JRS Education Agent
AI-native education technology, from project decisions to teaching content
A two-founder company running a ten-person workload through agentic automation
JINRUISHENG CO., LIMITED · Hong Kong
- USD 7–8billionglobal AI-in-education market
- 660K+K-12 students in Hong Kong
- 20–30institutional clients by Year 3
- Betawith paying POC customers
A Hong Kong OPC building agentic AI for education — with paying customers from day one
A Hong Kong-incorporated, AI-native education technology company, built around two agentic products and a lean OPC operating model — entering a USD 7–8 billion global market.
Incorporated in Hong Kong
18 September 2025 under the Companies Ordinance (Cap. 622); independent, with no parent company or group.
AI-native education technology
Two agentic products cover the chain from project decisions to teaching content.
Lean OPC operating model
Two founders orchestrate agent workflows; team capped at 10, scaling through automation, not headcount.
Paying customers from day one
Both products run at Beta with paying POC customers — the model is revenue-validated before scale-up.
Making knowledge performable, and decisions computable.
One-stop agentic support for education institutions — from project decision to content output.
Team credentials
Both founders are certified Damo Academy AI Trainers (Advanced) and alumni of the Oxford Saïd × UNESCO course "AI, Justice and the Rule of Law".
A USD 7–8 billion global AI-in-education market, compounding at over 20% a year
Global AI-in-education market size, USD billion
Forecast horizons: Grand View Research to 2033 (CAGR 25.9%); Mordor Intelligence to 2030 (CAGR 42.8%).
in Hong Kong
Hong Kong's school population has grown for two consecutive years (Education Bureau, 2025/26) — a stable institutional base.
What this means for us
Growth comes from institutional adoption, not consumer hype. Our entry points — project advisory and content production — sit on the two budgets schools already control.
Sources: Grand View Research (2025); Mordor Intelligence (2025); HK Education Bureau (2025/26)
Institutions pay twice for the same project — once for decisions, once for content
Two agentic products on one self-developed toolchain: an advisory agent for project decisions, and an edu-drama agent for content production.
Content production is slow and costly
A five-minute teaching animation takes a studio days of scripting, illustration and dubbing — rarely tailored to one knowledge point.
Project decisions rely on fragmented consulting
Compliance, market scans, cost estimates and financial models sit with different vendors; figures drift and contradict each other.
Generic AI content fails the classroom
Off-the-shelf AIGC drifts off-curriculum and breaks consistency across episodes — unusable for a course series.
One agentic toolchain, two products
Education Project Advisory Agent
Six specialist agents deliver source-traceable proposals in Word, Excel and PPT — estimates flow into financial models automatically.
Edu-Drama Agent
One knowledge point in, a story-driven teaching mini-drama out — with cross-module consistency checks.
Advisory Agent: six specialists in parallel, one traceable deliverable
Why it matters
Estimates flow straight into the financial model and are cross-validated — one auditable pack instead of six disconnected reports.
Where humans stay in the loop
Agents draft, cross-check and self-correct; our team performs only the final review — accountability kept, delivery time compressed.
Beta stage, with paying POC customers in delivery.
Edu-Drama Agent: one knowledge point in, a teaching mini-drama out
From studio days to agent minutes
The pipeline batch-produces episodes across subjects — geometry, classical Chinese, history — with the same characters and visual style held constant, so a course series stays coherent.
Prototype advancing to commercial rollout, with paying POC usage at Beta stage.
Knowledge point
one line of input
Scriptwriting
plot & dialogue
Storyboarding
shot breakdown
Character design
consistent persona
Voice & subtitles
dubbing, aligned
Rendering & review
synthesis + QA gate
Cross-module consistency checkcharacter design and visual style verified across every module before rendering
Agent × Skill × Tool: models are a commodity, orchestration is ours
Self-developed where it counts
Dispatch engine, orchestration, consistency validation and task verification are built in-house; each layer is independently replaceable.
LLM-agnostic
Foundation models are swappable generation capacity via API — no lock-in; costs track market prices down.
Why this is innovative
Shifts education AI from "features a human triggers" to agents that run the workflow end-to-end.
Three revenue lines, priced for institutional budgets
Three revenue lines priced for institutional budgets, a conservative bottom-up growth path, and a founding team pairing education operations with large-scale system architecture.
| Revenue line | What the client gets | Indicative pricing | Status |
|---|---|---|---|
| Education project advisory | Compliance-checked, financially modelled project proposals (Word/Excel/PPT) | HK$200K–300K per project, p.a. | In service — Beta, paying POC |
| Custom edu-drama production | Story-driven teaching mini-dramas built from single knowledge points | HK$150K–200K per project | In service — commercial rollout |
| Toolchain licensing | Self-serve access to the JRS one-stop AI creation toolchain | HK$50K–150K per institution, p.a. | Planned — pilot from Year 2 |
How to read this
Two lines are already in service at Beta stage with paying POC customers, so pricing reflects live delivery rather than theory. The licensing line is the recurring-revenue layer — deliberately marked as planned and piloted only from Year 2, once services prove productisation.
Why margins hold
Agentic production keeps marginal cost low and the team capped at ten — pricing power comes from speed and consistency, not headcount.
A bottom-up path to 20–30 institutional clients by Year 3
Beta → commercial
Complete productisation of both agents; convert Beta POC usage into reference clients across Hong Kong institutions.
Licensing pilot
Extend to multi-subject content and a wider education client base; pilot toolchain licensing as recurring revenue.
Bottom-up estimate
Education content production platform serving 20–30 institutional clients — approx. HK$5–10M annual revenue.
Traction today
Both products run at Beta with paying POC customers, and delivery capacity is already proven in live engagements. The Year-3 target is a bottom-up estimate from the indicative pricing on the previous page — a plan with stated assumptions, not a promise.
Founding team: education operations × large-scale system architecture
KUANG Haohua
- Research grounding
- BEng + MEng, Central South University; research at Curtin University; three co-authored journal papers (2011–2019).
- Education operator
- Shareholder of a licensed kindergarten (2018–2020); since 2020 runs a training business (~150 students, ~RMB 1.6M p.a.).
- AI credentials
- Damo Academy AI Trainer (Advanced); Oxford Saïd × UNESCO course alumnus.
KUANG Xiaocheng
- Computer science pedigree
- BSc, USTC (Anhui Provincial Outstanding Graduate); PhD, Institute of Acoustics, Chinese Academy of Sciences.
- Patents and papers
- Five granted patents (OSS, IPTV, billing, network management); four papers, one EI-indexed at IEEE ICCS 2008.
- Large-scale architecture
- Senior engineer, Perfect World; Senior Technical Architect, ByteDance (c. 2019–2025), with system-architecture patents.
- AI credentials
- Damo Academy AI Trainer (Advanced); Oxford Saïd × UNESCO course alumnus.
Every knowledge point deserves its own story.Every education project deserves a computable decision.
JINRUISHENG CO., LIMITED
KUANG Haohua, Founder & CEO · jinruisheng@yeah.net · +852 5761 4692 · jinruishengltd.com