The organizational brain for the enterprise 企业级组织大脑

Every system.
One brain.

所有系统
一颗大脑

One conversational brain across every system — PLM, ERP, MES, Jira, GitHub. Ask in plain language. Private by design.

一颗对话式大脑，覆盖所有系统——PLM、ERP、MES、Jira、GitHub。用中文提问。私有部署。

✓ Private deployment私有部署 ✓ All major PLMs兼容所有主流 PLM ✓ Live in 48 hours48 小时上线
Latent
Live · 4 systems synced 实时 · 4 个系统已同步
Which projects are at risk this week? 这周哪几个项目风险最高？
3 projects flagged — cross-referenced PLM + Jira + ERP:

• Helios-07 HIGH · 33% done, predicted 6-week slip
• Atlas-12 HIGH · 2 unassigned milestones, lead at 115% load
• Orion-03 MED · implementation phase 11 days overdue
识别到 3 个高风险项目 —— 跨 PLM + Jira + ERP 数据：

• Helios-07 HIGH · 完成 33%，预测延期 6 周
• Atlas-12 HIGH · 2 个里程碑无主，负责人负荷 115%
• Orion-03 MED · 实施阶段超期 11 天
Who owns Helios-07? Any blockers? Helios-07 负责人？目前卡点？
Owner: Michael Reed (Mechanical Design Lead).
5 overdue tasks, longest 18 days. 3 open Jira issues tagged blocker. Recommend a 1:1 today.
负责人：张明（机械设计组组长）。
5 项任务逾期，最长超期 18 天。Jira 上有 3 个 blocker 标签问题未处理。建议今天 1:1。
Ask anything about your company 问任何关于你公司的问题
Early design partner 首批共建伙伴
Xili Tech 西力科技
Smart metering · listed manufacturer 智能计量 · 上市制造企业
10 days 天
From kickoff to live on PLM从启动到 PLM 上线
120→5 min 分钟
Decision review cycle决策评审时长
2
Systems connected and answering已接入并可对话的系统
Why Latent · in one sentence each
三句话讲完 Latent

Unified across silos 打通所有系统孤岛

One question, complete context — across PLM, ERP, MES, GitHub, and Jira, read live.

一个问题，完整上下文——横跨 PLM、ERP、MES、GitHub、Jira 实时读取。

Private by design 私有部署，数据不出内网

On-prem deployment. AI runs inside your network. Audit-ready logs, no phone-home.

完全本地部署，AI 在内网运行，完整审计日志，无任何外联。

Live in 48 hours 48 小时上线

Pre-built connectors for every major PLM / ERP / MES. First insight before onboarding ends.

主流 PLM / ERP / MES 预制连接器。培训没做完，第一批洞察已经出来。

Integrations
系统集成

The only tool that reads
every system you already use.

连接你已有的
所有研发工具链

Stop asking engineers to export spreadsheets. Latent ships with pre-built, production-grade connectors for every major system in manufacturing R&D.

不需要让工程师手动导 Excel。Latent 提供生产级的预制连接器，覆盖制造业研发的所有核心系统。

  • ✓ Real-time read access, no batch imports — data is always fresh.实时读取，无需批量同步——数据始终最新。
  • ✓ Read-only by default — Latent never writes back unless you authorize it.默认只读——未经授权不会写入任何源系统。
  • ✓ Custom connectors available for legacy or in-house systems.支持定制连接器——老旧或自研系统也能接入。
  • ✓ Full audit trail — every query and data access is logged.完整审计日志——每次查询和数据访问均有记录。
PLM Product lifecycle 产品生命周期 Design 设计
SAPSAP PLM SiemensSiemens Teamcenter PTCWindchill 3DSDassault ENOVIA
ERP Enterprise resource 企业资源计划 Plan & order 计划 · 采购
SAPSAP ERP ORACLE 用友YonyouNC
MES Manufacturing execution 制造执行系统 Produce 生产执行
SiemensSiemens Opcenter 用友MES
R&D 研发 Code · tasks · docs 代码 · 任务 · 文档 Collaborate 协作
GitHubGitHub GitLabGitLab JiraJira ConfluenceConfluence LarkFeishu / Lark飞书 + Custom · any REST or SQL source + 定制 · 任何 REST / SQL 数据源
How Latent is different
Latent 的不同之处

Not a dashboard.
Not a chatbot. Not a project.

