Evose
Executive Briefing · 2026
evose.ai

构建 AI 原生组织Building AI-Native Organizations

让每一家企业,都长出自己的 AI 组织Every company grows its own AI organization

Evose Who We Are

Evose 是企业级的
AI Agent 协作与治理平台Evose is the enterprise platform for AI agent collaboration and governance

行业领先的企业 AI 落地解决方案An industry-leading way to put AI to work inside the enterprise

evose.ai
01

数字员工平台Digital employee platform

销售、客服、运营,搭成能上岗的数字员工Sales, support and ops roles, built into digital employees that actually work

02

多人协作Team collaboration

人和数字员工同在一个工作台,跨部门接力People and digital employees in one workspace, handing off across teams

03

治理与管控Governance & control

谁能用、干了什么、花了多少,一本账管到人Who may use it, what they did, what it cost — one ledger that covers people too

The Framework · AI First

企业用 AI 的三个阶段The three stages of enterprise AI

88% 的公司已经用起来 AI,但是做到 AI 转型的组织不足 1%88% of companies already use AI. Fewer than 1% have actually transformed.

Stage 01

人问模型People query a model

AI as Tool · 随用随关AI as Tool · open, use, close

AI 的角色AI's role

资料员Research clerk

输出规模Output

单点问答,局部提效One-off answers, local gains

业务状态The business

员工偶尔用,主流程不变Occasional use; core processes unchanged

最危险的错觉The most dangerous illusion

Stage 02

人配助手Everyone gets an assistant

AI as Assistant · 人安排工作AI as Assistant · people assign the work

AI 的角色AI's role

执行员Task executor

输出规模Output

单个任务提效,量级没变Faster tasks, same order of magnitude

业务状态The business

流程靠人一项项过,人是瓶颈Every step still waits on a person

Stage 03

AI 原生组织The AI-native organization

AI First · Agent 驱动全链条AI First · agents run the full chain

AI 的角色AI's role

业务调度官Operations dispatcher

输出规模Output

自动化业务流水线Automated business pipelines

业务状态The business

Agent 跑全流程,人定标准与决策Agents run the process; people set standards and make the calls

核心公式The core formula
业务成果Business results = 大模型能力Model capability × harness 工程Harness engineering × 企业的认知与审美The company's judgment and taste
The Core Distinction

用上 AI,和让 AI 上岗,是两件事"Using AI" and "putting AI on the job" are two different things

用上 AIUsing AI

工具逻辑Tool logic

经验Know-how

人走,经验就没了People leave, know-how leaves

提示词在个人收藏夹,打法从来没写下来。Prompts live in personal bookmarks; the playbook was never written down.

天花板The ceiling

组织提效 30%,到头了About +30% across the org, and that’s it

AI 只加速了员工手上那道工序,仍大量消耗在开会、来回确认。AI only speeds up the one step in the employee’s hands; the time still drains into meetings and re-aligning.

让 AI 上岗AI on the job

组织逻辑Organization logic

经验Know-how

人走,经验还在People leave, know-how stays

把岗位经验沉淀为企业知识库,换个人接手,产出质量一样。Role know-how becomes a company knowledge base — hand it over and the output holds.

天花板The ceiling

人效是传统组织的 10 倍以上10× the output per person of a conventional org

上下游不用互相等,一个人带一队智能体 7×24 自己跑,人只管审核和拍板。No waiting up and down the chain: one person runs a squad of agents around the clock and only reviews and decides.

The Problem

企业转型遇到的困难Where transformations get stuck

人、知识、协作、治理——四道关,过不去一道,AI 就停在个人工具People, knowledge, teamwork, governance — miss any one gate and AI stays a personal tool

① 人① People

熟练用 AI 的人,不到 10%Fewer than 10% really know how to use it

这几个人快了三倍,公司报表一点没变。Those few got 3× faster; the P&L didn’t move.

② 知识② Knowledge

AI 不懂公司的经验和流程AI doesn’t know your know-how or your processes

这些都在文档、群聊和老员工脑子里,AI 拿不到。It all sits in docs, group chats and veterans’ heads, out of reach.

③ 协作③ Teamwork

跨部门协作依赖人Cross-team work still runs on people

文件资料的流转靠人传,权限也不好管。Files and documents get passed around by hand, and permissions are hard to control.

④ 治理④ Governance

成本和收益,说不清Costs and returns don’t add up

没有一张总账,只能一直停在试点。No single ledger, so it stays a pilot forever.

Our Vision

让每一家企业,都长出自己的 AI 组织Every company grows its own AI organization

人定方向和标准,数字员工成建制执行。Evose 的全部产品,都为这一场转型而造。People set direction and standards; digital employees execute at scale. Everything Evose builds serves this one transition.

