Sofia Yan

Speaker Booking · 演講邀約 Speaker Booking

邀請一位長期在企業 AI 導入現場工作的人 和你的團隊拆解下一步。 Bring in someone who works inside enterprise AI adoption to help your team unpack the next step.

不從工具清單開始，也不把工程語言原封不動搬過來；
重點是把 AI Agent 的角色、界線和稽核紀錄放回日常工作。 Not a tool-list lecture or an engineering deep dive —
the work is putting AI agent roles, boundaries, and audit trails back into daily operations.

Sofia Yan（嚴世紀），Numbers Protocol 共同創辦人與 CGO。 政大華語教學碩士，長期在 AI 新創、內容溯源與企業導入現場工作。 專長是把企業 AI 導入、內容溯源、AI 治理整理成非工程團隊也能判斷和使用的語言。 Sofia Yan, Co-Founder & CGO of Numbers Protocol. MA in Teaching Chinese as a Second Language (NCCU), with long-running field work across AI startups, content provenance, and enterprise adoption. Her work is translating enterprise AI adoption, content provenance, and AI governance into language non-engineering teams can judge and use.

Numbers Protocol 共同創辦與成長策略 Numbers Protocol co-founding and growth strategy 企業 AI 導入 · 內容溯源 · 工作流治理 Enterprise AI adoption · provenance · workflow governance 中英文演講 / 內訓 / workshop Bilingual talks, training, and workshops SXSW · UNICRI · GNI · AAJA · APJF · Startup@Taipei · 經理人Manager Today
先寄邀約需求Send an enquiry 先看適不適合See if we're a fit

Best fit

什麼場合最適合？When is it the best fit?

我最能發揮的場景，是聽眾已經知道 AI 重要，但還卡在「怎麼讓團隊真的開始用」。如果你要的是工具清單或一小時速成課，可能會有更適合的人。 I'm at my best when the audience already knows AI matters but is stuck on "how do we get the team to actually use it." If you want a tool list or a one-hour crash course, someone else may fit better.

適合A good fit

  • 企業主管 offsite：需要對齊 AI 導入方向與團隊語言。 Leadership offsites: aligning on AI direction and a shared team vocabulary.
  • AI adoption workshop：想從一個真實 workflow 開始改。 AI adoption workshops: starting from one real workflow.
  • 新創 / 產品團隊：想用 AI 加速雛形與內部工具，但需要知道風險邊界。 Startup and product teams: using AI to move prototypes and internal tools faster, while knowing the risk boundaries.
  • 媒體、內容、創作者社群：需要理解內容溯源與 C2PA 為什麼重要。 Media, content, and creator communities: understanding why content provenance and C2PA matter.
  • 非工程背景團隊：想知道自己在 AI 時代可以站在哪裡。 Non-engineering teams: figuring out where they stand in the AI era.

不太適合Not a fit

  • 只想要一串最新 AI 工具教學，沒有組織情境或導入目標。 Just want a list of the latest AI tools, with no organisational context or adoption goal.
  • 單純 KOL 業配或產品背書。 Pure influencer promotion or product endorsement.
  • 期待一次演講直接取代內部流程設計與後續 owner。 Expecting one talk to replace internal process design and an ongoing owner.
  • 需要純工程細節、模型調參或 infra implementation 深潛。 Need pure engineering detail, model tuning, or an infra-implementation deep dive.
描述你的場景Describe your scenario 先看 AI 同事方法論Read the AI coworker methodology

現場照片In the room

不是只存在於簡報裡的 AI 導入 AI adoption beyond the slide deck

我比較常做的事，是把企業、媒體、新創和非工程團隊卡住的 AI 問題，放回真實工作現場裡拆。 Most of my work happens by taking AI problems from enterprises, media teams, startups, and non-engineering teams back into the room where work actually happens.

Sofia Yan speaking beside a Code with Vibe slide about guardrails for startup AI development
Startup@Taipei：Code with Vibe，新創團隊的 AI 開發避險課 Startup@Taipei: Code with Vibe, AI development risk management for startup teams
Sofia Yan speaking into a microphone during a Startup Taipei session
把 AI 開發速度和風險邊界放在同一張桌上講 Putting AI development speed and risk boundaries on the same table
Sofia Yan holding a microphone during an evening community talk
社群現場：用日常語言拆 AI 導入與團隊協作 Community session: unpacking AI adoption and teamwork in plain language
Sofia Yan presenting on a DIGITIMES stage
DIGITIMES 舞台：把技術趨勢翻成團隊聽得懂的判斷 DIGITIMES stage: translating technical trends into team-level decisions

Scope

我可以幫團隊拆的題目 What I can help your team unpack

如果你們已經知道「要用 AI」，但還卡在要從哪個流程開始、怎麼評估工具、誰來負責後續，我通常會從這幾種題目切進去。 If your team already knows AI matters but is stuck on where to start, how to evaluate tools, or who owns the follow-through, these are the entry points I usually work from.

