟风做的。从Kyligence成立第一天开始，一直在想怎么做智能化。今天的产业方向可能更加推崇OpenAI，但在2016年Kyligence创立时，公司名字就是Kylin，加上Intelligence，就是想把智能带到整个产品或者行业里。当时我公开讲了一个观点：把智能带给一个神兽，（让它）做一头聪明的神兽。所以今天非常坚定地投入大模型，产品做得不错是从公司第一天的使命开始的。

中间也探索了很多，比如2019年，在Kyligence用户大会上推出的AI增强引擎，用了机器学习算法，来做平台层面的自动化调优、自动化推荐。但更多是在后台平台，现在这个功能已经非常成熟，在多个客户稳定运行多年。

2021年Kyligence提出了「改变人类使用数据的习惯」的愿景。因为觉得整个数据的加工链路、使用，还是太重了。我当时打过一个比方：我们从家里到公司，可以用特斯拉自动驾驶了，但是到了公司用数据的技术还是20年前、30年前的方式方法，这肯定不对。我当时判断，肯定会发生变化，但是并不知道有大模型。

ChatGPT出来时，一开始也做了很多探索，突然发现，它符合我们的战略，就加速执行。

这一系列探索背后其实是一个理念——要做一只聪明的神兽，这是一直以来指导Kyligence的使命。我们一直在探索用什么样的方式才能提升用户体验，或者数据分析的效率。机器学习、大模型、自然语言对话、以及各种推理，这些是一脉相承的。本质上我们还是想造车，只是看到今天自动驾驶的能力好像成熟了，拿进来，变成自己的模块继续往前做。

Q2：Kyligence是Kylin+intelligence，2016年定的，这个前瞻性是怎么来的？

韩卿：对，这个渊源是我们的历史。我们最早创业时，是来自于eBay中国研发中心的大数据团队，当时做了Apache Kylin，是中国贡献到Apache软件基金会的第一个顶级项目，基于这个开源项目出来创业，这是起点。

另外，作为从业者对数据行业「痛苦」的感同身受，也是创业的初衷。外面的世界很智能化，但是数据行业做得挺痛苦。用户用起来不爽，还要天天被业务骂等等。当时判断，整个数据服务，必然需要一些自动化，让整个平台变得更加Smart，就取了Intelligence，这个初心一直没变，而且越做越好了。

Q3：让数据使用，变成普通大众都能使用的能力。你认为AI可以加速Kyligence想要解决的问题。

韩卿：对，我很兴奋。用得越多，越发现AI的技术变革对Kyligence整个战略非常利好。这种非常自然的语言对话，能解决一个核心问题——平权。以前使用数据分析做决策，是企业里一些专业人士甚至老板们的特权，只有他们有这个知识储备，只给他们配备专业的数据分析团队。但AI让数据使用这件事情，变成了普通大众都能使用的能力，（未来）每个人都可以跟机器对话，都有一个私人的数字助理。这正是我们改变使用数据的习惯这个大方向。

Q4：在「改变人类使用数据的习惯」的脉络上，Kyligence产品是怎么演进的？

韩卿：最早2016年到2019年底，我们从原先的开源框架，变成了一个商业化产品，验证了在中国卖一个标准化的软件产品，按年、按订阅收费的商业模式。

从2019年年底到去年，差不多三到四年，是比较痛苦的阶段，可以说，跨越鸿沟。Kyligence原来的产品非常专业，只能卖给专业的客户，像OLAP等，当时客户基本上都是大型银行、零售、制造等行业企业。当拓展更多客户时，就遇到了阻碍。因为专业的东西理解起来挺难，采用起来挺难，费用也挺贵。大量中腰部以上的客户想要一个开箱即用的产品来解决问题。

我们从客户侧琢磨后，做成了指标平台的产品线，在整个过程中，花了很大力气。因为原来在做一些底层的Infra、服务大型客户，整个公司产研的加工模式比较重。当需要一些更加轻量产品的时候，它就不太适应了，坦率说这里走了蛮多弯路，直到去年年底才开始走顺。

今年开始「AI革命」。真正动手开始把AI带来的更多可能性放到Kyligence产品里，是从今年四五月份开始。到现在，即使只有短短半年多的时间，产品本身、公司定位、战略等发生了翻天覆地的变化，但不是说用AI彻底做一个新产品，而是在原来的基础上。 CXO UNION-CXO联盟（cxounion.cn）

可以说，第一阶段是造重型发动机，只能卖给重型企业，卖的是专业的产品。第二阶段从做发动机变成做车了，卖的是量产轿车。今年AI出来之后，我们配了自动驾驶模块，使整个车更好卖，而且现在需求量看上去非常旺盛。

二、+大模型的抉择时刻

Q1：哪个瞬间，让你觉得大模型就是Kyligence一直要找的钥匙？

韩卿：ChatGPT出来到今年春节那段时间，外界都很焦虑，我觉得先静观其变。当时，Kyligence研发自发组织了一个AI兴趣小组，做了一个非常原始的原型，想在产品里增加一些自然语言搜索的能力。当他们要发布这个功能的前一天晚上被我阻止了，我觉得这个原型的功能还拿不出手。那个时候大家都抱怨，怎么最后一个时刻被叫停了？

转变发生在今年5月12日，当时组织Kyligence所有的销售、售前在上海进行今年第一次年度培训。傍晚结束之前我做总结和大家聊了聊，当时有人challenge我说，「外面AI已经这么火了，我们公司行不行？有没有在做AI？有没有这个东西？」

我的脾气有点不服输，觉得：那不行，我得给你们看看。我把兴趣小组做的原型拿出来给大家秀了一下。但坦率讲，当时大家看到结果还非常粗糙，觉得也就那样，可我心里却被触动了。

5月12日马上拉着产品等团队组了一支特攻队，整个公司在那段时间突击了两个月，到7月14日发布会，我给现场一千多人做Live Demo。很多朋友跟我说，今年看了这么多AI的发布会，就你做Live Demo，胆子也挺大。

Q2：什么契机下成立的兴趣小组？

韩卿：公司内部学习氛围挺强，几乎每周都会有技术分享会。当时，是由研发团队自发成立的兴趣小组。他们一方面研究大模型本身的能力、技术，一方面也在想怎么把它吸收到产品里。也会从另外一个角度思考，大模型出来之后，我们的产品是不是就没有价值了？当时做了一些研究，微软的GitHub Copilot、Excel Copilot、Windows Copilot等等，给了我们很多灵感。