Who We Are · Our Own Experiment

我们自己的 AI Native 实验Our own AI-native experiment

01 / 公司人机比01 / Human-to-agent ratio

1 : 71 : 7

20 个人 × 150+ 数字员工,生产环境在岗20 people × 150+ digital employees, on the job in production

02 / 人效02 / Output per person

3 倍3×

20 个人,撑起过去六七十人的盘子20 people carrying what used to take sixty or seventy

03 / 上手周期03 / Ramp-up time

1 周1 week

经验和工作流都沉在知识库里,新人一周就能上手Know-how and workflows already sit in the knowledge base — a new hire is productive in a week

04 / AI CEO04 / An AI CEO

十个智能体管着这家公司,也治理着公司的知识库Ten agents run this company — and govern its knowledge base

每天 1 小时1 hr/day

喂库、纠偏feeding and correcting the base

战略分析Strategy竞品研究Competitors客户洞察Customers经营分析Analytics决策复盘Retros融资材料Fundraising定价条款Pricing招聘助理Recruiting项目管理Projects会议整理Meetings
Imagine · The AI-Native Org

第三阶段的公司,长这样This is what a stage-three company looks like

上下文是核心：人给判断、给经验、做训练,调度和执行交给智能体Context is the core: people supply judgment, experience and training — scheduling and execution run on agents

战略Strategy

标准Standards

预算Budget

取舍Trade-offs

决策层Decision layer 人最少 · 权最重Fewest people · final say
组织上下文中心Organizational context hub
目标与约束Goals & limits业务事实Business facts任务与进展Tasks & progress成果与决策Outcomes & decisions知识与规范Knowledge & rules案例与复盘Cases & retros
按权限共享：谁能看什么,一条一条划Shared by permission — who sees what is set line by line
内部系统Internal systems

HR · ERP · CRM · 工单 · 财务HR · ERP · CRM · ticketing · finance

业务数据Business data

流量 · 转化 · 成本 · 客户行为Traffic · conversion · cost · behaviour

增长与营销 · 示例流程Growth & marketing · sample flow

把一个热点,变成一场投放Turning one trend into a live campaign

1 热点线索摘录Trend brief 热点发现Trend scout
2 内容与素材成稿Copy and creative 内容工厂Content factory
3 人工把关Human sign-off 预算与投放口径,等主管确认Budget and targeting — waiting on the lead 通过Approve拒绝Reject
4 多渠道排期投放Multi-channel scheduling 投放素材Ad creative

投放预算Budget 已拟定 · 待主管确认drafted · awaiting the lead

素材合规Creative compliance 已自检 · 待终审self-checked · awaiting final review

落地页Landing page 已就绪 · 可发布ready · clear to publish

增长与营销Growth & marketing
热点发现Trend radar内容工厂Content factory投放素材Ad creative发布上线Publishing
训练trains 审核reviews
增长主管Growth lead增长专员Growth associate

销售与交付Sales & delivery · 待办Queue 123/3

需要人工把关的决策Decisions that need a human

方案与报价结构Proposal and pricing 已通过Approved已通过Approved

客户背景、竞品对比、报价结构由智能体出齐,销售主管确认口径。Background, comparisons and pricing come from the agents; the sales lead confirms the line.

折扣线与付款条件Discount floor and terms 待主管审批Awaiting the lead已通过Approved

让到哪一步、怎么收款,超出折扣线必须销售主管签字。How far to concede and how to collect — past the floor it needs the sales lead’s signature.

通过Approve拒绝Reject
交付验收与回款Acceptance and collection 待执行Queued已完成Done

验收标准和回款节点,由交付负责人确认后关单。Acceptance criteria and payment milestones — the delivery lead closes it out.

销售与交付Sales & delivery
线索研究Lead research客户画像Account profiling方案生成Proposals交付验收Delivery checks
训练trains 审核reviews
销售主管Sales lead销售专员Sales associate

客服与运营 · 示例流程Support & operations · sample flow

把一通投诉,变成一次召回Turning one complaint into a win-back

1 工单归因摘录Ticket root cause 工单回复Ticket agent
2 服务记录质检Conversation QA 会话质检QA agent
3 人工把关Human sign-off 补偿方案,等主管确认Compensation — waiting on the lead 通过Approve拒绝Reject
4 召回名单与话术Win-back list and scripts 分层运营Segmented ops

补偿金额Compensation 已测算 · 待主管确认calculated · awaiting the lead

客户身份Customer identity 已核验 · 权益已对齐verified · entitlements aligned

召回话术Win-back script 已定稿 · 可下发final · clear to send

客服与运营Service & user ops
工单回复Ticket replies会话质检Chat QA分层运营Tiered user ops主动召回Win-back outreach
训练trains 审核reviews
运营主管Ops lead运营专员Ops associate

风控与合规Risk & compliance · 待办Queue 123/3

需要人工把关的决策Decisions that need a human

身份核验与异常筛查Identity checks and screening 已通过Approved已通过Approved

全量筛查由智能体完成,命中项转人工复核。Agents screen every case; hits go to a human for review.

账户冻结与放行Freeze or release 待主管审批Awaiting the lead已通过Approved

涉敏账户怎么处置,风控主管拍板。How a flagged account is handled — the risk lead calls it.

通过Approve拒绝Reject
规则变更留痕Rule changes on the record 待执行Queued已完成Done

为什么改这一条规则,写进可审计的链路。Why the rule changed, written into an auditable trail.

