NETWORK: DISCONNECTED AI INFERENCE: RUNNING ── Still thinking, fully offline.

The AI that keeps your secrets in-house.

Have you told a cloud AI about your trade secrets, your plans, or someone's personal data?
GPT-AILS 1 is a general-purpose local AI server that runs fully disconnected from the internet. Your data never leaves the building.

FULLY OFFLINE DEDICATED RAG + TRAINING JAPANESE-OPTIMIZED / MULTILINGUAL ENTERPRISE AI FOR SMEs

VISUAL OVERVIEW

ITO-japan provides app integration, deployment support,
and custom development end to end.

As the authorized distributor, ITO-japan supports more than the sale of GPT-AILS 1. We can provide employee smartphone app integration, receipt capture, invoice uploads, internal document search, approval workflows, custom application development, and deployment support. The system is designed for internal intranets or closed networks, keeping data inside your company.

HIGH SECURITY

Local AI and a closed network keep confidential data inside.

FLEXIBLE CUSTOMIZATION

Design apps, screens, and workflows for each department.

ROLLOUT SUPPORT

ITO-japan supports requirements definition, deployment, and post-launch improvements.

THE PROBLEM WITH CLOUD AI

"Are you sure you should be typing that in?"

Cloud-based AI has structural weaknesses for business use. These concerns come up everywhere:

"We don't know where the data is stored." "We don't know who manages it." "We can't tell if it's safe to enter confidential information."

RISK 01

Data leaves your environment

Accounting records, HR files, customer correspondence — all sent to someone else's server. "We can't enter this into an AI" is a problem at every company. This is cloud AI's greatest weakness.

RISK 02

You can't train it on your own data

Cloud AI providers generally prohibit serious fine-tuning on company data and closed, secure training of company-specific models. The result: you can never build "our own AI."

RISK 03

Fragile in a disaster

The moment the network goes down, so does your AI. Disasters, outages, network isolation, internal network policies — your operations stop with the connection.

THE ANSWER: LOCAL AI

GPT-AILS 1
solves all three.

A local AI runs entirely offline. AILS 1 learns from your data on the machine itself, protecting confidentiality while delivering answers.

Not one byte leaves the building

Input, training, and inference all happen inside the server. High-level information protection, by architecture.

Build "our own AI" on your data

Dedicated RAG and on-device training maximize the value of internal data and keep hallucinations low.

The network dies. Your AI doesn't.

Designed offline-first, AILS keeps working through disasters and outages.

Enterprise-grade AI, sized for SMEs

The first serious AI server designed to be deployable by small and medium businesses.

IMPROVEMENT PROPOSALS, BY THE AI

If AILS worked at your company —
here's what it would propose.

This isn't about exotic industries. Meeting minutes, invoices, policy questions, hunting for old quotes — the everyday office work every company has. Read these as proposals from AILS itself; every one runs without your data leaving the office.

FROM: AILS — TO: MANAGEMENT & ALL DEPARTMENTS (8 PROPOSALS)
PROPOSAL 001ALL DEPTS

Let me take the meeting minutes.

Hand me the recording of any meeting and Sound2txt will transcribe it — speaker by speaker — then summarize and pull out the decisions and action items. Even confidential topics never leave the office.

Sound2txtSummaries & action items Effect: minutes in minutes, not hours
PROPOSAL 002ACCOUNTING

That pile of invoices and receipts? I'll read it.

Invoices, receipts, and purchase orders — paper or scanned PDF — get digitized by Image2txt with distortion correction. Ask "last month's invoice from that supplier?" and the document appears. No vendor data ever goes to a cloud.

Image2txtRAG Effect: far less typing and searching
PROPOSAL 003GENERAL AFFAIRS / HR ADMIN

"What do the rules say about this?" — ask me.

Drop your work rules, expense policies, approval procedures, and past notices into a folder. I answer employee questions — travel allowances, leave entitlements — with the source document attached.