兴趣小组不是专职研究，本职工作都挺忙，只是凭兴趣，把这个事情一点一点做成了。最早就是一个玩具，看到价值点后，把它变成了一个产品。 CXO UNION-CXO联盟（cxounion.cn）

Q3：从5月到现在，Kyligence AI战略落地为什么这么快？

韩卿：真的是靠团队，感谢团队。这两年，Kyligence在组织能力建设上花了非常多心血，我一直在强调企业未来的竞争是组织能力的竞争，所以今天在整个战略方法论下往前冲的时候，能够有组织保障能力。

早些年大家一腔热血，一个技术就可以往前冲，做事情。但是现在发现，技术的进步太大了、太快了，如果只是抱着一个技术，当成壁垒，很有可能这个技术本身就会被替换掉。

第二，市场变化也很快，怎么能够快速变成一套体系，接得住市场前方传回来的需求，快速把后方研究的特色能力推出去。就像我们今年做的Copilot的AI能力，是后方兴趣小组做的，我们可以通过产品化能力快速实现，通过市场能力快速推出去，这是组织能力的体现。不是一个人或几个人的决策和能力，而是变成组织性的东西。这样，未来有新的技术和业务出现，也不担心抓不住机会。

Q4：今年在强化AI战略加大模型的过程中，最印象深刻的事情是什么？

韩卿：就是7月14日的用户大会上做Kyligence Copilot的Live Demo，我在台上很紧张，团队研发、产品在台下也很紧张。发完的那一刻，成了，能够看得到实打实的效果，这给我带来一个震撼。

尤其在今年，整个经济周期和企业服务赛道预期降低的情况下，Kyligence Copilot的发布给团队、客户和行业都打了个气，来的非常及时。毕竟，做软件产品的公司最终其实反映在客户层面，才是真正的根，产品做出来获得了大家认可，就很有信心，这对我触动很深。 CXO UNION-CXO联盟（cxounion.cn）

三、数据就在那里，AI解决了信息获取通道的问题

1. AI让「管理」变成「赋能」

Q1：+AI的半年时间里，Kyligence发生了翻天覆地的变化，哪些变化？

韩卿：整个模式发生了变化，产品的生命力、竞争力、差异化一下子显现出来，这是我非常兴奋的原因。

在客户层面，我们最早5月份拿原型与客户交流时，你会发现所有客户都在寻找类似的方向去突破，他们觉得你的场景很好、落地性很好。像这样客户给到的输入，叫市场的味道，背后是需求。

第二个变化是产品背后的方法论。AI的到来，能够大大增加老板们的管理半径，使管理的中间成本更低，管理效率更高。这对公司治理非常重要，因为站在公司层面，指挥要快，落地要准。

第三，AI让管理变成赋能。以前上系统后，一线员工的感受是，「你又搞一个系统来管我，来扣我奖金」，员工会选择消极怠工，填假信息。这是人性，管理就是反着人性来的。AI带来的东西反而非常积极，因为它是赋能，而人都是上进的。 CXO UNION-CXO联盟（cxounion.cn）

AI不仅仅告诉你KPI或指标做得不好、有风险，甚至建议你怎么做，帮助你做得更好，获得更多年终奖。通过一些技术工程，比如Kyligence正在做的外挂知识库，AI可以把更精准、更好用的建议给到员工。

Q2：Kyligence Copilot推出后，B端客户有没有对大模型幻觉等问题的担忧？

韩卿：我们其实没有什么幻觉问题，因为场景非常特定，在这个特定范围内，确保AI在有限空间里回答。

第二，所有数据、数值的计算全部是Kyligence平台自己做的，所以我们对于大模型的依赖降得很低，这是我们在使用大模型时的特色，或者最佳实践。把对大模型的依赖降到了最低，只有意图理解，找一个「弱智」一点，达到高中生水平的大模型也能用，只要它理解我想干什么，翻译成我要干的事情就行了。

在一些生成场景下，比如写总结，这是大模型擅长的事情，但是数据计算、对比等等，我们自己做，不依赖它，否则它会出现1+1=3。

Q3：Kyligence Copilot的推出，是否给Kyligence带来了新客户？

韩卿：Copilot一定是跟指标平台绑定在一起，Kyligence真正的门槛是指标平台背后的OLAP引擎。但，Copilot也给业务带来了三大变化。 CXO UNION-CXO联盟（cxounion.cn）

一个是老客户扩容，老客户看重Copilot能力，希望快速导入。Kyligence的优势在于头部客户，转换效率很高，愿意付费。

第二是拓宽了客户群，原先我们专注在金融一些大型制造行业和头部零售客户，主要是大客户。Kyligence把AI产品化后，现在能够接触到一些中小型客户，获得「意料之外」的客户。

第三是销售周期变短，销售的搭配组合更丰富，客单价更高。例如最近与智谱、百川在探讨一种bundle（组合销售），需求很旺盛，甚至大客户也希望尽快部署。

2. 人与人之间的不信任感，远远超过人和机器之间的不信任感

Q1：为什么AI使得管理半径更长了？

韩卿：比如现在我想知道风险预警，有10位高管汇报给我，每个高管至少要背几个KPI，KPI往下还有很多过程指标，加起来几十、几百个指标，甚至更多。每位高管过一遍状况，再跑一遍数据，再层层汇报，可能半天就过去了。西方管理哲学讲，一个人只能管7个人，因为没有时间与精力再管更多，这就是管理半径。

但现在，我可以很简单地生成AI报告和总结，知道大的方向。最重要的是，人机交互解决了一个问题：人与机器的交互更加信任，人与人之间会更复杂。 CXO UNION-CXO联盟（cxounion.cn）

举个例子，当我问销售他的客户情况，我仅仅是想知道具体情况，但销售心里的第一反应可能是：老板为什么直接找我？老板到底想问什么？他可能会说，昨天刚刚拜访之类的话，证明自己很努力，但我并不想问这个。中层的高管也会觉得不对劲，觉得「老板你怎么跳过我了？」类似这种人与人之间的复杂性太烦了。

因此，通过人和机器的交互代替人与人之间的交互，通过一个AI系统来做交互，情况会好很多，加快了信息获取的速度，管理半径就会大起来。

Q2：AI解决了这件事情，它可以直接下钻很多层，想要什么样的信息都有。

韩卿：钻取信息其实是指标平台本来就有的，与AI无关。但AI能够帮助我做总结、交互对话等等，使我的效率更高。其实数据本身一直在那里，与AI没关，但AI解决了信息获取的通道问题。