风控与合规Risk & compliance
身份核验Identity checks异常监测Anomaly watch合规预审Compliance pre-check留痕审计Audit trail
训练trains 审核reviews
风控主管Risk lead风控专员Risk associate
Two Halves

企业 AI 转型的上下半场The two halves of an AI transformation

上半场First half

场景突破Use-case
breakthroughs

突破增长 · 降本提效Growth · Cost and speed

谁下场Who runs it 业务负责人自己带,对结果背责——不是 IT,不是 AI 小组The business owner leads it and answers for the result — not IT, not an AI task force

下半场Second half

组织重构Organizational
rebuild

企业知识库 · 组织架构Knowledge base · Org structure

谁下场Who runs it 一把手,没有替代方案The CEO. There is no substitute.

Gallup 调研统计：主管自己带头用 AI,团队真用起来的比例是不带头的 8 倍(33% vs 4%)。Gallup: when the manager personally champions AI, adoption runs 8× higher than when they do not (33% vs 4%).

上半场 · 突破增长First half · Break growth open

上半场：优先突破增长First half: break growth open first

01

出海扩张Going global

一个市场跑通,复制到下一个Crack one, copy to the next

卡在哪Stuck on 打法在老员工脑子里,带不走The playbook lives in a few heads
怎么做How
  • 判断沉淀成智能体Distil judgment into agents
  • 换市场,只换本地策略Swap only local tactics
  • 相似市场,整套复用Similar markets reuse it all

复制的是理念,不是做法Copy the thinking, not the steps

02

线上获客Online leads

一个年轻人,顶一个内容团队One junior, one content team

卡在哪Stuck on 靠线下渠道,自媒体没做起来Offline only; media never took off
怎么做How
  • 智能体出稿、分发全平台Agents write and post
  • 年轻人把调性、回评论A junior owns the tone
  • 按线索数据,每周调Tune weekly by leads

一整个板块,从零做起来A whole channel, built from zero

03

人头瓶颈Headcount cap

靠铺人的生意,不再靠铺人Scale without adding people

卡在哪Stuck on 销售、内容电商,人越多卖越多Sales, content commerce: more people, more sales
怎么做How
  • 销冠打法,沉淀成智能体Turn top sellers into agents
  • 新号、新渠道,直接复制New channel? Copy it
  • 新人上手,不靠老人带No shadowing for new hires

想做多少做多少,不用先招人No hiring before you scale

我们的体会What we have learned

AI 只放大已经验证过的模式。没验证过的,得人先趟出来。AI only scales a model that is already proven. An unproven one still needs people to prove it first.

上半场 · 降本提效First half · Cut cost, move faster

上半场：降本提效First half: cut cost, move faster

01

客户服务Customer service

问题不再在部门之间转来转去One answer, no runaround

包括Covers 客服答疑、工单处理、售后跟进Q&A, tickets, after-sales
卡在哪Stuck on 答不了、办不了,就层层转交Support → ops → tech
怎么做How
  • 常见问题,智能体直接答完Agents answer FAQs
  • 工单自动分类、带上下文派单Tickets sorted and routed
  • 新问题每周补回知识库Docs updated every week

人只接真正难的问题People take only the hard ones

02

经营分析Business reporting

拼表、写周报,可以不做了No more stitching reports

包括Covers 经营报表、周报、项目复盘Reports, weeklies, retros
卡在哪Stuck on 各部门交数、对口径、层层汇总Teams submit, reconcile, stitch
怎么做How
  • 智能体直接从系统取数Agents pull from the systems
  • 统计口径写进知识库Definitions live in the docs
  • 异常自动标出,推给负责人Anomalies flagged to owners

省的不是时间,是一整个环节Not faster — the step is gone

03

风控合规Risk & compliance

从抽检到全量质检From spot checks to full checks

包括Covers 合同审核、内容质检、合规检查Contracts, content, policy
卡在哪Stuck on 人手只够抽查,出了事才知道Only a sample gets checked
怎么做How
  • 智能体逐条按规则过一遍Agents check every item
  • 有问题的标出来,写清原因Flags come with the reason
  • 法务、风控只看被标的Legal and risk see only flags

风险在出事前就被看见Risk is seen before it lands

我们的体会What we have learned

回去算笔账：一件事真正干活的时间占几成？剩下的都耗在部门交接上——提效,先砍交接。Go back and measure: how much of a task is actual work? The rest is handoffs between teams — cut those first.

下半场 · 知识治理Second half · Knowledge

下半场：搭建企业知识库Second half: build the company knowledge base

01

蒸馏人的认知Distil what people know

给人看的For people 会议录音、聊天记录、老员工脑子里的判断Meeting recordings, group chats, the judgment inside veterans’ heads
给 AI 检索的For AI 什么情况下怎么判断 · 底线在哪 · 谁说了算When to judge which way · where the line is · who decides

02

蒸馏公司的决策逻辑Distil how the company decides

给人看的For people 报表、数据看板、项目复盘Reports, dashboards, project retros
给 AI 检索的For AI 这个数为什么是这个数 · 当时为什么这么定 · 后来为什么改Why the number is what it is · why it was decided then · why it changed later

我们的体会What we have learned

堆文档不叫知识库——给人看的报表,要蒸馏成 AI 读得懂的判断和经验。Piling up documents is not a knowledge base. Distilling the reports and records written for people into judgment an AI can read — that is.