不是看板
不是聊天机器人
不是一个需要做的项目

The most common question buyers ask: "Isn't this just ChatGPT with our data?" No. Here's what happens when you compare Latent head-to-head with the three things people try first.

买家最常问的一句话："这不就是接了我们数据的 ChatGPT 吗？"——不是。下面是把 Latent 和通常会先尝试的三个方案逐项摆在一起的结果。

Latent Traditional BI传统 BI
Tableau · Power BI · DataV		 General AI Copilot通用 AI 助手
ChatGPT · Claude · Copilot		 Self-built RAG自建 RAG
Internal GPT project内部 AI 项目
Reads PLM / ERP / MES live实时读取 PLM / ERP / MES ✓Pre-built connectors预置连接器 —Manual export & ETL手动导出 + ETL ✗No data access无数据接入 ◐Build it yourself自己开发
One question, every silo一个问题，跨系统回答 ✓Joins across systems跨系统关联 —One dashboard per system每系统一个看板 ✗No context无上下文 ◐If you engineer it取决于实现
Time to first insight首个洞察耗时 ✓48 hours48 小时 ◐Weeks of BI work数周 BI 建模 ✗Never — no data无法——没有数据 —3–6 months3–6 个月
Runs on-prem / airgapped支持本地 / 内网隔离部署 ✓Default deployment默认部署方式 ◐Some vendors部分厂商支持 ✗SaaS only仅公有云 ✓If you host自托管可以
Audit trail + access control审计日志 + 权限管理 ✓Every query logged每次查询留痕 ◐Dashboard-level only仅看板级别 ✗None by default默认无 ◐Build it yourself自行实现
Ongoing maintenance长期运维 ✓Fully managed全托管 —Dedicated BI team需专职 BI 团队 —N/A不适用 ✗Your team, forever永远你的团队维护
✓ strong fit · ◐ partial · — weak · ✗ not applicable ✓ 强项 · ◐ 部分满足 · — 较弱 · ✗ 不适用
Live Intelligence
实时洞察

Every risk, surfaced.
Every blocker, named.

每个风险，浮出水面
每个卡点，有名有姓

Latent doesn't just answer questions. It proactively monitors your R&D org, scoring every project across 6 dimensions and flagging issues before they become fires.

Latent 不只是被动回答问题。它主动监控你的研发组织，从 6 个维度给每个项目打分，让风险在变成火情之前就被识别。

Project Risk Status 项目风险状态 3 at risk 3 高风险
HIGH Helios-07 (EV Motor) Helios-07 新能源电机
33%
HIGH Atlas-12 (Drivetrain) Atlas-12 传动系统
51%
MED Orion-03 (Chassis) Orion-03 底盘优化
78%
LOW Nova-05 (Braking) Nova-05 制动系统
92%
Review Task Queue 审核任务队列 11 pending 11 待审核
MRZM
Michael Reed 张明 18d overdue · 5 items 逾期 18 天 · 5 项
SKLH
Sarah Klein 李华 7d overdue · 3 items 逾期 7 天 · 3 项
DPWF
Daniel Park 王芳 On time · 3 items 今日审核 · 3 项
Active Warnings 活跃预警 View all 查看全部
CRITICAL
Helios-07 progress far behind plan. At current velocity, will miss deadline by 6 weeks. Helios-07 项目进度严重落后，以当前速度无法在截止日期前完成，预计延期 6 周。 → Schedule project review meeting today → 建议立即召开项目评审会
HIGH
Michael Reed has 12 active tasks (115% capacity). 5 overdue. 张明 当前 12 个活跃任务，负荷 115%，5 个逾期。 → Redistribute 3 tasks to balance load → 建议重新分配部分任务
MED
Implementation phase overdue by 11 days in 8/13 projects. Systemic bottleneck suspected. 实施阶段在 8/13 个项目中平均超期 11 天，可能存在系统性瓶颈。 → Audit implementation process → 审查实施阶段流程
Risk Dimensions 风险维度评分
Progress 进度
42
Resources 资源
58
Quality 质量
71
Comms 沟通
79
Technical 技术
83
Compliance 合规
88
Who uses Latent
谁在用 Latent

From the shop floor
to the C-suite.