四、AI会加速管理方法论在中国落地

1. Kyligence的核心是一套管理方法论

Q1：Kyligence在数据分析大的品类下，聚焦做数据指标，这会带来哪些独特性？

韩卿：不同阶段，我们的聚焦点和差异点其实不一样。在做开源Kylin时，当时最大的卖点是在超大数据集上提供高并发和高性能的分析，比如说PB级以上的数据集，我们能查出来，还能支撑一些比较高的并发。

当我们做整个企业级的OLAP工作时，高可用、安全、多租户、资源隔离，这些金融级特性是当时的竞争优势和壁垒。只做技术还不够，后来也做了一些创新，比如AI增强引擎，资源的自动分配等，通过智能化开始打磨新的差异化。

到做数据指标平台的时候，我们非常强调管理方法论。我们认为，指标远非一个简简单单的技术，一个数据点，本质上反映的是一个公司的管理方法以及这个方法论的实现工具，当人人都产生数据时，我们会强调用PDCA等的方式，通过价值指标、价值树的形式完成管理和决策帮助。尤其在国内，Kyligence希望从老板到员工，都能够清晰地理解整个指标情况。 CXO UNION-CXO联盟（cxounion.cn）

今天也一样，AI进来之后，你会发现Copilot或者说ChatBI大家都能做，但是特点不一样。Kyligence基于指标的这套系统，一直在探索如何能让AI帮你做一些分析、运营、管理，而不是简单的数据查询。比如，Kyligence目标分解或者目标评估的功能，可以直接帮你评估KPI、提供建议，帮你提升业绩、竞争能力。

Q2：Kyligence指标平台的核心竞争力，是打捞行业背后的最佳实践，或者说管理方法。

韩卿：对，叫管理方法，不是实践，最佳实践往往会变成项目。我觉得，软件本质上是一个管理方法论的工具，为什么要ERP？为什么要这套系统而不是那套？背后都是一套管理逻辑、方法论，这才是核心的核心。否则，做出来的系统要么变成项目，要么变成一个没法复制的东西。

另外，中国公司的管理方法论和美国公司的管理方法论其实不一样，这很明显。但我认为，现在到了中国公司开始创立自己管理方法论的时间节点。过去这么多年，我们都在学习西方先进企业的方法，但是回过头来，中国的文化、哲学不一样，相应地，管理方法论也完全不一样。 CXO UNION-CXO联盟（cxounion.cn）

我认为中国的软件未来应该有一套自己独特的东西，背后的根是中国的哲学，中国的管理方法论，没有这个管理方法论，没有办法指导软件。

Q3：软件承载了这套方法论，让客户的管理越来越好。但也有一种观点是，中国没有一套通用的管理方法论，各行各业、各家公司的方法可能不一样，你怎么看？

韩卿：创立Kyligence这些年，在市场一线跟很多客户交流后，我认为，美国不一样的地方是它先有管理（再有软件），已经过了我们（现在经历的）草莽阶段了。从德鲁克开始建立了现代公司治理的方法论，各种商学院、大公司已经沉淀出他们那个社会通用的最佳实践、或者理论基础，符合西方文化、哲学的方式方法。所以每个公司的管理逻辑和东西是一致的。

但是在中国不一样，中国用了40年的改革开放，干了人家200年的事情，所以在这个过程当中，成功企业有各自的方法，没有一样的。

第二，中国公司的老板都是拼出来的，很有狼性，同时有一个特点——谁也不服谁，「你跑过来跟我讲这个公司怎么管？你懂不懂？」不好意思，稍微有点成就的老板EGO都很大。但同时，你发现所有的老板都在学习管理学。因为大家都知道都没管好，需要更好的东西。于是，大家都急于求成，买各种各样的方法论，买各种各样的软件，最后变成「差生文具多，搞了一堆，然后发现没啥用。 CXO UNION-CXO联盟（cxounion.cn）

Q4：如果中国在形成一套通用的管理方法上，时机还未成熟，你却这么做产品，会不会走弯路？

韩卿：在这种情况下，我认为，中国的软件可以有一个不同的，符合当下中国企业现状的方式——底层维护一套以数据驱动的标准化能力，上层像搭乐高一样，拼出不同的管理能力。

也就是说，如果把整个体系和系统底层做得足够灵活，让每个老板能够把他的管理方法论和管理哲学用这个平台承载起来，这是可行的。这个平台不需要变成一个非常复杂的系统，「来教你干什么事情」，而是一个很好的工具，「你」一定有自己的管理方法论，用这套工具来承载，它帮你想要的东西推下去，因为指标平台本质是一个指挥系统。

Kyligence整个底层平台非常灵活，既能够适应不同的管理方法论，但是底层的数据、指标，又吸纳了西方数据治理，用数据做决策的核心逻辑。这样，是有机会闯出一条有特色的路，既能保证Kyligence底层产品的标准化，又能够适应不同公司管理方法论的差异化。

Q5：怎么做到这么灵活？PaaS化要集纳的共性需求也非常多。

韩卿：抓核心。一般来说，一家公司应该有三套系统。第一套系统是以ERP为核心的生产制造系统，解决「生产」的问题。第二个系统是以CRM为核心的营销系统，解决「卖出去」的问题。

第三套系统，以指标平台为核心的指挥决策系统。今天业务进展情况到底如何？管理意志和意图是不是被有效贯彻落地？如何更早地干预过程，来确保最终绩效能够达成。也就是说，管理指挥系统，是以指标平台为核心的。

因为每个公司最终都是为财务指标服务，从财务层面，都能被拆解到原子指标的程度。只要抓住这些不变量，在系统里沉淀好底层数据指标，保持数据定义的标准口径一致。至于上层怎么组合，更关心营销还是成本，其实是不同管理思路、方法的串联。这样一来，可以解决灵活性的问题。 CXO UNION-CXO联盟（cxounion.cn）

并且，AI能够让指挥管理系统运行得更好。AI通过强大的计算、推理能力，及时告知最新的整体状况，甚至是一些平时不会留意的指标洞察。这个时候，我认为AI会对整个管理能力、管理决策、指挥决策，带来巨大的帮助。

