Natural-Language Automation Runtime

From a sentence to the shop floor.

AI GO turns a plain-language instruction into a running, auditable automation loop — deployed to an on-site edge core in minutes, not weeks. No PLC programmers. No six-figure integration projects. Built for the small factories that traditional automation left behind.

STATUS: Pilot · validated on real hardware
TEAM: ALL GOOD — All for Good
ORIGIN: National Tsing Hua University · Taiwan
The Gap

Automation exists. Small factories can't reach it.

Taiwan alone has over 140,000 small and mid-size manufacturers. Most still log machine temperatures by hand and discover breakdowns the next morning — not because they don't want automation, but because every deployment today is a bespoke engineering project.

142K
SME manufacturers in Taiwan — the core market AI GO serves
~70%
Have not completed smart-manufacturing transformation (32% adoption vs ~47% globally)
Weeks+
Typical lead time for one custom integration, at six-figure NT$ cost

Every deployment is bespoke

PLC logic, protocol wiring and control code are written by hand for each site — a costly, non-repeatable project every single time.

Only two doors, both closed

Hire a PLC engineer or commission a system integrator. Both cost hundreds of thousands of NT$ and require ongoing maintenance a 40-person factory can't sustain.

Problems surface too late

Readings logged hourly by hand get missed. Downtime is discovered the next day — or at month-end, when the losses are already booked.

The bottleneck isn't AI capability. It's the missing execution interface between language and machines — one a factory can operate without a single engineer.

How AI GO Works

Understand the sentence. Execute the loop.

A complete automation is a loop — sense, understand, decide, act, report. AI GO runs the whole loop from one plain-language instruction, with a hard boundary between the AI that understands and the engine that acts.

Language & IntentUNDERSTAND

An LLM parses the operator's instruction — trigger, condition, target, action — into a structured intent spec. It never issues device commands directly.

↓ intent spec · confirmed by the operator
AI GO Runtime
Deterministic ExecutionDECIDE · ACT

A deterministic engine validates the spec field by field, then calls only pre-tested, permissioned Skills. Every action leaves a full audit trail.

plc.stop notify.push log.event report.daily query.metrics
↓ actions · ↑ telemetry
Devices & SignalsSENSE · EXECUTE

PLCs, sensors, conveyors, cameras and IoT devices — reached through industrial standards (Modbus, OPC-UA, MQTT), mostly via signals the factory already has wired to its PLC.

One interface covers the whole plant

The same natural-language surface handles monitoring, alerts, conditional control, scheduled reports and ad-hoc queries. "Log every machine's temperature every five minutes." "Notify me if any line stops for more than five minutes." "Which machine had the most downtime this week?" — one sentence each, no dashboard training, no code.

Devices are packaged as plug-and-play Skills behind a standard interface. Adding hardware means adding a Skill — not re-integrating the system. An open Skill Marketplace lets third parties ship modules for PLCs, robots and cameras, compounding into a network effect no single integrator can match.

And because inference and control run on the Edge AI Core inside the plant, automation keeps running when the network doesn't — with sensitive production data never leaving the site.

Safety by Architecture

The LLM never touches the machine.

"What if the AI misunderstands?" is the first question every factory owner asks. Our answer is structural, not procedural — a mistaken interpretation stops at the spec layer, where a human catches it, before anything physical can move.

Intent spec separation

Language becomes a structured, human-readable spec before anything executes. The model proposes; it never commands.

Human in the loop

The app reads the plan back in plain language. Nothing runs until the operator confirms; critical actions require explicit approval every time.

Deterministic execution

Only pre-tested Skills fire, gated by role permissions, action whitelists and dry-run simulation. Behavior is reproducible and provable.

Audited & air-gappable

Every action is logged for traceability. Cloud and edge are separated by design, keeping OT data inside the plant per industrial security practice.

The Product

Three parts. One deployment.

Everything a factory needs to go from spoken intent to running automation — sold as an affordable monthly subscription plus one piece of on-site hardware.

Mobile App

Speak, confirm, monitor

Natural-language input, plain-language confirmation, live dashboards, push alerts, and conversational queries — "what was line 2's yield last month?" answered with charts.

Cloud Platform

Manage every plant

Accounts, device sync, workflow management, cross-site analytics and remote oversight — the SaaS backbone that scales from one machine to a fleet.

Edge AI Core

Run inside the plant

On-premise inference, task planning and device control on commercial edge hardware. Keeps executing through network outages; keeps production data on-site.