从一线到董事会
每个角色都用得上

Different questions, same brain. Everyone gets answers from the same real-time data — so leadership, operations, and execution stay aligned.

不同角色问不同问题，用的是同一颗大脑。同一份实时数据，让管理层、运营层、执行层始终对齐。

CTO / VP Engineering
CTO / 研发总监

Strategic oversight
without the spreadsheets

战略掌控
告别报表堆

Every Monday: one question gets the week's critical issues ranked by priority, with owner names, deadlines, and recommended actions.

每周一早上一个问题，获得本周最关键的风险列表（按优先级排序），标注责任人、截止日期、建议动作。

Cut my Monday review from 2 hours to 15 minutes. I see what matters, not what's in the spreadsheet.
周一例会准备时间从 2 小时压缩到 15 分钟。看到的是真正重要的事，不是表格里的数字。
Engineering Manager
研发经理

Spot overload
before it burns out

识别超负荷
避免团队崩溃

Latent watches task assignments, overdue rates, and cross-project load. Flags engineers at 90%+ capacity before they burn out or miss deadlines.

Latent 持续监控任务分配、逾期率、跨项目负荷。工程师超过 90% 负荷时主动预警，避免在崩溃或错过节点时才发现。

Caught two overload situations this month before they became delays. Huge.
这个月发现了两个超负荷的同事，在变成延期之前就处理掉了。太有用了。
Project Manager
项目经理

One source of truth
across PLM, Jira, MES

跨系统
统一数据口径

No more reconciling progress data between PLM milestones, Jira tickets, and MES production status. Latent joins them automatically.

不再需要在 PLM 里程碑、Jira 工单、MES 生产状态之间对口径。Latent 自动跨系统聚合，口径始终一致。

I used to spend 3 hours/week just reconciling data. That's gone now.
以前每周光对数据就要花 3 小时。现在这件事没有了。
COO / Operations
COO / 运营负责人

See supply-chain risk
before it hits the floor

供应链风险
未到车间先预警

Latent joins ERP purchase orders, MES production status, and supplier signals. Know which parts are late, which lines will stall, which orders need a call — before the daily stand-up.

Latent 打通 ERP 采购订单、MES 生产状态、供应商异常。哪个零件延迟、哪条产线要停、哪个订单要紧急沟通——在晨会之前就知道。

Caught a supplier delay 10 days before it would've stopped the line. Saved a six-figure recovery.
提前 10 天发现供应商延迟，避免了生产线停工。挽回了六位数的损失。
CFO / Finance Lead
CFO / 财务负责人

Real project economics
this week, not this quarter

项目真实经济性
按周看，不按季度

Live join of ERP spend, PLM project phase, and team allocation. See which projects are burning above plan, which are coming in under — with the actual numbers, not a month-end export.

ERP 支出、PLM 项目阶段、团队分配的实时打通。哪些项目超预算、哪些低于预算——看的是真实数字，不是月末导出。

I finally see what each project costs this week, not when my FP&A team finishes the close.
我终于能看到每个项目本周的实际成本，不用等月度关账。
VP Sales / Rev Ops
销售 VP / 收入运营

Real deals vs.
ghost pipeline

真实订单
还是虚假管道？

Cross-check CRM pipeline against ERP quotes, product usage, and support tickets. Find deals dressed up to look hot and deals that are actually warmer than they seem.

交叉验证 CRM 管道与 ERP 报价、产品使用、支持工单。识别被包装得"很热"但实际不会成交的订单，也能发现比看起来更值得押注的真机会。

Killed $2M of fake pipeline in week one. Sales forecast is finally believable.
上线第一周砍掉了 200 万的假管道。销售预测终于可信了。
3hrs/wk
Saved per director, per week
每位研发总监每周节省
47
Avg risks identified in first 90 days
上线前 90 天平均识别风险数
48hrs
From kickoff to first insight
从部署到第一批洞察
0
Bytes of data that leave your network
数据离开企业内网的字节数
What it's worth
价值量化

Latent pays for itself
in weeks, not years.