2. AI+指标平台，让管理方法论走向成熟

Q1：AI可能会加速管理方法论的成熟。

韩卿：AI绝对可以加速。中国企业发展飞快，但缺少善于管理的人才。

不同于美国的职业经理人，中国企业发展太快了，大部分管理者都是临时被提拔起来的，怎么管人、怎么开人、怎么管成本都没做过，找不到经验丰富的，包括我自己也一样。这种情况下，讲什么管理方法论，讲学院派的东西，根本没有用。说得难听一点，有时候「管理」就变成了各种「斗争」。

我在公司讲，最需要被管理的是管理者本身。不可能把每个管理者送去读MBA，没这个钱也没这个时间。但我突然发现，AI好像可以做这个事情，助力每位管理者具备更好的管理经验。

比如，我们公司把所有高管的指标都做成了一个指标树，每周一AI都会帮着把这个指标树跑一个总结报告，给到整个管理层，但是内容加在一起又很多。我可以让AI给我总结成不超过500字的东西，它可以直接告诉我风险指标在哪里，应该在什么地方加强管理。所以AI能够赋能管理层。

Q2：像指标平台融合Copilot，听起来对管理者特别友好，通过AI，管理者可以有更强的掌控感。对员工呢？

韩卿：员工赋能才是我看重的，我们设计的指标平台和Copilot不是卖给专业用户的。对于公司而言，能使用数据能力的人不超过10%-15%是特权，通过Kyligence Zen指标平台加Copilot让其他所有人有机会通过自然语言获取数据，这样员工就被赋能了。 CXO UNION-CXO联盟（cxounion.cn）

公司的业绩一定是靠一线员工，他们是真正需要用数据的人，老板看数据从来不是问题。今天AI+指标平台，能够提供人人可用的数据服务和能力，你会发现整个公司的数据素养就起来了。

另外，赋能员工还体现在，AI可以不知疲倦，并且AI具有海量的知识推理能力，他的知识面宽度远远大于单个人类，给到你的参考远远大于你所能看到的。例如，我们产品里有一个叫归因分析的功能，甚至数据分析师可能都不一定掌握很好这种分析函数，但今天可以通过问AI，它马上会帮忙归因，做总结，这就是赋能。

Q3：未来公司会给每个人都配一个Copilot吗？

韩卿：我相信未来公司肯定会给每个人都配一个指标的Copilot或数据的Copilot，尤其中国一线的企业或民企。公司的管理总是需要一些数据或指标来体现，如果我不是给你一个工具来监控你，而是给你一个真正的助理赋能你，不管对于公司管理层还是业务人员，这都是价值。

五、大模型时代的合作与竞争

Q1：上个月，飞书7的发布会上，Kyligence作为独家北极星指标的合作伙伴出现，这个决策是怎么推动的？

韩卿：我们首先是飞书的重度用户，因为我们内部一直推崇管理的可观测性，也将飞书应用得很极致。这次与飞书智能伙伴的合作，也是借飞书的超级流量入口，把Kyligence的产品推向更多人使用。

与飞书的合作可以说一拍即合，飞书让大家更多集中在平台、办公协同等方面，而我们在做数据指标、北极星管理等方面更有优势。

Q2：在你看来，大模型的到来会让生态之间的合作会变得更加频繁吗？

韩卿：我觉得会。这些东西带来的变化在于，世界发展太快。大模型进来之后，每个公司更加焦虑了，要么用AI重塑自己，要么被AI替换掉，这件事情很明显。这个时候大家聊下来发现，以前没什么东西的时候，你做我的我做你的，现在不如合作。 CXO UNION-CXO联盟（cxounion.cn）

Q3：年底大家都在定战略，Kyligence在AI方面，整体的产品布局规划是什么样的？

韩卿：第一个是聚焦，聚焦以指标平台为核心，把AI的东西更深入地放进去。其实有一些客户跟我说能不能做一些类似知识问答的大模型能力或者相关的，但这不是我们想要做的。

第二个是技术的迭代，我最近一直在问技术团队一个话题，「你们觉得AI时代的数据仓库应该是什么样，AI时代的数据分析应该是什么样的」，希望大家不断思考根本性的问题，这个地方很可能是一些颠覆性的东西，一定要深入思考，因为AI带来的变化实在太快了。

第三个是更加快速地拓展我们的商业，我们产品的成熟度非常高，现在已有多家客户已经付费上线了，明年是一个可以做快速增长的年份，能够更快速地跑马圈地、占领用户心智。

Q4：今年创业和往年是不是有一些不一样？感受以及市场，有什么样的不一样？

韩卿：兴奋是很兴奋，但是很难。最近看了一个分析，「剩者为王」，剩下的剩。

我认为，整个中国企业服务进入了非常低的低谷期。但是我坚信中国未来会进入一个中国软件的黄金时代，可能这就是黎明前的黑暗，一般来说天亮前总是更暗，但是熬过去，接下来就是一个非常巨大的的爆发机会，这也是我们冲在第一线的感触，就看怎么布局，希望大家一起在春天绽放。 CXO UNION-CXO联盟（cxounion.cn）

翻译：

Dialogue with Kyligence Han Qing: An entrepreneur’s “+ big model” midfield story

OpenAI has been on a roll since the beginning of the year, and joining the AGI wave is a statement that is not difficult for a “layman” to make. But when an entrepreneur in a challenging environment, really need to make a “+ big model” key decision, not so simple to say.

Kyligence (跬), a data analytics and metrics platform, is in such a situation. However, in the field of domestic enterprise services, they are one of the fastest companies in the practice of “+ large model”.

Kyligence’s products maintain a standardized set of data-driven decision making capabilities from the ground up, allowing customers to build their own management processes and practices based on data and metrics like Lego. Customers can combine their own management methodology with Kyligence’s metrics platform to build their own decision command system. According to Han Qing, founder of Kyligence, such flexibility is in line with the current state of business management in China.

“The arrival of AI will accelerate the maturity of management methodology in China.” Han Qing said that because of AI’s understanding and reasoning ability, it can make the command and management system of the index platform run better. At the same time, AI that compresses “world knowledge” can assist managers to make better choices and decisions, and improve their management capabilities.

In a sense, there is a natural combination of large models and data analysis – indicators. But even if the big model has been burning through half the sky, as the founder of Han Qing, taking the first step is not easy.