Speaks the plant's native protocols

PLC SystemsModbus TCP/RTUOPC-UAMQTTTemperature SensorsConveyorsSmart CamerasAGVsIoT Devices PLC SystemsModbus TCP/RTUOPC-UAMQTTTemperature SensorsConveyorsSmart CamerasAGVsIoT Devices
Market & Model

A blue ocean the integrators skip.

Global smart manufacturing is heading toward US$620B by 2026, yet the economics of traditional integration only work for large plants. We don't fight incumbents for factory-wide contracts — we serve the tier they structurally cannot reach.

TAM Taiwan's industrial SMEsThe full population of small and mid-size industrial firms ~327KCOMPANIES
SAM Semi-automated or manual small factoriesPriced out of traditional SI — managed today with Excel and LINE 30–50KNT$1.5–4B / YR
SOM 3-year target via subsidy channels & pilot sitesSingle-digit share of SAM — ample headroom for an early-stage company 500–1,000NT$25–80M / YR
Engine 01 · Recurring

Subscription

Tiered monthly and annual plans — the cash-flow backbone that grows naturally with each customer's devices and features.

Engine 02 · Hardware

Edge AI Core

One-time hardware revenue that anchors the deployment and drives follow-on sensors, upgrades and Skill bundles.

Engine 03 · Platform

Skill Marketplace

Revenue share on third-party device modules — near-zero marginal cost, network effects, and the long-term moat.

Where We Are

Pilot stage — proven on real hardware.

The core runtime has been built and tested against real embedded devices, sensors and control workflows, validating the architecture across multiple hardware scenarios.

Idea
Prototype
Pilot
Growth
Scale
  • Working runtime on real devicesLocal-first runtime connecting AI agents to embedded hardware, sensors and control systems.
  • Natural language → device action, validatedEnd-to-end workflows where instructions become monitored, executed and logged actions.
  • Production-proven stackThe founding team has already shipped an LLM-driven mobile product for a traditional industry — the same stack AI GO runs on.
  • De-risked go-to-marketLaunching with read-only monitoring and alerts at pilot factories, riding Taiwan's NT$46B+ industrial digitalization subsidies as an acquisition channel.
Recognition & Roadmap

Hult Prize NTHU 2026 — Campus Champion

First place among campus teams, validating both the problem and the team behind it.

0–3 moMVP: monitoring / alerts / reports + core Skills
3–6 moPilot POCs at 1–2 small factories
6–18 moSubsidy channel · control actions · replication
18 mo +Skill Marketplace & developer ecosystem
The Vision

Every factory should use automation like it uses a chat app.

AI GO's long-term position is the AI Control Layer between language and physical equipment — starting with Taiwan's small manufacturers, then Southeast Asia's, wherever labor shortages meet an automation barrier that shouldn't exist.

Why FII Innovators Pitch

Physical AI can't be proven in a simulator.

AI GO's challenge is more than technical. Systems that act on real equipment demand operational deployment environments, hardware partners, and feedback from organizations automating in complex settings — none of which software alone can supply.

FII's international network, investment ecosystem, and focus on AI-driven societal impact align directly with our mission: scaling AI GO from pilot deployments into globally deployable automation infrastructure for the factories every other solution skips.

01

Real-world deployment environments

Operational settings and industry collaborators to validate natural-language automation where it matters most.

02

Edge infrastructure scaling

Guidance on scaling edge AI across diverse hardware ecosystems while staying affordable for resource-constrained factories.

03

Mentorship & expertise

Strategic mentorship from industrial-automation leaders, robotics experts and infrastructure-focused investors.

04

Cross-sector partnerships

Connections to the industries and regions — including Southeast Asia — that stand to gain most from accessible automation.

ALL GOOD
All for Good

AI GO needs two kinds of fluency — how factories actually run, and how AI systems are actually built. The founding team covers both ends, out of National Tsing Hua University.

HC
Co-Founder · CEO

Hector Chiu

Industrial Engineering + Computer Science, NTHU. Translates production-floor pain points into product requirements — leading strategy, factory relationships, and the government-subsidy channel.

Industrial Eng.Ops & PMGo-to-Market
YC
Co-Founder · CTO

Yuyi Chang

EE + CS, NTHU. Multi-startup CTO with internationally published research in NLP, LLM agents and model efficiency — architecting the runtime, intent engine, cloud-edge system and Skill SDK.

LLM AgentsCloud-EdgeFull-Stack & HPC

⌁ Originating from National Tsing Hua University · Hsinchu, Taiwan

Let's build the control layer

Give every factory a voice its machines understand.

We're connecting AI GO with the partners, pilot sites and investors who will bring natural-language automation to the factories the industry forgot.