Yaqi Li · Zoom · Product & Design

Product Manager · Product Designer · Innovation Specialist · Zoom Workplace

Yaqi Li 李亚淇

AI Product Builder · Social Anthropologist · 10+ Years Across 4 Countries

My background sits at the intersection of two disciplines that rarely overlap: social anthropology and AI workflow engineering. The former trains rigorous analysis of how people actually communicate. The latter enables systematic transformation of that communication into structured outputs. Zoom Workplace operates precisely at that intersection.

6archetypes

User need profiles extractedfrom 10+ cross-cultural Focus Groups

30min

AI workflow rebuild of product cycledown from 5-7 days

¥4M+

Revenue from 15 products shippedat Yilv Cultural Travel

AI Workflow Architecture Ethnographic User Research 0→1 Product Delivery Cross-Cultural GTM Enterprise Stakeholder Empathy LLM Orchestration

00 · Areas of Interest

The following reflects my thinking across Zoom's product, design, and innovation domains. The analysis, proposals, and capability evidence below apply across all three directions.

Product Management

Zoom Docs · AI-Powered Collaboration

Defining the meeting-to-document conversion experience. Owning AI features that transform transcripts and meeting context into actionable enterprise assets.

Conversation → Document Pipeline

Product Design

Zoom Meeting · Next-Generation Interaction

Designing the human-AI interaction layer within live meetings. User research, interaction principles, and the behavioral logic of real-time collaboration interfaces.

Human-AI Interaction Design

Innovation & Strategy

Zoom Workplace · Innovation Tools

Defining the product roadmap for next-generation collaborative tooling. AI commercialization, enterprise adoption strategy, and 0→1 feature delivery within a global platform context.

AI Commercialization · B2B Strategy

01 · Strategic Analysis

My reading of Zoom's current position

The following reflects my independent analysis based on publicly available information. These are the strategic contexts I consider most directly relevant to Zoom Workplace's product direction.

INSIGHT 01

Removing "Video" from the legal name was a strategic positioning decision, not a rebrand

In late 2024, Zoom Communications Inc. formally dropped "Video" from its name. The change reflects a deliberate repositioning: from single-purpose meeting software to an AI-first work platform. Zoom Docs and the broader Workplace suite carry the weight of substantiating that narrative with actual product capability.

PLATFORM REPOSITIONING IN EXECUTION

INSIGHT 02

Zoom's primary competitive asset is its position inside the meeting, not its video technology

Zoom processes millions of meetings daily, generating a corpus of high-signal conversational data that competitors cannot replicate without the same meeting infrastructure. Most of this data currently has no downstream lifecycle. The strategic opportunity is converting that conversational corpus into structured, reusable enterprise knowledge — a problem no other platform is better positioned to solve.

STRUCTURAL DATA ADVANTAGE

INSIGHT 03

AI Companion 3.0 reveals the architecture: Zoom is building an orchestration layer, not a model

AI Companion 3.0 routes inference tasks across Zoom's own LLMs, OpenAI, Anthropic, Meta's Llama, and NVIDIA Nemotron based on context. The product moat is therefore depth of workflow integration and enterprise trust, not model performance. The relevant product question is not which model is best — it is at which point in the meeting lifecycle which transformation should be triggered, and under what governance conditions.

ORCHESTRATION OVER CAPABILITY

INSIGHT 04

The field is shifting from AI-augmented tools toward AI-native workflows — Zoom has a narrow window

The emerging competitive pressure is not from tools adding AI features to existing interfaces, but from products built AI-native from the ground up — where context is persistent, the document is the workflow, and human input shifts from execution to review. Zoom's advantage is its position at the conversation source. The question is whether the product organization can move fast enough to define the AI-native collaboration paradigm before it is defined by others.

PARADIGM TIMING RISK

02 · Product Proposals

Four directions worth exploring

These proposals emerged from my own experience managing cross-cultural distributed teams — specifically the gap between what is discussed in meetings and what actually gets executed. They are shared as hypotheses, not prescriptions.

Proposal 01 · Docs + Meeting

Dynamic Task State Machine

Core friction: meeting commitments are spoken, then manually re-entered into separate project tools — producing information loss and execution latency.

When AI detects a spoken commitment in a meeting, it auto-generates an actionable Smart Block in Zoom Docs — with assignee, deadline, and status pre-populated. The document becomes a live project tracker, not a static record. Two-way sync with external tools eliminates the manual handoff entirely.

North Star MetricTask completion rate via AI-generated Docs vs. manually created entries from the same meeting

Proposal 02 · Meeting Design

Real-Time Creative Mirroring

Core friction: verbal description of visual concepts is imprecise. Teams lose time re-aligning on what was meant rather than evaluating what was proposed.