On July 14, Kyligence launched the Kyligence Copilot (AI data intelligence Assistant) product, which aims to use the understanding and interaction ability of large models to help enterprises make decisions with data more naturally, flexibly and accurately. For example, improve the management radius through AI’s automatic analysis and recommendation of indicators; Through natural language interaction, all business personnel are empowered to use data, and agents are used to perform complex tasks instead of humans.

Han Qing told Geek Park that as early as this Spring Festival or even before, they paid attention to the progress of ChatGPT, and the company spontaneously set up different interest groups to do some research and Demo. In May this year, when a “bold” colleague pointed his nose and questioned “outside AI is so hot, whether the company can do it or not, whether it has AI”, Han Qing showed a very early prototype at the time, and at the user conference two months later, Live Demo of their new AI Copilot, and proposed to change the organization’s operation and management concept with AI. Win recognition and attention from all sides CXO UNION-CXO联盟（cxounion.cn）

Kyligence’s story is a stark example of how innovation and adaptation are key to survival and success in today’s corporate services landscape. Recently, Geek Park and Kyligence co-founder and CEO Han Qing had a nearly 2-hour dialogue, Han Qing told us about the past year of “+ big model” from Kyligence’s internal perspective, and shared a series of Kyligence in the process of promoting the big model encountered challenges and gains cognition.

Here is the full text of that conversation, compiled by Geek Park.

1. Kylin+Intelligence, Kyligence’s mission is to be an intelligent god beast

Q1: The big model is probably the fastest consensus in history. From Kyligence’s perspective, why did you decide to +AI?

Han Qing: We do AI this thing, not to follow the trend. Since the first day of Kyligence’s establishment, I have been thinking about how to make it intelligent. Today’s industry direction may favor OpenAI, but when Kyligence was founded in 2016, the company’s name was Kylin, and Intelligence was intended to bring intelligence to an entire product or industry. At that time, I publicly stated a point of view: to bring intelligence to a divine animal, [let it] be a wise divine animal. So today’s very strong commitment to the big model, the product doing well starts with the mission of the company on day one.

The middle has also explored a lot, such as the AI enhancement engine launched at the Kyligence User Conference in 2019, which uses machine learning algorithms to do automatic tuning and automatic recommendation at the platform level. But more in the background platform, now this feature is very mature, in multiple customers stable operation for many years.

In 2021, Kyligence set out a vision to “change the way humans use data.” Because I feel that the processing link and use of the entire data are still too heavy. I made an analogy at that time: we can use Tesla autopilot from home to the company, but to the company to use data technology is still 20 years ago, 30 years ago, which is definitely not right. I judged that it was bound to change, but I didn’t know there was a big model. CXO UNION-CXO联盟（cxounion.cn）

When ChatGPT came out, we did a lot of exploration at first, and we suddenly realized that it fit our strategy, and we accelerated our execution.

Behind this series of explorations is the idea of being an intelligent divine beast, which has always guided Kyligence’s mission. We are always exploring ways to improve the user experience or the efficiency of data analysis. Machine learning, large models, natural language dialogue, and all kinds of reasoning are all in the same vein. In essence, we still want to build cars, but we just see that the ability of autonomous driving today seems to mature, take it in, turn it into its own module and continue to do it.

Q2: Kyligence was established by Kylin+intelligence in 2016. How did this foresight come about?

Han Qing: Yes, this origin is our history. When we first started our business, we came from the big data team of eBay China R&D Center. At that time, we made Apache Kylin, which was the first top-level project contributed by China to the Apache Software Foundation. Based on this open source project, we started our business.

In addition, as a practitioner of the data industry “pain” empathy, is also the original intention of entrepreneurship. The world out there is smart, but the data industry is miserable. Users are not happy to use, but also be scolded by the business every day and so on. At that time, the judgment was that the whole data service must need some automation to make the whole platform more Smart, so we took Intelligence, and this original intention has not changed, and it is getting better and better.

Image Artificial intelligence operation Excel | Source: DALL·E

Q3: The ability to make data use accessible to the general public. You think AI can accelerate the problem Kyligence is trying to solve.

Han Qing: Yes, I’m very excited. The more we use it, the more we find that the technological change in AI is very good for Kyligence’s overall strategy. This very natural language dialogue addresses a core issue – equality. In the past, using data analysis to make decisions was the privilege of some professionals and even bosses in the enterprise, only they had this knowledge reserve, and only they were equipped with professional data analysis teams. But AI makes the use of data accessible to the general public, and everyone can talk to machines and have a personal digital assistant. This is the general direction of changing the way we use data. CXO UNION-CXO联盟（cxounion.cn）

Q4: How does Kyligence evolve in the context of “changing people’s habits of using data”?

Han Qing: From 2016 to the end of 2019 at the earliest, we changed from the original open source framework to a commercial product, verifying the business model of selling a standardized software product in China, charging per year and per subscription.

From the end of 2019 to last year, almost three to four years, is the more painful phase, so to speak, across the chasm. Kyligence’s original products were very professional and could only be sold to professional customers, such as OLAP, who were basically large banks, retail, manufacturing and other industries. When it came to reaching more customers, it hit a roadblock. Because professional things are very difficult to understand, very difficult to use, and very expensive. A large number of customers from the mid-waist up want an out-of-the-box product to solve a problem.

After thinking from the customer side, we made the product line of the indicator platform, and spent a lot of effort in the whole process. Because I was doing some low-level Infra and serving large customers, the processing mode of the whole company’s production and research was relatively heavy. When the need for some more lightweight products, it is not quite adapted to, frankly speaking, there are quite a few detours, until the end of last year began to go smoothly.

The “AI revolution” began this year. The real start to put more possibilities brought by AI into Kyligence products began in April and May this year. Until now, even in just over half a year, the product itself, company positioning, strategy, etc., have undergone earth-shaking changes, but it is not to say that AI is used to completely make a new product, but on the basis of the original.

It can be said that the first stage is to build heavy engines, which can only be sold to heavy enterprises and sell professional products. The second stage changed from making engines to making cars, selling mass production cars. After the AI came out this year, we equipped the autonomous driving module to make the whole car more sellable, and now the demand seems to be very high.

2. + the decision moment of the large model

Q1: What moment made you think that the big model was the key Kyligence had been looking for?

