Anna Joung

I make things scale — products, teams, and now my own tools. Eight years shipping in Seoul and Tokyo. Currently in Berkeley.

I have spent my career taking products from small to large. At LINE I launched OpenChat in Japan and ran the experiments that scaled it past 3M monthly users. At Kakao Pay Securities I built the systems that brought dormant users back and moved $165M in trading volume. I ran my own consultancy, Valuelab, and took a YC-backed startup into the Japanese market from zero. Most recently, as an AI product manager at Channel Corp, I launched the company's first small-business freemium tier and converted its first paying account.

Scale is also why I write. I co-authored a book on U.S. stock investing for Korean readers in their twenties — 60,000 copies, twenty printings over six years. Making a hard subject plain enough that a beginner acts on it is the same problem as product: you are not simplifying the thing, you are removing what stands between the person and the thing.

I am now a first-year MBA at Berkeley Haas. These days I build with AI every week — the work is below.

Portfolio

Products I have owned, and what changed because of it.

AI Product Manager · Channel Corp
Feb 2026 – Jul 2026
Korea's leading AI customer service platform · 250K+ business customers · Series D

Launched the company's first small-business freemium tier. Thirty customer interviews surfaced the mismatch: we were selling customer-service efficiency to owners who wanted revenue. Repackaged three AI products around that reframe, established product-market fit, and converted the first free account at $110 per month.

Defined the small-business scope for the company's AI analytics agent. Twenty-five founder interviews and thirty live sales calls, validated with a Claude Code prototype before specs. Shipped two products and elevated a cash-flow report from P2 to P0 against the internal roadmap.

Drove AI sales agent answer accuracy from 36% to 90%. Designed the conversation state flow and hallucination guardrails across five QA rounds, piloting the agent to replace a manual lead-qualification process that converted at 7%. Selected the pilot customer from 19 candidates on purchase-intent data.

Forward-deployed a voice AI agent across three use cases, including satisfaction calls for Korea's largest childcare commerce platform, doubling survey completion. Killed the debt-collection track for regulatory risk.

Founder & CEO · Valuelab
Mar 2024 – Jan 2026
Product and go-to-market consultancy for early-stage startups

Rebuilt the user journey and product architecture for InAnswer, an AI education startup — sign-up from three steps to one, SKU portfolio from five to three, driving 10x month-over-month sign-up growth.

Ran Brandazine's (YC W22) Japan market entry from zero. Designed the localized go-to-market motion and the testing framework behind it, doubling lead volume at 80% lower acquisition cost.

Growth Product Manager · Kakao Pay Securities
Sep 2022 – Jan 2024
Korea's leading fintech firm · $2.8B+ assets under management

Built a two-tier dormant-user reactivation system. Behavioral analysis surfaced the milestones that predicted retention; aligning eight cross-functional teams reactivated 50% of 10,000 dormant users on day one, at 80% retention.

Designed a tiered stock-transfer incentive program that generated $19M in inflows and $165M in trading volume, enabling management's acquisition talks with U.S. brokerage Siebert Financial.

Growth Manager · LINE PLUS / LINE Corp.
Apr 2018 – Aug 2022
Global messaging platform · 220M+ monthly active users · Seoul and Tokyo

Pioneered the Japan launch of OpenChat, LINE's interest-based public group chat product. Ran 100+ experiments to scale past 3M monthly active users and rank among LINE's global top three products by growth.

Rebuilt the sign-up flow after six months of user data showed 10% of sign-ups came from in-app referrals — eliminating landing pages lifted conversion 26% and cut three ops hours per week.

Building

One thing a week, built with AI. I am deliberately mixing B2B and B2C, and rotating through agent design, evals, and the generative models I have not used yet. The goal is to learn the stack, not to grow users.

Agents that review my work before anyone else does
Ongoing
Agent harnessInternal tool

A set of agent systems that encode a reviewing standard and run against my own drafts. One simulates my CEO's and my manager's objections before I hand a document over. Another runs unattended on a schedule and groups my recurring mistakes by pattern. I would rather build the internal tool than ask for it.

Prototypes instead of specs
2026
B2BClaude Code

At Channel Corp I prototyped in Claude Code and pre-sold the prototype across 30 live sales calls before writing a single specification. Demand was validated by the customer, not by the roadmap.