Latent 的回报周期
以周计，不以年计

Move the sliders below. Numbers update live. Based on pilot data — your real deployment will differ, but the order of magnitude holds.

拖动下方滑块，数字实时更新。模型基于 pilot 数据——你的实际部署会有差异，但数量级不会偏。

R&D headcount研发团队规模 200
Hours/week hunting data每周找数据耗时 4 hrs小时
Systems in play涉及系统数 6
680
hours recovered per week每周节省工时
$3.4M
annual value recovered年化价值
3
weeks to payback周数回本
Assumes 85% reduction in cross-system lookup time (observed at Xili Tech). Hourly cost modeled at $100 fully-loaded. 模型假设跨系统查找时间减少 85%（西力科技 pilot 实测）。全负担人工时薪按 ¥500 计算。
Customer story
客户案例

Before Latent, my Monday review meeting ran two hours. We'd argue about which project was actually at risk, then half the data was wrong anyway. Now I walk in, open Latent, and in 5 minutes I know exactly who needs help and what to unblock. It replaced an entire reporting layer.

用 Latent 之前，我每周一的评审会开两个小时，还在为哪个项目真正有风险争论，等到最后发现一半数据不对。现在我开会前打开 Latent，5 分钟就知道谁需要帮助、什么要解决。它替代了我们整个一层汇报体系。

DCCW
David Chen 陈伟
VP of Engineering · Xili Tech (pilot customer) 研发副总裁 · 西力智能科技（首批共建客户）
Security & Compliance
安全与合规

Your data never
leaves your network.

数据始终
留在你的内网

Private deployment isn't an enterprise-tier upsell. It's the only way we ship. Latent runs inside your VPC, your on-prem servers, or your airgapped environment.

私有部署不是企业版加钱才有——这是 Latent 的唯一交付方式。Latent 运行在你的 VPC、本地服务器、或完全隔离的内网环境中。

Fully private deployment

完全私有化部署

Installed on your infrastructure. No telemetry, no phone-home. Source code escrow available for regulated industries.

部署在客户自有基础设施。无遥测、无外联。受监管行业可提供源代码托管。

SOC 2 & ISO 27001 aligned

对齐 SOC 2 与 ISO 27001

Security practices mapped to SOC 2 Type II and ISO 27001. Full SSO, RBAC, audit logs, and data encryption.

安全实践对齐 SOC 2 Type II 和 ISO 27001 标准。完整 SSO、RBAC 权限、审计日志、数据加密。

On-prem LLM option

支持本地部署 LLM

Bring your own model — Claude, GPT-4, Gemini, or open-source (Llama, Qwen, DeepSeek) running entirely inside your network. AI never sees the public internet.

自带模型——通义千问、DeepSeek、Llama 等开源模型完全在内网部署运行，或接入 Claude、GPT-4、Gemini。AI 推理完全本地化，模型不接触公网。

SOC 2
SOC 2 Type IIAligned · audit scheduled对齐 · 审计排期中
ISO27001
ISO 27001Aligned · audit Q3 2026对齐 · Q3 2026 审计
等保三级
China MLPS 2.0 等保 2.0 Level 3 ready三级准备就绪
GDPRCompliant合规
Deployment Architecture 部署架构
🏢
Customer VPC / On-prem
客户 VPC / 本地数据中心

Where everything runs

所有组件运行位置

Private 私有
↓
⬡
Latent Core
Latent 核心服务

Connectors · Analysis engine · API

连接器 · 分析引擎 · API

↓ ↑
🔒
Your source systems
你的源系统

PLM · ERP · MES · Jira · GitHub (read-only)

PLM · ERP · MES · Jira · GitHub（只读）

↓ ↑
🤖
LLM (choose deployment)
LLM（可选部署方式）

On-prem · Private cloud · API (opt-in)

内网部署 · 私有云 · API 调用（可选）

About Latent
关于 Latent

We're building the brain
every company deserves.