Han Qing: Between the time ChatGPT came out and the Spring Festival this year, the outside world was very anxious. I think we should wait and see what happens. At that time, Kyligence R&D spontaneously organized an AI interest group and made a very primitive prototype to add some natural language search capabilities to the product. When they were about to release the feature the night before I stopped them, I didn’t think the prototype was ready yet. At that time, everyone complained, why was the last moment stopped?

The change took place on May 12 this year, when all of Kyligence’s sales and pre-sales were organized in Shanghai for the first annual training of the year. Before the end of the evening, I made a summary and chatted with everyone, when someone challenged me and said, “AI has been so hot outside, is our company OK? Is it doing AI? Is there this thing?”

My temper is a little unconvinced, think: that can’t, I have to show you. I showed you the prototype that the interest group made. But frankly, at that time, everyone saw that the results were still very rough and thought that it was just like that, but my heart was touched.

On May 12, immediately led the product team to set up a special attack team, and the whole company attacked for two months during that time. By July 14, I had made a Live Demo for more than 1,000 people on the scene. Many friends told me that I have seen so many AI conferences this year, and you are brave enough to do Live Demo. CXO UNION-CXO联盟（cxounion.cn）

Q2: Under what circumstances was the interest group established?

Han Qing: The learning atmosphere within the company is very strong, and there are technology sharing meetings almost every week. At that time, it was a spontaneous interest group formed by the research and development team. On the one hand, they study the capability and technology of the large model itself, and on the other hand, they also think about how to absorb it into the product. Will also think from another Angle, after the large model comes out, our product is not valuable? We did some research at that time, and Microsoft’s GitHub Copilot, Excel Copilot, Windows Copilot, etc., gave us a lot of inspiration.

Interest groups are not full-time research, their work is very busy, just with interest, to make this thing little by little. It started as a toy, saw the value point, and turned it into a product.

Q3: From May to now, why is the Kyligence AI strategy landing so quickly?

Han Qing: It really depends on the team, thank the team. In the past two years, Kyligence has spent a lot of effort on organizational capacity building. I have always emphasized that the future competition of enterprises is the competition of organizational capacity. Therefore, when we move forward under the whole strategic methodology, we can ensure the organizational capacity.

In the early years, everyone was passionate, and a technology could rush forward and do things. But now it is found that the progress of technology is too big, too fast, if you just hold a technology as a barrier, it is likely that the technology itself will be replaced.

Second, the market is also changing quickly, how can it quickly become a system, catch the demand from the front of the market, and quickly push out the characteristics of the rear research. Just like the AI capability of Copilot we did this year, which was done by the rear interest group, we can quickly realize it through productization capability and quickly launch it through market capability, which is the embodiment of organizational capability. Not the decisions and abilities of one person or a few people, but something that becomes organizational. In this way, there will be new technologies and businesses in the future, and there will be no worry about seizing opportunities.

Q4: What was the most impressive thing in the process of strengthening the AI strategy scaling model this year?

Han Qing: I did the Live Demo of Kyligence Copilot at the user conference on July 14. I was very nervous on the stage, and the team’s research and development and products were also very nervous under the stage. The moment I finished, I was able to see the real effect, which gave me a shock.

On July 14, 2023, Han Qing did a live demo demo at the Kyligence User Conference

Especially this year, when expectations throughout the economic cycle and the enterprise services circuit were lowered, the release of Kyligence Copilot came at a very timely time for the team, customers and the industry. After all, the company that does software products is ultimately reflected in the customer level, which is the real root, and the product has been recognized by everyone, and it is very confident, which touched me deeply. CXO UNION-CXO联盟（cxounion.cn）

3. The data is there, and AI solves the problem of access to information

1. AI turns “management” into “empowerment”

Q1: In the past six months of +AI, Kyligence has undergone earth-shaking changes, what changes?

Han Qing: The whole model has changed, and the vitality, competitiveness and differentiation of the product have suddenly appeared, which is the reason why I am very excited.

At the customer level, when we take the prototype and communicate with customers as early as May, you will find that all customers are looking for a similar direction to break through, and they think your scene is very good and the landing is very good. The input given by customers like this is called the taste of the market, and behind it is the demand.

The second change is the methodology behind the product. The arrival of AI can greatly increase the management radius of bosses, so that the intermediate cost of management is lower and the management efficiency is higher. This is very important for corporate governance, because standing at the company level, the command must be fast and the landing must be accurate.

Third, AI enables management. After the previous system, the feeling of frontline employees is, “You have a system to manage me and deduct my bonus”, and employees will choose to be passive and fill in false information. This is human nature, management is against human nature. What AI brings is very positive, because it is empowering, and people are motivated.

AI doesn’t just tell you that KPIs or indicators are bad or risky, it even suggests what you can do to help you do better and get more year-end bonuses. With some technical engineering, such as the plug-in knowledge base that Kyligence is working on, AI can deliver more accurate and useful suggestions to employees. CXO UNION-CXO联盟（cxounion.cn）

Q2: After the launch of Kyligence Copilot, do B-side customers have concerns about the illusion of large models?

Han Qing: We don’t really have any illusion problem, because the scene is very specific, within this specific range, to ensure that the AI can answer in a limited space.

Second, all the data and numerical calculations are done by the Kyligence platform itself, so our reliance on large models is very low, which is our characteristic, or best practice, when using large models. Keep the reliance on the big model to a minimum, only intention understanding, and find a “retarded”, high school level big model can also work, as long as it understands what I want to do, translated into what I want to do.

In some generation scenarios, such as writing summaries, this is something that large models are good at, but data calculation, comparison, etc., we do it ourselves, do not rely on it, otherwise it will appear 1+1=3.

Q3: Has the launch of Kyligence Copilot brought new customers to Kyligence?

Han Qing: Copilot must be tied to the indicator platform, and the real threshold for Kyligence is the OLAP engine behind the indicator platform. But Copilot has also brought three major changes to the business. CXO UNION-CXO联盟（cxounion.cn）

One is the expansion of old customers, old customers value Copilot capabilities and want to quickly import. Kyligence’s strength lies in the head customer, the conversion efficiency is high, and the willingness to pay.

The second is to broaden the customer base, originally we focused on some large manufacturing industry and head retail customers in finance, mainly large customers. By productizing AI, Kyligence is now able to reach some small and medium-sized customers and gain “unexpected” customers.

The third is that the sales cycle is shorter, the sales mix is richer, and the customer price is higher. For example, recently, we are discussing a bundle (combined sales) with Zhipu and Baichuan, and the demand is very strong, and even large customers want to deploy as soon as possible.