A meeting sidebar canvas that renders AI-generated visual references in real time, triggered by semantic detection of creative intent. When a team discusses a design direction, the sidebar surfaces visual examples immediately. Human input shifts from description to selection and refinement.

North Star MetricTime from creative brief discussion to first visual reference — meeting-based vs. async follow-up baseline

Proposal 03 · Innovation Tools

Invisible Reporting

Core friction: status reporting is largely performative — a summary of information all stakeholders already have, formatted for communication rather than decision-making.

Progress surfaces automatically from actual work artifacts: document edits, meeting updates, task completions. A real-time dashboard updates as work happens, without manual reporting cycles. The interaction model shifts from "write what you did" to "review what was captured."

North Star MetricReduction in time spent on status reporting per contributor per week

Proposal 04 · AI-Native Context Layer

Persistent AI as Organizational Memory

Core friction: enterprise AI tools operate session-by-session. Organizational knowledge — decisions made, rationale discussed, commitments given — is not retained or surfaced when relevant.

Rather than treating AI as a feature within documents or meetings, position Zoom Workplace as the persistent context layer for an organization's work. AI that retains every meeting, every document edit, every commitment — and surfaces the right history at the right moment. The competitive differentiator is not capability, but memory. This is where Zoom's conversation data advantage becomes structurally defensible.

North Star Metric% of decision contexts where AI-surfaced prior meeting history was referenced before action was taken

    ↳ The above reflects external-observer analysis. Internal data, technical constraints, and existing research would significantly shape how any of these directions are prioritized or scoped.
  

03 · Fit Assessment

Capabilities across three role directions

The following maps my background against the core requirements across the product, design, and innovation directions. Each entry is supported by specific evidence.

Capability Dimension	PM · Docs	Designer · Meeting	Expert · Innovation	Evidence
AI product shipping experience	Strong	Medium	Strong	Gemini 6-node workflow (Yilv); WeChat mini-app 4,600+ users; OpenClaw multi-agent architecture
Conversation data processing (transcripts, summaries)	Strong	Medium	Strong	LLM processing of 10+ English Focus Group recordings; multilingual semantic clustering; structured output generation
User research & qualitative insight extraction	Strong	Strong	Strong	LSE ethnographic methodology (Distinction); 300+ deep interviews; 6 user archetypes from 10+ Focus Groups
English communication (written & spoken)	Strong	Strong	Strong	LSE MSc (English, Distinction); OSU BSBA (English); 10+ years cross-cultural professional communication
Enterprise systems & stakeholder coordination	Medium	Medium	Medium	Deloitte PMO: 23 BUs, 40+ stakeholders, large-scale ERP migration. Full-stack development gives technical alignment capability.
Interaction design & visual design execution	Building	Building	Building	Product design judgment from full-cycle PM work; UI design for own applications. No professional design portfolio at SaaS scale.
Enterprise SaaS PM at scale / team management	Building	Building	Building	No direct SaaS PM role at scale or formal team management. Deloitte PMO and independent 0→1 delivery provide adjacent foundations.

* Strong = direct evidence; Medium = transferable with ramp-up; Building = acknowledged gap.

A note on experience level My background does not include a decade of Enterprise SaaS product management, professional design practice at scale, or formal team management. Enterprise collaboration tools frequently fail not due to technical limitations, but because product decisions are made against assumed behavioral models rather than observed ones. My training in ethnographic research provides a systematic method for identifying that gap. If the primary requirement is an operator with an established SaaS or design scale background, I am not the strongest candidate for roles where that depth is the primary requirement. If the requirement includes demonstrated AI product execution, rigorous qualitative research methodology, and direct experience with the conversation-to-structured-output problem — those I can evidence.

04 · Relevant Work

Selected projects with direct relevance to Zoom Workplace's core product challenges.

Yilv Cultural Travel · 2025–Present

AI Workflow: Conversation → Structured Document

Designed and deployed a 6-node Gemini workflow to convert supplier interviews and research conversations into production-ready product SOPs. Architecture handles source selection, hallucination suppression, structured extraction, and formatted output. Product research cycle: 5-7 days → 30 minutes. 15 products shipped, ¥4M+ revenue.

Conversation → Document Hallucination Suppression LLM Workflow Architecture

Trailblazer China · 2024–Present

Multilingual Qualitative Data → Structured GTM Output

Conducted 10+ English-language Focus Groups with European and North American participants. Processed multilingual recordings via LLM pipeline — semantic clustering, insight extraction — producing 6 user need archetypes and 5 primary pain points. Final output: a structured GTM strategy document covering channel strategy, persona-driven messaging, and conversion logic.