我们在打造
每家公司都该有的大脑

A company born in the factory.
Built for every enterprise.

诞生于工厂
服务于所有企业

Latent started from a simple observation: decision-makers spend more time reconciling data across 5+ systems than actually deciding. Systems don't talk. Teams burn out. Opportunities slip. And nobody has a complete picture until the quarterly review.

Latent 源于一个简单的观察：决策者花在"跨 5+ 个系统对数据"上的时间，远多于真正做决策的时间。系统互不连通，团队过载，机会溜走，等到季度评审才看见完整问题。

We started in manufacturing R&D because that's where the pain is sharpest. But the brain we're building applies to every org that runs on fragmented systems — which is every org.

我们从制造业研发起步——那是痛点最尖锐的地方。但我们构建的这颗大脑适用于所有靠多个系统运转的组织——也就是所有组织。

Where
在哪

Hong Kong · Global deployment

香港 · 全球部署

Team
团队

One founder, an AI-native stack. We eat our own dog food.

一个创始人，一套 AI 原生工作流。我们自己就是 Latent 的用户。

Mission
使命

Make every company as data-driven as the best tech company

让每家公司都像顶尖科技公司一样数据驱动

Principle
原则

Your data never leaves your network. Ever.

你的数据不会离开你的内网。永远不会。

Timeline
时间线
2025 Nov
Founded
Latent 成立

Spun out of a manufacturing PLM consulting engagement.

从一个制造业 PLM 咨询项目中孵化。

2026 Mar
First pilot live
首个 pilot 上线

Xili Tech connected in 10 days. 2 systems live and answering.

西力科技 10 天上线，接入 2 个系统并可对话。

2026 Q2
Connector library shipped
连接器矩阵全量上线

PLM (SAP, Siemens, PTC, Dassault), ERP, and MES — all major systems covered.

PLM（SAP、西门子、PTC、达索）、ERP、MES——主流系统全覆盖。

2026 H2
You?
下一个是你？

Next pilot cohort opening. Book a call.

下一批 pilot 正在开放。欢迎预约沟通。

Ziyi Song

Hi — I'm Ziyi. Every modern company runs on five, ten, twenty systems — PLM, ERP, MES, Jira, GitHub — and nobody has a single brain that sees them all. Every decision waits for someone to reconcile spreadsheets, and every wait is an opportunity dying quietly.

Latent is that brain — the one I wanted for my own team first. If you've felt this pain, write me directly. Every demo is still run by me.

我是 宋子钇。每家现代公司都跑在 5 个、10 个、20 个系统上——PLM、ERP、MES、Jira、GitHub——但没有一个大脑能看到全貌。每次决策都在等某个人手动对 Excel，每等一次，机会就溜走一次。

Latent 就是那颗大脑——最初是我给自己团队造的。如果你也有这种痛，直接写信给我。每个 demo 还是我自己演。

Ziyi Song · Founder, Latent · Hong Kong 宋子钇 · Latent 创始人 · 香港
ziyisong@asklatent.com
FAQ
常见问题

Straight answers to
the questions buyers ask.

买家必问
我们直接回答

Still have questions?

还有其他问题？

The fastest way to get every question answered is a 30-minute demo with our solution team. We'll show Latent on your real systems, walk through the deployment model, and scope a pilot that fits your timeline.