下半场 · 组织架构Second half · Org structure

下半场：重建组织架构Second half: rebuild the org

1 组织单元Unit of work

最小单元：从岗位变成任务From roles to tasks

先设岗再招人,按技能串联接力Set roles, hire, relay by skill发布任务,按能力并行认领Publish tasks, claim in parallel

BuilderBuilder 拆解任务、搭建上下文,把经验做成智能体Breaks down tasks, builds the context
ReviewerReviewer 认领结果,做判断与取舍,对质量负责Claims the outcome, judges, signs off
AgentAgent 认领执行,7×24 运转Claims the execution, 24/7

认领任务的是「认知节点」——可以是人,也可以是智能体Tasks are claimed by cognitive nodes — people or agents

2 协作方式How teams work

部门内端到端闭环Close the loop inside the team

跨部门接力,靠开会对齐Cross-team relay以 AI 为中心,人机共生AI at the center

AIAI 接管组织运营:派活、流转、对齐、沉淀Runs operations: routing, handoffs, alignment
人People 专注认知共创:提出问题、判断取舍Co-create insight: frame problems, judge
其他部门Others 把规则和专业能力放进上下文,随时被调用Put rules and skills into the shared context

一场活动接力五个部门 → 一个部门跑完全程Five-team relay → one team runs it all

我们的体会What we have learned

AI 接管了组织的运营,组织比拼的只剩一件事:认知升级的速度。AI now runs the operations; what is left to compete on is how fast the organization learns.

观点参考:曾鸣《智能:AI 时代的商业、组织与战略的本质》(2026)Drawing on Zeng Ming, Intelligence (2026)

Case Study 01 · Finance

客户案例：全球化交易所的运营体系重构Case study: a global exchange rebuilds its operations

全球前十大交易所之一 · 私有化部署 · 千人组织——第一刀切在最痛的全球营销,三个月铺满整个运营体系One of the world's top-10 exchanges · self-hosted · 1,000 people — the first cut was their most painful scenario; the whole operation in three months

01 · Operations & Growth

运营与增长Ops & growth头号场景No.1 scenario

全球活动策划Global campaign planning SEO 与内容获客SEO & content acquisition 社媒与社群运营Social & community ops 用户增长与留存User growth & retention 竞品动态追踪Competitor tracking

02 · Channel Ops

渠道运营Channel ops

全球本地代理发现Local agent discovery 渠道触达与建联Channel outreach 渠道运营管理Channel management 区域与渠道分析Region & channel analytics

03 · Risk & Security

风控与安全Risk & security

提币人脸风控Withdrawal face checks 资金异常监测Fund anomaly detection KYC 资料审核KYC review 反薅羊毛Anti-farming 可疑交易识别Suspicious-activity detection

04 · Research

研究与投研Research

潜力项目挖掘Project scouting 上币尽调Listing due diligence 币种热度监控Token heat tracking KOL 与舆情追踪KOL & sentiment tracking

05 · Analytics

数据分析Analytics

经营大盘分析Business dashboards 活动效果分析Campaign analytics 渠道与区域分析Channel & region analytics 留存与成本分析Retention & cost analytics

06 · Support

客服与用户运营Support & user ops

工单辅助回复Ticket assistance 申诉质检Appeals QA VIP 服务VIP service
治理层贯穿全程Governance runs through everything

权限管控 · 行为审计 · 合规卡点 · 全链路可观测 · 成本与 ROI 可见Permission control · behavior audit · compliance gates · full observability · visible cost & ROI

底座 · 打通Foundation · connected

内部业务数据 · 内部知识文档 · 组织权限与身份——让 Agent 安全地用上公司自己的数据与系统Business data · internal docs · org identity & permissions — agents safely use the company's own data and systems

Case Study 02 · Consumer

客户案例：年销 40 亿内容电商品牌的 AI 转型Case study: an RMB 4B content-commerce brand goes AI-first

私有化打通数据与权限,从最痛的内容生产切入——上线 1 个月,约 1000 人日常使用,铺满整条运营链路Self-hosted, with data and permissions wired in; started from the worst pain — content production. One month in: ~1,000 daily users across the whole chain

01 · Merchandising

选品Merchandising

趋势捕捉Trend scouting 爆品复盘Bestseller retros 竞品与价格监控Competitor & price watch 选品推荐Pick recommendations 评论需求洞察Review insights

02 · Content Planning

内容策划Content planning头号场景No.1 scenario

选题策划Topic planning 脚本与文案Scripts & copy 卖点与科普Selling points & explainers 多平台版本适配Per-platform versions

03 · Creative

图片与视频Creative

商品图与海报Product images & posters 短视频脚本Short-video scripts 分镜与剪辑辅助Storyboard & edit assist 直播话术与切片Livestream scripts & clips

04 · Influencer

达人营销Influencer marketing

达人发现与筛选Discovery & vetting 达人—商品匹配Creator–product match 专属简报与物料Custom briefs & kits 触达建联跟进Outreach & follow-up 带货效果复盘Sales retros

05 · Paid Growth

投放与增长Paid growth

投放素材批量生成Ad creative at scale 投放数据监控Spend monitoring ROI 归因ROI attribution 人群与货品匹配Audience–product match