AI 需求盤點與優先順序 AI needs mapping and prioritization

把客服、營運、內容生產、知識管理裡真的值得導入 AI 的場景先找出來，不從工具清單開始。 Find the customer-service, operations, content, and knowledge workflows that are actually worth applying AI to before picking tools.

AI Agent 工作流程設計 AI Agent workflow design

定義 agent 的角色、邊界、交接、稽核紀錄和人類 review 點，讓它真的進到日常流程。 Define agent roles, boundaries, handoffs, audit trails, and human review points so AI can enter daily work safely.

AI 工具與廠商評估 AI tool and vendor evaluation

協助團隊看懂不同解決方案的能力、限制、資料風險和導入成本，避免買完才發現接不起來。 Help teams compare capability, constraints, data risk, and implementation cost before committing to a solution.

AI 開發避險與技術檢核 AI development risk review

幫新創與產品團隊拆哪些原型可以放心用 AI 快速試錯，哪些牽涉金流、資安、資料品質或長期維護時一定要工程檢核。 Help startup and product teams separate safe AI-assisted experiments from work that needs engineering review because of payments, security, data quality, or long-term maintenance.

AI 治理與 Vibe Coding 實作 AI governance and vibe-coding practice

用 TAEA、Vibe Coding 和原型實作，把非工程團隊訓練成看得懂風險、也做得出東西的人。 Use TAEA, vibe coding, and prototype practice to help non-engineering teams understand risk and build working artifacts.

主打主題Signature topics

文組人也能領導的 AI 導入 AI adoption a humanities person can lead

為什麼很多企業 AI pilot 沒有進到日常流程？常見問題不是模型不夠強，而是 AI 沒有被安排角色、界線和 review 方法。這個主題會從非工程背景的 AI 新創現場經驗切入。 Why do many enterprise AI pilots fail to become daily work? The problem is often not model strength, but the lack of role design, boundaries, and review practice. This topic starts from field notes inside an AI startup, from a non-engineering operator's view.

AI 同事方法論：Numbers 案例拆解 The AI coworker methodology: a Numbers teardown

用 Numbers 的內部案例，拆解一個團隊如何把 AI 同事分成清楚角色：Amy、Jordan、Clara、小鳳、阿張、老科，各自有工作範圍、權限邊界和人類 review 點。 Using Numbers' internal practice, this topic breaks down how a team assigns AI coworkers into clear roles: Amy, Jordan, Clara, Xiao-Feng, A-Zhang, and Lao-Ke, each with scope, boundaries, and human review points.

Code with Vibe：新創團隊的 AI 開發避險課 Code with Vibe: AI development risk management for startup teams

AI 能讓非工程團隊快速做出產品雛形和內部工具，但金流、資安、資料品質和長期維護不能靠「看起來可以跑」就過關。這個主題會拆什麼情境可以用 AI 快速試錯，什麼時候必須請工程師把關。 AI can help non-engineering teams build prototypes and internal tools quickly, but payments, security, data quality, and long-term maintenance cannot pass just because a demo runs. This topic separates safe rapid experiments from work that needs engineering review.

內容溯源與 AI 時代的信任戰役 Content provenance and the trust war of the AI era

當生成式 AI 幾秒就能造一張逼真影像，新聞、創作、IP 怎麼保護？C2PA、ERC-7053、TAEA 框架。Numbers 與 UNICRI / Google News Initiative / 伙影的實戰案例。 When generative AI can fake a photorealistic image in seconds, how do we protect news, creative work, and IP? C2PA, ERC-7053, the TAEA framework — with real cases from Numbers' work with UNICRI, the Google News Initiative, and 伙影.

過往演講與課程Past talks & courses

2024-2026 主題課程 / Keynote 2024–2026 courses / keynotes

國際 panel / 媒體International panels / media

  • SXSW 2023 Pitch（Metaverse & Web3 冠軍，Numbers 團隊） SXSW 2023 Pitch (Metaverse & Web3 winner, Numbers team)
  • Asia Pacific Journalism Forum 2025 講者(speaker)
  • AAJA Webinar 「Surviving the AI Flood: Trust & Revenue」（主持，2025/06）"Surviving the AI Flood: Trust & Revenue" (host, 2025/06)
  • PL Summit 23 | LabWeek23 (Istanbul)
  • 倫敦 2023：AI 於創意產業之變革性技術 London 2023: Transformative tech for the creative industries
  • 日內瓦 2024：如何在 AI 驅動的世界對抗假訊息 Geneva 2024: Fighting disinformation in an AI-driven world
  • UNICRI （聯合國跨區域犯罪暨司法研究所）合作(UN Interregional Crime and Justice Research Institute) collaboration
  • Google News Initiative 補助獲獎（2025）grant recipient (2025)
  • NTU TEC Talk （台大創創中心 S2EP1）(NTU entrepreneurship center, S2EP1)
  • 伙影通訊社（與余志偉攝影記者） 伙影 news agency (with photojournalist Yu Zhi-Wei)

可以拆解的內部案例 Internal cases we can unpack

這些數字保留在 proof context 裡使用，演講時可以回到背後的流程、限制與適用條件，而不只看結果。 These numbers stay inside a proof context. In a talk, we can trace the workflows, constraints, and conditions behind them, not only the result.