最快的方式是和我们解决方案团队做一次 30 分钟演示——在你真实的系统上跑给你看，讲清楚部署方式，根据你的时间线设计 pilot 方案。

How does Latent connect to our existing systems? Latent 如何连接我们现有的系统？
Pre-built connectors for SAP, Oracle, Siemens, PTC, Dassault, Yonyou, GitHub, GitLab, Jira, Confluence, WeCom, Feishu, DingTalk — plus REST and SQL bridges for anything with an API. Custom connectors for legacy or in-house systems are included in enterprise onboarding at no extra cost.
预制连接器覆盖 SAP、Oracle、西门子、PTC、达索、用友、GitHub、GitLab、Jira、Confluence、企业微信、飞书、钉钉——加上 REST 和 SQL 桥接层，只要有 API 的系统都能接入。老旧或自研系统可定制连接器，企业版实施包内免费提供。
Can Latent write back to our source systems? Latent 会写回到我们的源系统吗？
No, read-only by default. If you want write-back — e.g. updating a Jira ticket or closing a PLM milestone — it requires explicit per-user authorization, is scoped to specific actions, and every write is logged. Your systems remain the source of truth.
默认只读。如果你需要写回——比如更新 Jira 工单或关闭 PLM 里程碑——需要明确的用户级授权，限定具体动作，每次写入都有完整日志。你的系统始终是唯一真实来源。
How do you handle permissions and access control? 权限和访问控制怎么处理？
Latent inherits permissions from your source systems. If a user can't see a project in PLM, they can't ask Latent about it either. Integrates with SSO (Okta, Azure AD, 飞书身份), SCIM for user provisioning, and your existing RBAC policies. No separate permission layer to maintain.
Latent 继承源系统的权限。如果用户在 PLM 看不到某个项目，也问不了 Latent。支持 SSO（Okta、Azure AD、飞书身份），SCIM 用户同步，以及你现有的 RBAC 策略。不需要维护独立的权限层。
What about AI hallucinations? How accurate are the answers? AI 幻觉怎么办？答案有多准确？
Every answer is grounded in real data from your systems — with clickable citations for every claim. Users can trace any statement back to the source record (a specific PLM entry, a Jira ticket, an ERP transaction). When data is insufficient or conflicting, Latent says I don't have enough data to answer this confidently rather than guess.
每个答案都基于你系统里的真实数据——每条结论都有可点击的引用。用户可以追溯到具体的源记录（某条 PLM 记录、某个 Jira 工单、某笔 ERP 交易）。当数据不足或存在冲突时，Latent 会说"数据不足以给出可靠答案"，而不是编造。
How long does deployment take? 部署需要多长时间？
Typical timeline: 48 hours to first insight on a pilot dataset, 2 weeks to full production rollout across your primary systems. Our implementation team is on-site or remote with your team during the first week. Enterprise rollouts with complex legacy systems can run 4–6 weeks.
典型时间线：48 小时在 pilot 数据集上跑出第一批洞察，2 周完成主要系统的正式上线。第一周我们的实施团队驻场或远程跟你们团队一起工作。涉及复杂遗留系统的企业级部署通常 4-6 周。
Can we use our own LLM or keep AI inference on-prem? 可以用我们自己的 LLM 或本地部署 AI 吗？
Yes. Latent supports the top frontier models (Claude, GPT-4, Gemini) via direct API or your private cloud (Azure OpenAI, AWS Bedrock), as well as open-source models (Llama, Qwen, DeepSeek) deployed entirely inside your network. For regulated environments, full on-prem inference means no tokens ever leave your infrastructure.
支持。你可以自选模型：开源模型（通义千问、DeepSeek、Llama）完全在你的内网运行；或接入海外顶级模型 Claude、GPT-4、Gemini（通过 Azure OpenAI、AWS Bedrock 或直连 API）。对受监管环境，完全本地推理意味着 token 永远不离开你的基础设施。
What's the pricing model? 定价模式？
Enterprise annual contracts priced by number of seats, connected systems, and deployment model (cloud vs. on-prem). Pilot programs are available at a reduced scope to prove value before full rollout. Book a demo for a quote tailored to your environment.
企业级年度合约，按席位数、连接系统数、部署方式（云或本地）定价。提供缩小范围的 pilot 方案，让你先验证价值再全面推开。预约 demo 我们给出针对你环境的报价。
Your company deserves one brain 你的公司值得有一颗大脑

Every system.
One brain.

所有系统
一颗大脑

See Latent running on your real systems, your real data.
30-minute demo tailored to your stack and use cases.

用你自己的系统、真实的数据，看 Latent 实际跑起来。
30 分钟定制化演示，展示最贴近你业务的部分。

No credit card · No commitment · Private demo environment

无需信用卡 · 无合约承诺 · 独立私有演示环境