PRICING

三种交付方式Three delivery models

三者功能相同,区别在于代码部署在何处、由谁维护、能否改造成只属于你们的版本。标准云按服务器规格报价,不按人数计费;另外两种按许可与项目计价。The features are identical in all three. What differs is where the code runs, who maintains it, and whether it can be reshaped into a version that is yours alone. Standard cloud is priced by server spec rather than headcount; the other two are priced by license and by project.

STANDARD CLOUD

选择最多MOST CHOSEN

标准云Standard cloud

我们托管,开通后即可使用。多数团队从这里开始。We host it and you sign in to start. Most teams begin here.

八个模块全部开放All eight modules

智能体注册表和用量统计Agent registry and usage metering

标准 SOP 模板,节点可编辑Standard SOP templates, nodes editable

内置邮件,用你们现有的邮箱Built-in mail, using your existing addresses

与主线版本同步,每两周更新Tracks mainline, updated every two weeks

不限席位,按算力计费Unlimited seats — you pay for capacity

按服务器规格报价BY SERVER SPEC

SELF-HOSTED

私有化Self-hosted

部署在你们自己的机房或专有云,数据与模型均不出内网。Runs in your own data center or private cloud. Neither the data nor the models leave your network.

标准云的全部功能Everything in Standard cloud

部署在你们的环境里Deployed in your environment

模型可以换成自有推理服务Swap in your own inference service

审计日志导出、留存策略按你们的规定Audit export and retention to your policy

升级窗口由双方约定,不强制Upgrade windows agreed, never forced

许可 + 支持LICENSE + SUPPORT

CUSTOM BRANCH

定制版Custom branch

为你们单独维护一个代码分支,功能可以与主线不同。Your own code branch, maintained separately. Its features can differ from mainline.

私有化的全部内容Everything in Self-hosted

独立分支,不必和其他客户对齐A separate branch, not aligned to other customers

单据格式、审批链按你们的做法实现Document formats and approval chains built your way

主线更新按你们的节奏合并Mainline updates merged on your schedule

专属回归测试与升级支持Dedicated regression suite and upgrade support

项目 + 年度维护PROJECT + ANNUAL

定制版是什么WHAT THE CUSTOM BRANCH ACTUALLY IS

标准云与私有化运行同一份主线代码。你们提出的需求排进主线,完成后所有客户一并获得 —— 好处是维护成本摊薄,代价是需求要与其他客户的做法折中。Standard cloud and Self-hosted run the same mainline code. A request you make enters mainline and every customer receives it — the maintenance cost is shared, but the request must also work for everyone else.

何时值得选择这条路径WHEN IT IS WORTH IT

定制版为你们保留独立分支,你们需要的字段与流程只存在于这个分支,无需说服他人。分支须单独维护,因此有年度维护费,主线的新功能也会延后到达。业务做法确实与其他公司不同时,这条路径比改造主线更快。A custom branch keeps your fields and flows to yourselves, and no one else has to agree to them. It is maintained on its own, so there is an annual fee and new mainline features arrive later. When your way of working genuinely differs from other companies', this path is faster than reshaping mainline.

ROI CALCULATOR

用你们的数字测算一遍Run it on your numbers

两条假设写在下面,都可以拖动调整。算法与系统内使用的一致:节省的工时按工作单元折算,AI 成本按每节省一小时的单价计算。Both assumptions are written out below, and either can be dragged. The calculation is the same as in the product: saved hours derived from work units, AI cost per saved hour.

你们的数字YOUR NUMBERS

月单量Jobs per month

近三个月平均Average of the last three months

操作人数Operations headcount

用于计算团队工时与占比Sets team capacity and share

人均月薪(元)Average monthly salary (CNY)

两条假设TWO ASSUMPTIONS

每票节省的人工Time saved per job

包含单据解析、字段预填、跨单核对与邮件抄写四项之和。Covers document parsing, prefill, cross-document checks and re-typing from mail, taken together.

每省一小时的 AI 成本AI cost per saved hour

{{ costPerHour }} CNY

实际运行中一般在 30–50 元。上线之后这个数值由系统自行计算。In practice it falls between 30 and 50 CNY. Once live, the system computes the figure itself.

月节省工时HOURS SAVED

{{ hoursSaved }}h

占团队工时OF TEAM CAPACITY

{{ sharePct }}%

相当于释放EQUIVALENT

{{ fte }}人FTE

月净节省(已扣 AI 成本)NET MONTHLY SAVING (AFTER AI COST)

{{ netSaving }}CNY

人力口径LABOR {{ grossSaving }} AI 成本AI COST −{{ aiCost }} 时薪HOURLY {{ hourlyCost }} 人均每天省PER PERSON / DAY {{ minPerPersonDay }} min 团队月人力TEAM PAYROLL {{ payroll }}

CHECK

需要节省的工时已超过这些人一个月的工作总时长,而未曾花费的时间无法节省。上面的数值已按团队上限封顶 —— 单量、人数、每票分钟数中有一项与实际不符。The hours saved here exceed what this many people work in a month, and time nobody spent cannot be saved. The figures above are capped at team capacity — one of volume, headcount or minutes per job does not match reality.