2. Distrust between people is far greater than distrust between humans and machines

Q1: Why does AI make the management radius longer?

Han Qing: For example, now I want to know the risk warning, there are 10 senior executives reported to me, each executive must memorize at least a few KPIs, KPI down there are many process indicators, add up to dozens, hundreds of indicators, or even more. Each executive goes through the situation, runs the data again, and reports the layers again, and it may be half a day. Western management philosophy says that a person can only manage seven people, because there is no time and energy to manage more, which is the management radius. CXO UNION-CXO联盟（cxounion.cn）

But now, I can easily generate AI reports and summaries and know the big picture. Most importantly, human-computer interaction solves a problem: human-machine interactions become more trusting, and person-to-person interactions become more complex.

For example, when I ask a salesperson about his clients, I just want to know the details, but the salesperson’s first reaction may be: Why did the boss come to me directly? What does the boss want to ask? He might say something like, “Just visited yesterday,” to prove he’s working hard, but I don’t want to ask that. Mid-level executives also feel something is wrong, thinking, “Boss, why did you skip me?” This kind of human complexity is annoying.

Therefore, through human and machine interaction instead of human interaction, through an AI system to do interaction, the situation will be much better, accelerate the speed of information acquisition, management radius will be large.

Q2: AI solves this problem, it can drill down many layers directly, and it has what kind of information it wants.

Han Qing: Drilling information is actually an indicator platform that has nothing to do with AI. But AI can help me with summaries, interactive conversations, and so on, making me more productive. In fact, the data itself has always been there, and it has nothing to do with AI, but AI has solved the problem of access to information. CXO UNION-CXO联盟（cxounion.cn）

4. AI will accelerate the implementation of management methodology in China

1. The core of Kyligence is a management methodology

Q1: Kyligence focuses on data indicators under the category of data analysis. What uniqueness will this bring?

Han Qing: At different stages, our focus and differences are actually different. When doing open source Kylin, the biggest selling point at that time was to provide high concurrency and high-performance analysis on very large data sets, such as data sets above the PB level, we can find out, and also support some relatively high concurrency.

When we did OLAP work across the enterprise, high availability, security, multi-tenancy, resource isolation, these financial grade features were competitive advantages and barriers at the time. Only doing technology is not enough, and later we also did some innovations, such as AI enhanced engines, automatic allocation of resources, etc., and began to polish new differentiation through intelligence.

When it comes to the data metrics platform, we put a lot of emphasis on management methodology. We believe that the indicator is far from a simple technology, a data point, in essence reflects a company’s management method and the implementation of this methodology tools, when everyone generates data, we will emphasize the use of PDCA and other ways, through the form of value indicators, value trees to complete management and decision-making help. Especially at home, Kyligence wants a clear understanding of the overall metrics, from bosses to employees. CXO UNION-CXO联盟（cxounion.cn）

Today is the same, when AI comes in, you will find that Copilot or ChatBI can be done by everyone, but with different characteristics. Kyligence’s metrics-based system has been exploring how AI can help you do some analysis, operations, and management, rather than simply querying data. For example, the Kyligence goal decomposition or goal evaluation function can directly help you evaluate KPIs and provide recommendations to help you improve performance and competitiveness.

Kyligence Copilot product image | Source: official website

Q2: The core competence of the Kyligence index platform is the best practices, or management methods, behind the salvage industry.

Han Qing: Yes, it’s called management methods, not practices. Best practices often turn into projects. In my opinion, software is essentially a tool for management methodology, so why ERP? Why this system and not that one? Behind is a set of management logic, methodology, this is the core of the core. Otherwise, the resulting system will either become a project or something that cannot be replicated.

In addition, the management methodology of Chinese companies is actually different from the management methodology of American companies, which is obvious. But I think it’s time for Chinese companies to start creating their own management methodology. For so many years, we have been learning the methods of advanced Western enterprises, but looking back, the culture and philosophy of China are different, and accordingly, the management methodology is completely different.

I think Chinese software should have a unique set of things in the future, and the root behind it is Chinese philosophy and Chinese management methodology. Without this management methodology, there is no way to guide software. CXO UNION-CXO联盟（cxounion.cn）

Q3: Software carries this methodology and makes customer management better and better. But there is also a view that China does not have a universal management methodology, and the approach may be different from industry to industry and from company to company.

Han Qing: Over the years of founding Kyligence, after communicating with many customers on the front line of the market, I think that the difference in the United States is that it has management first (and then software), and has passed the grassy stage that we (now experiencing). Since Drucker started to establish the methodology of modern corporate governance, various business schools and large companies have deposited the best practices or theoretical foundations common to their society, and the ways and methods in line with Western culture and philosophy. So every company’s management logic and things are the same.

But China is different. China has spent 40 years of reform and opening up, and 200 years of doing what others have done. Therefore, in this process, successful enterprises have their own methods, and none is the same. CXO UNION-CXO联盟（cxounion.cn）

Second, the bosses of Chinese companies are spelled out, very Wolf, and there is a characteristic – no one disagrees with who, “You ran to tell me how to manage this company? Do you understand?” I’m sorry, bosses who are even remotely accomplished have big Egos. But at the same time, you find that all the bosses are studying management. Because we all know we’re failing. We need something better. So, everyone is eager to succeed, buy a variety of methodologies, buy a variety of software, and finally become a “poor student stationery, made a bunch, and then found that it is useless.”

Q4: If China is not ripe for a common management method, will you make a detente by doing so?

Han Qing: In this case, I think that Chinese software can have a different way that is in line with the current situation of Chinese enterprises – the bottom layer maintains a set of data-driven standardized capabilities, and the top layer is like building Lego to spell out different management capabilities.

In other words, if the whole system and the underlying system are made flexible enough that each boss can bring his management methodology and management philosophy to bear on the platform, it is feasible. This platform does not need to become a very complex system, “to teach you what to do”, but a good tool, “you” must have your own management methodology, with this tool to carry, it helps you to push down what you want, because the essence of the indicator platform is a command system.

The entire underlying platform of Kyligence is very flexible, which can adapt to different management methodologies, but the underlying data and indicators, and absorb the core logic of Western data governance, using data to make decisions. In this way, there is an opportunity to forge a distinctive path that not only ensures the standardization of the underlying products of Kyligence, but also can adapt to the differences in management methodologies of different companies.

