Claudeforce評測:Salesforce把整套CRM搬進Claude | Claudeforce Review: Salesforce's CRM Moves Into Claude
By Kit 小克 | AI Tool Observer | 2026-08-29
🇹🇼 Claudeforce評測:Salesforce把整套CRM搬進Claude
Claudeforce 是 Salesforce 與 Anthropic 在 2026 年 8 月 26 日聯手發布的新合作案,簡單講就是把 Salesforce 的 CRM 系統整套塞進 Claude 裡,讓業務員不用開 Salesforce App,直接在 Claude 裡問「這筆案子卡在哪」就能拿到答案、甚至直接動手改資料。消息公布後股價在盤後交易一度跳漲 14%,市場顯然買單,但實際能用的東西目前還很少。
Claudeforce到底裝了什麼
這次合作分兩個方向走:
- Salesforce in Claude:一個 CoWork 外掛,內建 37 個預先做好的銷售技能,像是會議前準備、案件健檢、業績管道檢視,全部用 Claude 的推理能力跑
- Claude in Salesforce:Claude 透過 Amazon Bedrock 進駐 Agentforce,跑在 Salesforce 自家的 Trust Boundary 裡,符合受監管產業的資安要求
- Slack 整合:Claude 變成 Slackbot 和 Slack Code 的預設模型,Salesforce 自己表示內部 83% 員工已經在用
現實:離「不用再開Salesforce App」還很遠
Salesforce 執行長 Benioff 講得很滿:「這波 SaaSpocalypse 唱衰論,我覺得該停了」,擺明是在回應「AI 助理會直接吃掉 SaaS 介面」的市場焦慮。但拆開時程表看,Salesforce in Claude 目前只開放給「精選試點客戶」,公開 Beta 要等 9 月才上線,更多技能得等到 2026 年底才會陸續補齊。換句話說,發布會上講的完整版本,現在還沒人能真的摸到。
Claudeforce對企業真正的風險:定價權旁落
比功能清單更值得留意的,其實是商業模式的變化。Salesforce 過去靠「每人每月收授權費」賺錢,Claudeforce 這種用量制的 AI 互動模式,很可能比它自己想像的更快侵蝕原本的座位授權收入——而且介面主導權一旦轉移到 Claude 這一端,議價籌碼也會跟著往 Anthropic 傾斜。Salesforce 今年稍早已經把對 Anthropic 的投資估值抬到 50 億美元,這次等於是主動把自己最值錢的資產——CRM 資料與工作流——接上別人家的大腦。
對用 Salesforce 的業務團隊來說,現階段能做的就是先觀望:功能還在試點階段,真要導入至少等 9 月開放 Beta 之後,看實際上手體驗跟資安審核結果,再決定要不要跟進。
好不好用,試了才知道。
🇺🇸 Claudeforce Review: Salesforce's CRM Moves Into Claude
Claudeforce is the new partnership Salesforce and Anthropic announced on August 26, 2026 — in short, Salesforce is folding its entire CRM into Claude, so sales reps can ask Claude "why is this deal stuck" instead of opening the Salesforce app, and even take action right from the chat. The stock jumped as much as 14% in after-hours trading the day it dropped. The market bought it. What is actually usable today is a lot thinner than the pitch.
What Claudeforce Actually Ships
The deal runs two directions:
- Salesforce in Claude: a CoWork plugin with 37 prebuilt sales skills — meeting prep, deal health checks, pipeline review — all running on Claude reasoning
- Claude in Salesforce: Claude plugs into Agentforce via Amazon Bedrock, running inside Salesforce own Trust Boundary for regulated industries
- Slack integration: Claude becomes the default model behind Slackbot and Slack Code — Salesforce says 83% of its own workforce already uses it
Reality Check: "You Will Never Need the App Again" Is Not Live Yet
CEO Marc Benioff did not hold back: "This nonsense of the SaaSpocalypse, I think it is time for it to stop" — a direct answer to fears that AI assistants will swallow the SaaS interface layer entirely. But check the actual rollout: Salesforce in Claude is live for "select pilot customers" only, open beta does not land until September 2026, and most of the promised skills will not ship until late 2026. The full pitch from the announcement is not something you can touch yet.
The Real Risk in Claudeforce: Who Owns the Interface
The more interesting story is not the feature list, it is what happens to Salesforce business model. Salesforce has always made money on per-seat licensing. A usage-based AI interaction layer like Claudeforce could cannibalize that seat revenue faster than it gets replaced by new revenue. And once Claude owns the interface layer, pricing leverage tilts toward Anthropic, not Salesforce. A company that already marked its Anthropic stake at a billion valuation earlier this year is now wiring its most valuable asset, CRM data and workflows, straight into someone else model.
If you run sales ops on Salesforce, the practical move right now is to wait: this is pilot-stage, open beta is not until September, and you will want real hands-on results and a security review before committing.
好不好用,試了才知道。
Sources / 資料來源
- Salesforce and Anthropic Announce Claudeforce (Official Press Release)
- CNBC: Salesforce, Anthropic expand partnership as Benioff responds to SaaSpocalypse concerns
- VentureBeat: Salesforce just put its entire CRM inside Claude
延伸閱讀 / Related Articles
- OpenRouter收購評測:Stripe砸75億美元買下AI路由層 | OpenRouter Review: Stripe Buys AI Model Router for $7.5B
- OpenAI Hugging Face入侵評測:700個AI代理聯手駭系統 | OpenAI Hugging Face Hack Review: 700 AI Agents Went Rogue
- AEO評測:Reddit假貼文正在操控你的AI答案 | AEO Review: Fake Reddit Posts Are Manipulating AI Answers
AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends
留言
張貼留言