CHECK

按这两个数字,AI 一小时的成本高于人工一小时,不存在净节省。这种情况下不宜引入 AI,建议先核对时薪与 AI 成本。At these numbers an hour of AI costs more than an hour of a person's time, so there is no net saving. This is not a case for automation — it is worth checking the hourly rate and the AI cost first.

CHECK

节省的工时占团队总工时四成以上,这个比例偏高。可能是每票节省的分钟数估得过高,也可能是这批人承担的单量本身就不正常。Saved hours exceed 40% of the team's capacity, which is high. Either the minutes per job are estimated generously, or this headcount is already carrying an unusual load.

时薪按 21.75 天、每天 8 小时折算。这里得出的是一个量级,不是报价 —— 准确数字需以你们自己的单据实测。 Hourly cost assumes 21.75 days and 8 hours a day. What comes out here is an order of magnitude, not a quote — for a firm number we run it on your own documents.

用我们的单据实测一遍 →Run it on our documents →

AI COST

AI 费用如何计算How AI usage is billed

token 费用单独计算,不并入订阅费。系统按模型与业务线记录调用量、支出、折算的工时,以及每节省一小时的成本。可按月设定上限。Token spend is billed separately, not folded into the subscription. Calls, cost, the hours they convert to and the cost per saved hour are all tracked by model and by lane. You can set a monthly ceiling.

METERED

按调用与工作单元计量,月底出明细Per call and per work unit, itemized at month end

BUDGET

设定月度上限,超限后自动降级为影子模式Set a monthly ceiling; once it is hit, agents fall back to shadow mode and write nothing.

BYO MODEL

用你们自己的模型,此项仅保留计量With your own models, only the metering is left

ROLL-OUT

四到六周上线,AI 放在最后一步Live in four to six weeks, with AI last

这个顺序有其道理。先将主数据与权限整理干净,再配置流程,然后两套系统并行运行一段时间,确认账目一致,最后才让 AI 开始生成提案。次序颠倒会出问题。The order matters. Clean master data and access first, then the process, then a parallel run until the books tie out, and only then let AI start proposing. Reversing that order causes problems.

主数据与账套Master data

导入组织、港口与费用代码,配置账套与币种。Import organizations, ports and charge codes; set up ledgers and currencies.

流程与权限Process & access

按你们的流程调整节点,确定哪些环节不可绕过,权限授到动作。Adjust nodes to your process, decide which steps cannot be skipped, grant rights per action.

并行试跑Parallel run

两套系统同时录入一批真实单据,以账目一致为准。Enter real jobs in both systems; the run passes when the books tie out.

切换Cut-over

新单据进入新系统,在途单据在旧系统内完成,旧系统转为只读。New jobs come here, in-flight jobs finish where they started, the old system goes read-only.

W5–6

放开 AIAI switched on

先以影子模式核对准确率,再开放预填,最后才允许生成提案。Shadow mode first to check accuracy, then prefill, then live proposals.

这是单公司、一两条业务线的常规节奏。分支机构较多、需对接的系统较多、主数据需要清洗的,周期都会更长;实施前会提供一份按周排定的计划。定制版另行计算,分支建立之后还需完成一轮回归测试。That is the usual pace for one company and one or two lanes. More offices, more integrations, or master data that needs cleaning all take longer; you receive a week-by-week plan before implementation. A custom branch is counted separately — once the branch exists, it needs a regression pass of its own.

预约演示Book a demo

请带一票真实数据,30 分钟从建单推进到结算。AI 在流程中提出了什么、消耗多少成本,我们一并查看。Bring one real job. Thirty minutes from entry to settlement, and we review together what the AI proposed along the way and what it cost.

演示环境按你们的业务线预置:海运出口、空运出口或拼箱The environment comes pre-set for your lane: sea export, air export or LCL

可当场修改一个流程节点,观察限制如何生效Change a process node on the spot and see the restriction take effect

以你们真实的单量与人数重算上面那笔账Run the numbers above again with your real volume and headcount

RECEIVED

已收到。We have received it.

一个工作日内我们会通过你留下的联系方式与你联系,先确定时间,再确认演示环境按哪条业务线预置。We will be in touch within one business day, using the contact details you left — first to agree a time, then to confirm which lane the demo environment should be pre-set for.

公司名称Company

联系人 / 职位Contact / role

邮箱或手机Email or phone

团队规模Team size

交付方式Delivery model

一个工作日内回复具体时间。We reply with a time within one business day.