06 · CS & Ops

客服与经营CS & operations

售前咨询Pre-sales chat 评论管理与回复Review management 经营日报Daily business reports 活动效果Campaign results 异常归因Anomaly attribution
合规审查Compliance review 贯穿全环节Every step
宣称红线预审Claims pre-check 广告法审查Ad-law review 平台规则核查Platform rules check 资质与留痕Credentials & records
选品、内容、投放——每一步先过合规Picks, content, ads — everything clears compliance first
底座 · 打通Foundation · connected

产品与配方知识 · 爆款内容样本 · 达人画像档案 · 合规规则库 · 电商经营数据Product & formula knowledge · winning-content samples · creator profiles · compliance rulebook · commerce data

The Receipts

花多少钱,换来什么What it costs, what it returns

200 美金 / 人 / 月$200 / person / mo

大约是薪资的 10%About 10% of salary

10 万 / 月$100K / mo

那家交易所,500 人天天在用The exchange — 500 daily users

20 万+ / 月$200K+ / mo

我们所有客户加起来All our clients combined

增量Growth

流量 5 倍5× traffic

某全球化内容平台A global content platform

  • 3 倍 同样预算能做的增长动作3× the growth moves on the same budget
  • 8,600 次/周 交易所风控从抽查变全查8,600/wk exchange risk checks — from sampling to all

降本Savings

编制 −14%−14% headcount

那家千人交易所,盘子没小The 1,000-person exchange — same business

  • +80% 整体运营效率+80% overall ops efficiency
  • 10% AI 出设计,成本只有外包的一成10% — AI design at a tenth of outsourcing
The Agent Factory

一个底座,
长出千行百业的数字员工One foundation.
Digital employees for any industry grow out of it.

换行业,不换底座——只换知识库和「岗位说明书」。新行业的数字员工,以「周」为单位孵化交付。Change the industry, not the foundation — you swap the knowledge base and the job description. Agents for a new industry ship in weeks, not quarters.

行业Industries

金融 / 交易所Finance & exchanges

电商 / 消费品牌E-commerce & consumer

SaaS 与科技公司SaaS & technology

媒体与内容平台Media & content

游戏与泛娱乐Gaming & entertainment

专业服务 / 连锁Professional services

×
岗位Roles

市场Marketing

选题策划 · 文案生成
多语言本地化 · 合规审查Campaign planning · Copywriting
Localization · Compliance review

客服Support

7×24 工单回复 · 会话质检
VIP 服务 · 申诉处理24/7 first response · QA on every case
VIP handling · Escalations

风控Risk

身份核验 · 欺诈监控
政策审查 · 反滥用Identity checks · Fraud detection
Policy review · Abuse prevention

研究Research

竞品追踪 · 线索挖掘
舆情监控 · 每日简报Competitor tracking · Lead sourcing
Sentiment · Daily briefings

经营Analytics

经营大盘看板 · 活动效果
留存与成本报表Business dashboards · Campaign impact
Retention & cost reporting

渠道Partnerships

伙伴发掘 · 触达建联
区域分析 · 效果复盘Partner discovery · Outreach
Regional analysis · Post-mortems

Platform Architecture

Evose 的产品架构The Evose platform architecture

evose.ai
Evose Platform
员工每天用Staff use daily
AI
工作台AI Workspace

建数字员工Build digital employees

说人话建岗,不写代码,当天上岗——像招一个新同事Describe the job in plain words — no code, on the job same day, like hiring a new colleague

搭业务工作流Assemble workflows

多个数字员工接力干活,一句话触发整条链路Digital employees work in relay; one sentence triggers the whole chain

沉淀企业知识库Grow the knowledge base

员工产出持续沉淀,人走,经验留下Output keeps accruing — people leave, know-how stays

团队协作空间Team workspace

人和数字员工同屏配合,分工进度全员可见People and digital employees side by side; everyone sees who's doing what

老板每天看Bosses check daily
AI
管控中枢AI Control Center

权限与规则Permissions & rules

能看什么、能做什么,越权当场拦What it sees, what it does — overreach blocked on the spot

安全与审计Security & audit

危险动作先审批,全程留痕可追责Risky actions wait for approval; every step logged, accountable

成本与预算Cost & budget

部门/岗位/人三级封顶,超支自动停Caps by department / role / person; auto-stop on overrun

产出与追踪Output & tracking

每个数字员工干多少、值不值,一屏看清What each digital employee does, and whether it's worth it — one screen

IT 维护IT maintains
AI
GatewayAI Gateway

模型 GatewayModel Gateway

贵模型只干重活 · DeepSeek / Qwen / OpenAI / Anthropic / GeminiPricey models only do heavy lifting · DeepSeek / Qwen / OpenAI / Anthropic / Gemini

数据 GatewayData Gateway

企业数据统一进出口,接上组织架构,AI 才懂你的公司One gate for company data, wired to your org chart — that's how AI understands your company

系统 GatewaySystem Gateway

API / MCP 标准协议接你现有系统,不换软件API / MCP standards into the systems you already run — no rip-and-replace