Multilingual Processing Semantic Clustering Qualitative → Structured Output

Independent · 2025–Present

"Ming Li You Xi" — 0→1 AI Product, Full Development Lifecycle

Independently managed the complete development lifecycle of a WeChat mini-application: requirements, UI/UX, front-end and back-end development, LLM integration, deployment. Prompt architecture incorporates anthropological frameworks. 4,600+ registered users. Demonstrates end-to-end product ownership and hands-on technical capability in AI product development.

Full-Cycle Development 4,600+ Users Prompt Architecture

2025.09 — PRESENT · HANGZHOU, CHINA

Product Manager — Yilv Culture · Global Grand Tour

AI-Powered 0→1 Product · Full Lifecycle · GTM Strategy

Led 0→1 development of 15 high-end overseas study tour products. Built AI-powered workflow (Gemini, 6 nodes) to compress product research cycles. Responsible for full product lifecycle: market research, design, supply chain, marketing, customer experience, and retrospective. Applied JTBD and Demand Space frameworks to reverse-engineer GTM strategy from user decision paths.

products shipped

80%

group fill rate

2022–2024 · HANGZHOU, CHINA

Project Manager — MSC Consulting · Xiang Xing She (Rural Revitalization NGO)

Immersive Fieldwork · Qualitative Research · MVP Validation

Led cross-regional immersive fieldwork projects across 10+ ethnic minority villages in 4 provinces. Conducted 300+ deep interviews and oral history sessions. Incubated and validated a rural experiential product concept (40+ seed users). Orchestrated a 200-person cross-cultural participatory event.

300+

interviews

40+

seed users

2018 INTERN · 2019–2020 FULL-TIME · CHICAGO, USA

Business Technology Analyst — Deloitte

Technology Consulting · Enterprise Systems · PMO

Led PMO for a large-scale ERP system upgrade at a major energy company — the kind of multi-stakeholder enterprise environment that B2B SaaS products are built to serve. Coordinated 23 business units and 40+ team members without direct authority. Audited and resolved 200K+ data conversion errors. Produced a ground-level understanding of how large organizations evaluate, resist, and ultimately adopt new technology.

BUs coordinated

200K+

errors resolved

2021–2022 · LONDON, UK

MSc Social Anthropology — London School of Economics

Distinction · Highest Classification

Thesis: Taoist ethics and free will through yin-yang balance. Trained in ethnographic methodology and deep-structure cultural analysis. These methods directly underpin my approach to user research and product discovery.

LSE

London

2015–2019 · COLUMBUS, USA

BSBA Finance, Minor Nonprofit Management — Ohio State University

GPA 3.86 · Pace Setter Award (Top 1%, Fisher College of Business)

The Pace Setter Award is OSU Fisher's highest honor, recognizing the top 1% of students across academic achievement, leadership, and community contribution.

OSU

Columbus

05 · What I Bring

Three capabilities directly applicable to this work

The following is not a plan. It is an assessment of specific capabilities I can contribute from day one, based on existing track record.

Capability 01

Ethnographic Decoding of User Behavior

Enterprise collaboration tools frequently fail at adoption due to misalignment between assumed and actual user behavior. My ethnographic research training provides a systematic framework for identifying latent friction points that analytics do not surface — the gap between what users report and what they do. Applied to Zoom Workplace, this means diagnosing the real failure modes in human-AI collaboration flows before they manifest as retention or adoption problems.

Ethnographic Methods Behavioral Gap Analysis Qualitative Signal Extraction

Capability 02

Demonstrated AI Workflow Execution

The core technical challenge of Zoom Docs — transforming unstructured conversational data into structured, actionable documents — is a problem I have already built production solutions against. Using LLM-based workflows of my own design, I have processed multilingual qualitative recordings and produced structured strategic outputs at significantly compressed timelines. The design patterns, failure modes, and architectural constraints I have encountered are directly applicable.

LLM Workflow Architecture Unstructured → Structured AI Product Execution

Capability 03

Enterprise Adoption Intelligence

AI features in enterprise contexts succeed or fail based on whether organizations are willing to trust and operationalize them — driven by governance clarity, data security, and output credibility. At Deloitte, I worked directly inside large-scale organizational change: coordinating 23 business units through a technology migration they did not initiate. That experience produced an operational understanding of how enterprise stakeholders actually evaluate new tools — which directly informs how product decisions should be framed for adoption, not just capability.

Enterprise Change Management AI Trust & Governance Stakeholder Credibility

A question I would prioritize early

At which specific point in the meeting-to-document conversion flow do enterprise users most frequently disengage from AI-generated output — and is that moment currently instrumented?