ESTsoft · Product Owner
Alan LLM service
- LLM Agent
- AI Search
- Monetization
Problem
An LLM service in trial operation at a research org (AI Agent Lab) had to become a real commercial service and earn revenue. With users moving from search sites to LLM services, it also had to feel natural to people used to search.
Key decisions
- 01Moved the trial-stage service onto separate Dev, QA, Stage and Release servers to bring it up to commercial-service standards
- 02Offered the LLM agent service in a search style to lower the barrier for existing search users
- 03Slides in stages, not one shot — an agent orchestration split into intent, scope, research, template, rendering and revision, with a step where users check and edit intermediate results (also built image/YouTube search and report generation agents in-house)
- 04Priced on evidence — modeled margin from inference cost per call, then set the Pro plan's price and usage limits from competitor pricing and expected paid conversion
Outcome
Led a service in trial operation at the research org through commercial launch and monetization. As PO I led the whole monetization — pricing and plan design, payment and cancellation flows, terms and notices, and the launch announcement — and launched the Pro subscription in July 2025 (announcement), bringing in the service's first paid revenue (figures not disclosed). On that basis I planned an enterprise product, which later led to institutional contract revenue. I also shipped deep research and slide generation and built the back-office operations tool. (myalan.ai)
Details
- Project name
- Alan (LLM based Agentic AI Search Engine Service)
- Client · Domain
- ESTsoft / LLM-based AI service
- Period
- 2025.04 - 2026.01
- Scope of work
- LeadAs PO, led a service in trial operation at the research org through commercial launch and monetization
- LeadRan the AI researcher and app developer (FE, BE) team — feature planning, schedules and risk management
- LeadArchitected the Azure, FastAPI, React and LangGraph stack; ran separate Dev, QA, Stage and Release servers
- LeadBuilt image search, YouTube search & summary and report generation features, plus specialized agents, in-house
- Hands-onPlanned the slide generation feature — benchmarked Genspark and other services, defined user scenarios and the generation flow
- Hands-onDesigned the slide generation agent — staged orchestration (intent → scope and outline → research → template, style and fonts → HTML rendering → revision), with human-in-the-loop review of intermediate results
- Hands-onPlanned the monetization of an AI service — modeled margin from inference cost per call and set the Pro plan's price and usage limits from competitor pricing and expected paid conversion
- LeadPaid launch — plan entitlements, payment, cancellation and refund flows, revised terms and paid-conversion notices (2025.07, announcement)
- CollabRan quality testing with in-house QA, auth, payments and infra teams; prepared the launch announcement with marketing and business
- LeadShipped the deep research and slide generation services; patch notes and back-office operations tool
- In commercial operation (myalan.ai)
- Role
Product Owner
(+ Project Manager, + Development Leader)
- Tech stack
- GPT
- Gemini
- MCP
- Notion
- Google Analytics
- Git
- Azure
- Figma
- MS Docs
- Slashpage