工具与 Skill 库Tools & Skill library

全公司统一注册、统一管控Registered once, governed centrally

The Landscape

跟市面主流产品的差别,一张表看清How Evose differs from the mainstream — one table

具备Yes △部分 / 个人级Partial / personal-grade 不具备No
产品Product 定位Positioning 智能问答Q&A 自主干活
(做 PPT、Word)Gets work done
(slides, docs) 搭数字员工Build agents 业务工作流Workflows 企业知识库Knowledge base 组织协作Team collaboration 企业治理
(权限·审计·成本)Governance
(access · audit · cost)

ChatGPT、豆包ChatGPT, Doubao

对话助手Chat assistant

△ △

Manus、GensparkManus, Genspark

通用智能体General agent

Dify、n8nDify, n8n

工作流编排Workflow orchestration

△

WorkBuddy、CodexWorkBuddy, Codex

办公助手Office copilot

△ △

EvoseEvose

企业协作与治理Enterprise collaboration & governance

前面几项是给个人的,后面三项才是给组织的——我们是从后面三项开始做的。The first columns are for individuals. The last three are for the organization — that is where we started.

上岗第 1 步 · 建岗On the Job · Step 1 · Build

搭建数字员工Build a digital employee

说人话建岗,当天上岗Describe the job in plain words — on the job the same day

智能体 / AI 创建Agents / AI Create
Evose
B

我想建一个「活动策划专家」——给它一个节日或热点,它能出一套完整的营销活动方案。I want a "Campaign Strategist" — give it a holiday or a trend, and it produces a full campaign plan.

我先理解需求,然后一步步把这个岗位配起来。Let me understand the need, then configure the role step by step.

① 写提示词Write the prompt 大白话 → 专业岗位说明书plain words → a real job spec
② 加载技能Load skills 活动方案框架 · 竞品活动分析campaign framework · competitor teardown
③ 加载工具Load tools 小红书 · 抖音 · 联网搜索 · 网页读取RED · Douyin · web search · page reader
④ 关联知识库Attach knowledge 营销活动历史案例库past campaign case library
⑤ 关联模型Pick the model Claude Sonnet 4.6Claude Sonnet 4.6

活动策划专家已创建。右边可以直接改,也可以进配置继续调。Campaign Strategist created. Edit it on the right, or open setup to keep tuning.

策C

活动策划专家Campaign Strategist

草稿已保存Draft saved

把一个节日或热点,变成可执行的营销活动方案：主题、玩法、素材清单与预算建议。Turns a holiday or trend into an executable campaign: theme, mechanics, asset list and budget.

营销Marketing 活动策划Campaigns 增长Growth
提示词Prompt 能力 6Skills 6 知识 3Knowledge 3 悬停切换hover

# 角色# Role

你是 {{brand}} 的活动策划专家,负责把一个节日、热点或品牌节点,拆成可执行的营销活动方案。流程严格分两步:先调研,再出案——没有调研支撑的创意,不要写进方案。You are the campaign strategist for {{brand}}. You turn a holiday, a trend or a brand moment into an executable plan. Work in two strict phases: research first, then propose — never put an idea in the plan that research cannot back.

## 一、调研阶段(必须先做)## 1 · Research phase (always first)

出现以下情况必须调用工具,不要凭记忆硬写：Call tools instead of relying on memory whenever:

· 判断热点当前热度与生命周期 → 小红书搜索 / 抖音搜索· you must judge a trend’s heat and life cycle → RED / Douyin search

· 竞品最近在做什么活动 → 联网搜索 + 网页读取原文· competitors’ current campaigns matter → web search + page reader

· 目标人群怎么说 → 抓评论区原话,不要转述· the audience’s own words matter → pull raw comments, don’t paraphrase

字符 2,847 · 变量 ×22,847 chars · 2 variables {{brand}} {{budget}}

工具 · 4Tools · 4

小红书搜索RED search 抖音搜索Douyin search 联网搜索Web search 网页读取Page reader

技能 · 2Skills · 2

活动方案框架Campaign framework 竞品活动分析Competitor teardown

营销活动历史案例库Past campaign library

38 篇往期活动复盘 · 带效果数据38 retros, with results

品牌调性与话术规范Brand voice guide

写出来就是自家的味道so it sounds like you

广告法与平台红线清单Ad-law & platform rules

出方案时自动避开avoided automatically

「AI 创建」流程示意 · 一句大白话,AI 逐步把岗位配起来：写提示词 → 加载技能 → 加载工具(小红书/抖音/联网)→ 关联知识库 → 选模型。这个岗位,就是后面那条营销活动工作流的第一棒"AI Create," illustrated · one plain sentence and AI configures the role step by step: write the prompt → load skills → load tools (RED / Douyin / web) → attach knowledge → pick the model. This role is the first leg of the campaign workflow you'll see next

一套配置,就是一个岗位One config = one role

大模型 + 提示词 + 技能 + 知识库,四样配齐即上岗——不写一行代码。Model + prompt + skills + knowledge base: four pieces and it's on the job — zero code.

提示词,一键优化One-click prompt polish

你用大白话写目标,AI 展开成专业岗位说明书——AI 先写,你当审核。Write the goal in plain language; AI expands it into a professional job description — AI drafts, you review.