RAGAnswers with sources Effect: fewer interruptions to admin staff
PROPOSAL 004SALES ADMIN

Even a quote from three years ago — found in a second.

Trained on past quotes, orders, delivery slips, and deal notes, I answer "what did we quote them last time?" instantly. Vector search finds it even from vague wording — and I'll draft the next quote from similar cases.

RAGDrafting Effect: faster quoting and retrieval
PROPOSAL 005CUSTOMER SERVICE

Customer replies — I'll write the first draft.

Given an inquiry and your past correspondence, I draft the reply. Testing showed strong results in emotionally-aware responses — apologies and polite refusals read naturally. Recurring questions become an FAQ.

Emotion-aware repliesFAQ building Effect: better replies, faster
PROPOSAL 006HR / TRAINING

Training videos and veteran know-how, turned into manuals.

Mov2txt transcribes training sessions and how-to videos, then summarizes them into manuals — capturing "only-they-know-how" work before retirements and transfers. New hires can then ask me instead of interrupting seniors.

Mov2txtSummarize → manualize Effect: less key-person risk, faster onboarding
PROPOSAL 007MANAGEMENT

Daily reports, monthly reports, meeting docs — summarized for you.

I learn the reports and meeting documents that pile up, and answer cross-cutting questions like "what happened in the field this month?" or "where does that project stand?" — even for documents with financial and personnel data, since nothing leaves.

RAGCross-search & summary Effect: faster situational awareness
PROPOSAL 008COMPANY-WIDE / BCP

On the day the network dies, I'll still be working.

When disasters or outages take every cloud service down, I keep running offline — policy lookups, document work, and record-keeping continue as usual. Make me the core of your business continuity plan.

Offline operationBCP Effect: office work survives outages

* Results vary with your data and operations. In a demo, AILS itself will generate proposals tailored to your business.

PROVEN IN TESTING

67,000+ court cases learned.
Searched by meaning, in ~10 seconds.

In our tests, AILS learned the full set of 67,000+ court precedents held by the National Diet Library, and achieved intent-aware vector search — not mere keyword matching — in about 10 seconds (AILS pro model). On your documents, the same power works in far more familiar ways.

67,000+
Court precedents learned
~10s
Vector search response
1
Folders needed to ingest

The corpus mixed dense legal text, images, tables, and scanned documents — yet a single drop folder was all it took to organize everything. Emotion-aware responses and analysis also showed strong results.

LINEUP

Three grades, sized to your workload.

Every grade includes dedicated RAG tuning on your company's data. Full specs on the product page.

ENTRY

GPT-AILS E1

¥3,600,000 (tax incl.)

  • GPURTX PRO 4000 24GB
  • MemoryDDR5 64GB
  • RAG tuning7 days (56h)
See details
RECOMMENDED

PRO

GPT-AILS P1

¥4,800,000 (tax incl.)

  • GPURTX PRO 5000 48GB
  • MemoryDDR5 96GB
  • RAG tuning10 days (80h)
See details

MONSTER

GPT-AILS M1

¥12,000,000 (tax incl.)

  • GPURTX PRO 6000 96GB
  • MemoryDDR5 256GB
  • RAG tuning30 days (240h)
See details

* Specifications and prices are indicative and vary with order timing and customization. Optional monthly support (from ¥10,000) includes system version upgrades plus email & phone support.

CONTACT

Tell us about the work
you couldn't give to AI.

Demos, quotes, and consultations are handled by ITO-japan, the authorized distributor. In a hurry? The chat in the corner works too.

INQUIRY FORM

Talk to ITO-japan CORP

AUTHORIZED DISTRIBUTOR — YOUR CONTACT

ITO-japan

SERVICES
Sales, consultation, demo arrangement, and domestic support for the GPT-AILS series
MAIL
[email protected]

DEVELOPER / MANUFACTURER

MIDOSY Inc.

ADDRESS
5-9-10 Nishitemma, Kita-ku, Osaka 530-0047, Japan