Investor and partner brief · job4autism.com · September 2026
AI4Autism operates rentable AI factories. An autistic founder brings an ability and an order; the factory brings the agents, the tools, the executive functions and the human reviewers, and returns a product that can be sold. We take a share of what it earns.
Pre-revenue. Phase 1 is 10 pilot factories. Every figure on this page is a model assumption, not an operating result.
Across ASEAN, autistic adults leave school into a labour market with no pathway built for them. Employment support, where it exists, aims at placing a person into a job description written for someone else, and the jobs available rarely match how the person actually works best.
Meanwhile the tooling that would let one person run a production business now exists, but it is fragmented across dozens of separate AI products, each demanding its own account, prompt craft, payment method and integration work. That fragmentation is a barrier for anyone; for many autistic users it is disqualifying.
The gap is therefore not talent and not technology. It is the absence of assembled infrastructure: a place where ability turns into output without the person first having to become an operations manager.
Country-level employment and prevalence figures are being compiled from cited public sources for the data room rather than estimated here. We would rather show one verified number than five plausible ones.
Six product layers, one operating system. The park is the commercial surface; FactoryOS is the defensible part.
The contribution ledger is a switching cost for tool providers and a trust asset for buyers. Neither a general AI platform nor a local jobs programme can produce it without rebuilding both halves.
Autism intervention has explicit, decomposable methodology — task analysis, prompting hierarchies, visual supports — which maps unusually well onto agent workflows. The domain is a competitive advantage, not a charitable framing.
A large young population, low digital production costs, an underserved special-education market, and an existing operator on the ground with Vietnamese-language material already built.
Three layers, so that a tenant who cannot pay up front is still a viable customer and a tenant who succeeds becomes a materially larger one.
| Layer | What is charged | Illustrative level |
|---|---|---|
| Subscription | Factory, infrastructure, C-suite, hybrid team, workspace, storage | $19–499 / month |
| Usage | Tokens, API calls, video generation, GPU, workflow runs | Cost-plus |
| Revenue share | Products, services, marketplace sales, licensing, royalties | 5–15% of gross |
| Enterprise & franchise | Dedicated factories for schools and organisations; country operators | Contracted |
Move the inputs to see how the three layers combine. This is an arithmetic model of the architecture for discussion, not a forecast, and it assumes a founder keeps 60% of gross product revenue after the park, tool and infrastructure shares and production cost.
The model deliberately shows the founder line next to the park line. If the founder number is not meaningful at a given scale, the park number is not a business we want to build.
Ten factories across autism education, K12, video modeling, music, film, design, software, research, marketing and e-commerce. Each with 10–20 agents, one AI C-suite, 3–5 human experts and 10–20 pilot users. The test is narrow and binary: does Factory → product → customer → revenue close at all, and what does one product actually cost to make?
Factory Compiler turns a written description into a runnable factory configuration, so new factories stop requiring our team. Most are logical factories on shared infrastructure, so marginal cost falls as the tool marketplace fills in.
Country operators in ten ASEAN markets run local factories, language, curriculum, marketplace, payments and expert networks on the shared OS. Franchise and licensing become a third revenue line alongside subscription and revenue share.
The economic value an autistic person creates with AI support, without depending on traditional hiring. It is reported per founder, per factory and per country, and it is the number we expect to be held to.
Supporting measures: founders earning, hours of suitable work, products sold, customers served, countries active, and the share of factories still active after six months.
| Human in the loop | No product reaches a customer without a named human reviewer, recorded in the ledger. |
|---|---|
| IP ownership | The founder owns the product. The park holds a licence to list and sell it. |
| Minors' data | Learner material is produced without identifying data by default; centres and schools contract separately for anything else. |
| Attribution | Tool providers are paid on measured contribution, auditable per product. |
| Risk | How we are addressing it |
|---|---|
| Demand risk: buyers may not pay for AI-assisted output | Phase 1 sells into segments with an existing budget line — schools, intervention centres, publishers — rather than into a new category. |
| Quality risk: a weak product damages trust in autistic producers | Mandatory human review, published Factory Passports, and QA gates that block listing rather than flagging after the fact. |
| Cost risk: model inference cost per product | Usage is a pass-through layer, and the tool marketplace lets us route work to whichever provider is efficient for each step. |
| Concentration risk: dependence on a single model provider | FactoryOS is built against MCP, API and SDK interfaces so tools are replaceable at the registry level. |
| Execution risk: scope is wide | Phase 1 is deliberately 10 factories, not 1,000, and each one has to reach a paying customer before the next cohort opens. |
We are opening Phase 1 to partners rather than running it alone. Terms, the financial model and the Phase 1 budget are in the data room on request.
List tools in the registry and take royalties on measured contribution. For a model or infrastructure provider, this is distribution into a market that cannot otherwise buy per-seat licences, with usage attributed product by product.
Discuss a tool listingFund the Phase 1 cohort of 10 factories and the ten pilot founders behind them. Returns are reported on both lines: park revenue and founder income, with impact measured by the same ledger that runs payouts.
Request the data roomSchools, intervention centres, universities and ASEAN autism organisations that will place real orders and supply the human reviewers. This is the fastest way to test whether the output holds up in practice.
Propose a pilotUse of Phase 1 funds, indicatively: factory build and agent engineering, human expert panel and QA, founder stipends during the pilot, compute, and localisation. The allocation is set with the lead partner rather than published here.
| Entity | Job4Autism AI Native Co., Ltd |
|---|---|
| Tax ID | 0319635269 |
| Headquarters | Thu Duc, Ho Chi Minh City, Vietnam |
| Properties | job4autism.com (the park), job4autism.vn (Vietnam), ai4autism.vn |
| Stage | Pre-revenue; Phase 1 pilot in preparation |
| Contact | leduonglam58@gmail.com |
A working library of Vietnamese-language production tools and learning material built with the same agent-and-review method the factories use, plus an operating architecture for the storefront, payment and QR provenance layer. Phase 1 assembles these into ten factories with customers attached.