智能体广场,现成模板Agent gallery, ready-made

市场分析师、行情专家、文案写手……一键复用,少数人建、全组织用。Market analyst, research expert, copywriter… reuse in one click — a few build, the whole org uses.

上岗第 2 步 · 干活On the Job · Step 2 · Run

工作流Workflows

一句话,一支小队把活干完——像给团队定 SOPOne sentence, and a small squad finishes the job — like writing an SOP for your team

工作流 / 全球营销活动上线Workflow / Global campaign launch
运行中 · 并行 17 路Running · 17 in parallel 17 个市场 · 已发布17 markets · live

活动策划Campaign plan

campaign_briefString

marketsArray<String>

brand_guideString

批处理Batchmarkets

17 个市场 · 并行17 markets · in parallel
ENJAKOESFRDEPTARRUTHVIIDTRITNLPLHI

文案翻译Copy translation

多语文案 · 品牌调性localized copy · on-brand

失败重试 3 次retry ×3

翻译质检Translation QA

代码校验术语库glossary check in code

Javascript

活动海报Campaign poster

按市场出图 · 多尺寸per-market, all sizes

失败重试 3 次retry ×3

开发落地页Landing page

本地化页面 · 可发布localized & publishable

失败重试 3 次retry ×3
JA
KO
···

同一条链路,17 个市场的物料同时在做One chain, 17 markets' assets produced at once

B

人工审核Human review

17 个市场 · 终审17 markets · final gate

等待审批Awaiting approval已批准Approved

活动发布Campaign live

result 批处理 / resultbatch / result

17 个市场 · 同步上线17 markets · live together

工作流示意 · 一场全球营销活动：活动策划 → 批处理按 17 个市场并行(文案翻译 → 代码校验术语库 → 活动海报 → 开发落地页)→ 人工审核 → 活动发布。AI 干活,人只在最后一道关口拍板Workflow, illustrated · one global campaign: plan → batch fans out to 17 markets in parallel (translate → glossary check in code → poster → landing page) → human review → go live. Agents do the work; the human decides at the final gate

可视化画布Visual canvas

拖拽编排,支持条件分支、循环、意图识别——业务流程长什么样,画布就长什么样。Drag-and-drop with branches, loops and intent detection — the canvas looks exactly like your process.

批处理并行Batch in parallel

一个批处理节点,17 个市场的物料同时做;失败自动重试,夜里也在跑。One batch node produces assets for 17 markets at once; failures retry themselves, and it runs overnight.

人工卡点Human gates

审批节点插在任意环节——关键动作停下来,等人审批。Drop an approval gate anywhere — critical actions stop and wait for a human call.

营销客户用同一张画布,把全球活动上线从 4 天压到 2 小时。One marketing client used a single canvas to cut global campaign launches from 4 days to 2 hours.

上岗第 3 步 · 协作On the Job · Step 3 · Co-work

多人协作Co-working

排好的活走工作流,没排过的活在这里干Scripted work runs in workflows; everything else happens here

同一个上下文One shared context

文件、对话、任务,人和 AI 看同一份Files, chats, tasks — people and AI see the same thing

随时组队,不用编排Assemble on the fly

@ 一下数字员工,就开工@ a digital employee and start

一人建,全组织用Build once, everyone uses

沉淀在组织里,不在个人账号里Lives in the org, not in personal accounts

谁在干什么,一屏可见Everyone visible, one screen

不用追问「做到哪了」No more "where are we on this?"

市场营销 / 任务 · 看板Marketing / Tasks · Board
B
7 名成员 · 4 个数字员工7 members · 4 AI
待办To do 32

活动上线发布Campaign launch

待启动To startB

投放数据监控Ad spend monitoring

待启动To start

社群预热内容Community warm-up

待接手Unassigned已接手Picked up
进行中In progress 3

KOL 合作名单与报价KOL shortlist & quotes

询价中PricingB

投放素材批量生成Ad creative at scale

执行中Running

多语言物料制作Multilingual assets

待人审批Awaiting sign-offBingo 已批准Approved by BingoB
已完成Done 34

上期活动复盘Last campaign retro

已交付DeliveredB

热点追踪与选题Trend watch & topics

已交付Delivered

活动主题与预算确认Campaign theme & budget

已批准ApprovedB

电商客户 1 个月约 1000 人用起来,靠的就是这一层。This layer is how an e-commerce client got ~1,000 people on board within a month.