Sofia 在 Numbers Protocol 帶領的成果 Results Sofia led at Numbers Protocol

  • Numbers 平台社群 1,000 → 130,000 人（+12,900%） Numbers community 1,000 → 130,000 (+12,900%)
  • Numbers 內容驗證平台覆蓋 190 個國家 · 月流量 60 億 Numbers content-verification platform: 190 countries · 6B monthly views
  • 國際媒體報導 Numbers / Sofia 工作：Politico、Meet Global、Tech Company News、ANI International coverage of Numbers / Sofia: Politico, Meet Global, Tech Company News, ANI
  • Vibe Coding 訓練：15 分鐘打造並上線行銷頁面 Vibe-coding training: build and publish a landing page in 15 minutes

2025 年 4 月全面 AI 導入後（Numbers 內部） After full AI adoption in April 2025 (Numbers internal)

  • 內部 MVP 從需求到完成 ≤ 24 小時（原本需要數週） Internal MVP from request to done in ≤ 24 hours (previously weeks)
  • 年度 SaaS / Infra 支出 -60% Annual SaaS / infra spend −60%
  • 全職人類同事剩 4 位 · 線上 AI 同事 15 位 + 實習 8 位 4 full-time human coworkers · 15 online AI coworkers + 8 interns
  • 每位人類同事平均同時跟 3 位 AI 同事協作 Each human works with ~3 AI coworkers at once
  • 團隊已落地超過 100 個微自動化流程 Team adoption has shipped 100+ micro-automation workflows

Numbers Protocol 是 C2PA 會員、被 NIST AI Risk Management Framework 引用、共同設計 ERC-7053 和 ERC-7517 標準、與 Adobe / Microsoft / Nikon / Reuters 合作、67M+ 數位資產註冊。 Numbers Protocol is a C2PA member, cited by the NIST AI Risk Management Framework, co-designed the ERC-7053 and ERC-7517 standards, works with Adobe / Microsoft / Nikon / Reuters, and has 67M+ digital assets registered.

FAQ

Sofia 通常用什麼語言演講？ What language does Sofia present in? +

中英文皆可。台灣企業內訓多以中文進行；國際場合（國外總部、跨國團隊）用英文。雙語場合也常見，視聽眾組成決定。 Both Chinese and English. Corporate training in Taiwan is usually in Chinese; international settings (overseas HQs, cross-border teams) in English. Bilingual sessions are common too, depending on the audience.

演講主題有哪些？可以客製嗎？ What topics are available? Can they be customised? +

四大主軸：（1）文組人也能領導的 AI 導入，（2）AI 同事方法論（Numbers 案例拆解），（3）Code with Vibe：新創團隊的 AI 開發避險課，（4）內容溯源與 AI 時代的信任戰役。可依貴公司產業、團隊規模、AI 成熟度客製。 Four pillars: (1) AI adoption a humanities person can lead, (2) the AI coworker methodology (a Numbers teardown), (3) Code with Vibe: AI development risk management for startup teams, (4) content provenance and the trust war of the AI era. Each can be tailored to your industry, team size, and AI maturity.

Sofia 接受怎樣的演講邀約？ What kinds of bookings does Sofia take? +

企業內訓 (in-house corporate training)、產業論壇 keynote、公開研討會、podcast / 媒體訪談。不接 KOL 業配、不開公開線上課。 In-house corporate training, industry-forum keynotes, public seminars, and podcast / media interviews. No influencer promotions and no public online courses.

海外活動可以嗎？ Are overseas events possible? +

可以。Sofia 過去曾在 SXSW (Austin)、AAJA-Asia、UNICRI 等國際場合演講，能處理跨國時差與旅程。 Yes. Sofia has spoken at SXSW (Austin), AAJA-Asia, UNICRI and other international venues, and is comfortable with cross-border travel and time zones.

為什麼是 Sofia？ Why Sofia? +

Sofia 的組合比較少見：政大華語教學碩士背景，加上多年 AI 新創共同創辦與成長實務。她擅長把 C2PA、agentic workflow、AI 治理這類工程與治理概念，整理成非工程團隊可以討論、判斷和採取行動的語言。 She's one of the few speakers who pairs an MA in Teaching Chinese as a Second Language (NCCU) with 7+ years of hands-on AI-startup co-founding. She explains C2PA / agentic workflows / AI governance in language non-technical people understand, instead of just copying engineer concepts.

邀約方式How to book

目前請以下列任一方式聯絡，48 小時內回覆： Reach out through any of the following — reply within 48 hours:

報價會依形式、地點、規模與客製深度而定。歡迎在 email 裡先描述場景，我們可以快速判斷適合的規格。 Pricing depends on format, location, scale, and customisation. Describe your scenario in the email and we can quickly suggest the right format.