Q5: How can you be so flexible? There are also many common requirements for PaaS integration.

Han Qing: Get to the core. In general, a company should have three systems. The first system is a production and manufacturing system with ERP as the core to solve the problem of “production”. The second system is a marketing system with CRM as the core to solve the problem of “selling out”.

The third system is the command and decision system with the index platform as the core. How is the business going today? Are the will and intentions of management effectively implemented? How to intervene earlier in the process to ensure that final performance is achieved. In other words, the management command system is based on the index platform. CXO UNION-CXO联盟（cxounion.cn）

Because every company ultimately serves financial indicators, from the financial level, it can be broken down to the degree of atomic indicators. As long as these invariants are grasped, the underlying data indicators are precipitate in the system, and the standard caliber of the data definition is consistent. As for how the upper level is combined, it is more concerned about marketing or cost, in fact, it is a series of different management ideas and methods. In this way, the problem of flexibility can be solved.

And AI can make command-and-control systems work better. Through powerful computing and reasoning capabilities, AI can inform the latest overall situation in a timely manner, and even some indicators that it would not normally pay attention to. At this time, I think AI will bring great help to the whole management ability, management decision-making, command decision-making.

2. AI+ indicator platform makes management methodology mature

Q1: AI may accelerate the maturation of management methodologies.

Han Qing: AI can definitely accelerate. Chinese companies are growing fast, but they lack people who can manage them well.

Unlike professional managers in the United States, Chinese enterprises are developing too fast, most managers are temporarily promoted, how to manage people, how to open people, how to manage costs have never done, can not find experienced, including myself. Under such circumstances, it is absolutely useless to talk about management methodology and academic things. To put it crudely, sometimes “management” becomes a variety of “struggles.” CXO UNION-CXO联盟（cxounion.cn）

I said in the company that the people who need to be managed most are the managers themselves. It is impossible to send every manager to an MBA without the money or time. But I suddenly found that AI seemed to be able to do this thing and help every manager have better management experience.

For example, our company has made an indicator tree for all senior executives, and every Monday, AI will help run a summary report on this indicator tree to the entire management, but the content adds up to a lot. I can ask the AI to give me something in no more than 500 words, and it can tell me directly where the risk indicators are and where the management should be strengthened. So AI can empower management.

Q2: Indicators like Copilot sounds particularly friendly to managers, who can have a stronger sense of control through AI. What about the employees?

Han Qing: Employee empowerment is what I value. The metrics platform we designed and Copilot are not for professional users. For companies, no more than 10 to 15 percent of the people have access to data power is a privilege, and by adding Copilot to the Kyligence Zen metrics platform to give everyone else the opportunity to access data through natural language, employees are empowered.

The performance of the company must depend on the front-line employees, they are the people who really need to use the data, the boss to see the data is never a problem. Today’s AI+ metrics platform can provide data services and capabilities that are available to everyone, and you will find that the data literacy of the entire company is up.

In addition, empowering employees is also reflected in the fact that AI can be tireless, and AI has massive knowledge reasoning ability, his knowledge breadth is far greater than that of a single human, and the reference to you is far greater than what you can see. For example, there is a function called attribution analysis in our product, and even data analysts may not have a good grasp of this analysis function, but today you can ask AI, and it will immediately help with attribution and summary, which is enabling. CXO UNION-CXO联盟（cxounion.cn）

Q3: Will companies give everyone a Copilot in the future?

Han Qing: I believe that in the future, the company will definitely assign an indicator Copilot or data Copilot to everyone, especially the first-line enterprises or private enterprises in China. The management of the company always needs some data or indicators to reflect, if I do not give you a tool to monitor you, but give you a real assistant to empower you, whether it is for the management of the company or the business people, it is valuable.

5. Cooperation and Competition in the era of large models

Q1: Last month, at the launch of Flying Book 7, Kyligence appeared as a partner of the exclusive Polaris Indicator. How was this decision driven?

Han Qing: First of all, we are heavy users of flying books, because we have always respected the observability of management internally, and also applied flying books to the extreme. This cooperation with Flybook smart partners is also the super traffic entrance of Flybook to push Kyligence products to more people.

The cooperation with Fei book can be said to hit it off immediately, Fei book lets you focus more on the platform, office coordination and other aspects, and we have more advantages in data indicators and North Star management. CXO UNION-CXO联盟（cxounion.cn）

Q2: In your opinion, the arrival of large models will make ecological cooperation more frequent?

Han Qing: I think so. What these things have changed is that the world is moving too fast. After the big model came in, every company became more anxious, either to reinvent itself with AI or to be replaced by AI, this thing was obvious. At this time, we chatted down and found that when there was nothing before, you did my work and I did yours, and now it is better to cooperate.

Q3: At the end of the year, everyone is making a strategy. What is the overall product layout plan of Kyligence in terms of AI?

Han Qing: The first one is focus, focusing on the index platform as the core, putting AI things more deeply into it. In fact, some customers have asked me if we can do some kind of big model capability like trivia or related, but that’s not what we want to do.

The second is the iteration of technology, I have recently been asking the technical team a topic, “What do you think the data warehouse in the AI era should be like, what should the data analysis in the AI era be like?”, I hope you keep thinking about fundamental issues, this place is likely to be some subversive things, must think deeply, because the changes brought by AI are too fast.

The third is to expand our business more quickly, the maturity of our products is very high, now there are a number of customers have paid online, next year is a year that can do rapid growth, can more quickly run the horse circle, occupy the user’s mind.

Q4: Are there any differences between starting a business this year and previous years? How are feelings and markets different?

Han Qing: It’s exciting, but it’s hard. Recently saw an analysis, “the leftover is king”, the rest of the remaining. CXO UNION-CXO联盟（cxounion.cn）

I think the whole Chinese business service has entered a very low trough period. But I firmly believe that China will enter a golden age of Chinese software in the future, maybe this is the darkness before dawn, generally speaking, it is always darker before dawn, but through it, the next is a very huge opportunity to break out, which is also our feeling at the front line, depending on how the layout, I hope everyone will bloom together in the spring.

由CXO UNION-CXO联盟（cxounion.cn）转载而成，来源于极客公园；编辑/翻译：CXO UNIONCXO联盟小U。

如需加入CXO UNION（CXO联盟）高管社群，请联系社群小伙伴哦~

免责声明: 本网站(http://www.cxounion.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。

本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。