上岗第 4 步 · 管控On the Job · Step 4 · Govern

企业管理台The admin console

一张看板,管住全公司的 AIOne dashboard to govern all the AI in the company

成本 · 近 7 天 · 全组织Cost · last 7 days · org-wide 导出报表Export
本期消耗Spend16%
$1,019 近 7 天 · 全组织last 7 days · org-wide
总 TokenTotal tokens
1.4B 近 7 天last 7 days
调用次数Calls
26,516 近 7 天last 7 days
预算使用Budget used
68% 部门 / 岗位 / 人三级封顶capped by team / role / person
08-2308-2508-2708-29

成本看板示意 · 本期消耗、环比、Token 与调用次数一屏看清——可按部门 / 成员 / API Key 下钻Cost dashboard, illustrated · spend, trend, tokens and calls on one screen — drillable by department / member / API key

应用追踪 · 智能体消耗 Top 5App tracking · top 5 agents by spend 实时Live
智能体Agent 消耗Spend 调用Calls 状态Status
线索研究助手Lead research$68.4312合理Normal
内容工厂Content factory$52.11,204合理Normal
竞品监测Competitor watch$41.796超预算Over budget
工单质检Ticket QA$23.52,860合理Normal
海报生成Poster generator$18.2430待观察Watch

应用追踪示意 · 每个智能体花了多少、被调用多少次;超预算和异常增长自动标出来App tracking · what each agent spends and how often it runs; over-budget and abnormal growth are flagged automatically

用量趋势 · Token 消耗Usage trends · token burn

用量与成本Usage & cost

每个数字员工花多少、干多少——对话次数与消耗一屏看清,月底考核。What each digital employee spends and delivers — conversations and burn on one screen, reviewed monthly.

成员与部门 · 角色权限Members & departments · roles

权限管理Permissions

人和 Agent 按部门统一管,能看什么、做什么——越权当场拦。People and agents managed by department — what they see, what they do; overreach blocked on the spot.

订阅管理 · 计费Subscriptions · billing

预算与配额Budgets & quotas

部门、岗位、个人三级封顶——超了自动停,账单不再失控。Caps at department, role and individual level — auto-stop over the line; no runaway bills.

产出 · 追踪Output · tracking

产出与追踪Output & tracking

哪个智能体在干活、花了多少、产出什么——超预算自动预警,异常当场标出来。Every step logged — who did it, who approved it; nothing goes public without a human yes.

头部客户对比完市面主流产品后选择我们——决定性差异,就在这一层。A leading client chose us after comparing every mainstream product — this layer was the deciding difference.

Security & Compliance

数据安全,靠这三件事Three things that keep your data safe

01

私有化部署Self-hosted deployment

数据不出你的机房Your data never leaves your building

  • 交付 Docker 镜像包,跑在你的机房或私有云Delivered as a Docker image, running in your data centre or private cloud
  • 无遥测、无业务数据回传(调用外部工具市场除外)No telemetry, no business data sent back (except calls to the external tool market)
  • 每个任务在独立沙箱里跑,跑完即销毁Every task runs in its own sandbox and is destroyed when it finishes

02

自定义模型通道Your own model channel

模型不经 Evose 之手The model never passes through Evose

  • 模型由你选,在你的内网调用You choose the model; it is called inside your own network
  • Evose 不经手任何推理数据Evose never touches the inference data
  • 敏感任务可锁定到你指定的模型通道Sensitive tasks can be pinned to a model channel you designate

03

安全审计Security audit

每一步都有记录Every step is on the record

  • 谁、哪个数字员工、动了什么、花了多少,全程留痕Who, which digital employee, what was touched, what it cost — all logged
  • 按部门、角色、工作空间授权,越权当场拦Access by department, role and workspace; anything out of scope is blocked on the spot
  • 高风险动作先审批,对外发布必须人点头High-risk actions need approval first; anything published externally needs a human yes

SOC 2 · ISO 27001 · GDPR 认证进行中;DPA、SLA、审计权可写进合同SOC 2 · ISO 27001 · GDPR certifications in progress; DPA, SLA and audit rights can go into the contract

已在千人规模金融机构的生产环境运行一年以上Running in production at a 1,000-person financial institution for over a year

带走这一句One thing to take with you

工具是买来的,
组织是长出来的Tools are bought.
Organizations are grown.

别停在买工具,去长你自己的 AI 原生组织Stop at buying tools and you stop. Grow your own AI-native organization

动手Get started

别等看懂了再开始Don’t wait until you understand it

现在 · 十分钟Right now · ten minutes

1

注册Sign up

打开 evose.ai,一分钟Open evose.ai — one minute

2

说一句话Say one sentence

你想要什么岗位,用大白话讲Describe the role you want, in plain words

3

它就上岗了It starts working

不写代码,不装东西,十块钱以内No code, nothing to install, under ten yuan

回去 · 第一周Back at the office · week one

1

挑一件事Pick one thing

最烧钱、最费人的那一件,别贪多The one that burns the most money and people — nothing else

2

定一个人Name one person

业务一把手挂帅,3–5 人先锋队,不丢给 ITA business leader out front, a squad of three to five — not IT

3

算一笔账Do the math

现在花多少人、多少时间、多少钱,90 天后对着验收People, time and money today — check it against day 90

今天在座的,谁搭出来了,发给我看看——搭不出来的,我帮你搭。Whoever gets one running today, send it to me. If you cannot, I will build it with you.

evose.ai

1 / 22
幻灯片总览All slides
点击跳转 · 按 Esc 关闭Click a slide to jump · Esc to close

快捷键Keyboard shortcuts

下一页(点击画面也可)Next slide (or click anywhere)→ / Space / PgDn
上一页Previous slide← / PgUp
第一页 / 最后一页First / last slideHome / End
总览Overview gridO
切换语言Cycle languageL
全屏FullscreenF
跳到章节Jump to chapter1 – 9
关闭面板Close panelsEsc