You can talk to us before requirements are fixed.
Guide: start with a ~10-person pilot; build quoted per project, usage-based fees|See pricing
Private AI Workspace
AI that reaches your own data — paid by usage
The main reason AI tools go unused is that they cannot reach the data people work with. We put a familiar chat screen under your control, connect it to AI providers on usage-based pricing, and link it to your CRM and business tools via MCP — so requests like "list this month’s deals" or "draft a proposal for this client" are answered from your own data. We start small: about ten people, read-only.
Common Situations
- You rolled out an AI tool, but it cannot reach your CRM or internal systems, so nobody uses it
- Your flat-rate internal chat AI keeps answers short to stay under its cap
- Some staff pay for AI personally, and the company cannot see what goes into it
- Every time you need CRM data, someone exports a CSV and reshapes it by hand
- Premium copilots and connectors cost too much to license for everyone
- You need to confirm that data stays inside and nobody can act beyond their permissions
What We Build
A familiar chat screen, in-house
An open-source chat platform on your own cloud or servers, switching between Claude, GPT, Gemini and others as the task needs.
Connected to your tools via MCP
We connect CRM, groupware and task tools through MCP, the protocol that lets AI use other software — building connectors via API where no official one exists.
Read-only first
We start with read-only credentials. Worries such as "could it advance a deal stage on its own?" are solved by making writes impossible.
Sign in with company accounts
Single sign-on with Microsoft or Google accounts, so leavers lose access through your existing identity management.
Usage caps and logs
Set caps per person or team and record who used how much, so you can adjust caps and compare cost with results.
Spreading what works
We collect how your best users work and make it available to everyone. Role-specific training is available separately.
Compared With Per-Seat AI Tools
| Per-seat AI tools | Private AI workspace | |
|---|---|---|
| Pricing | Same fee for light and heavy users | Pay for usage, with caps |
| Internal data | Often needs a premium tier or another tool | Connected directly via MCP |
| Models | Limited to the vendor’s models | Switch between Claude, GPT, Gemini and more |
| Hosting and logs | Vendor environment | Environment you control |
| Time to start | Immediate after signing | About a month to prepare and build |
* If you already use a business AI plan, we can design around using both.
How We Start
Preparation (about 2 weeks)
A short survey on current AI use and what people wanted but could not do; then we choose the tools to connect and a pilot group of about ten. We check your internal guidelines here too.
Build (2–3 weeks)
Screen, sign-in, caps, logs, read-only connections and security settings. Security settings take most of the time.
Pilot (2 months)
About ten people use it, sharing techniques weekly. We set the same cap as your current chat AI to compare cost and results.
Decide whether to expand
Use the pilot numbers to decide whether to widen access, allow writes, or add search over internal documents (RAG).
How costs work
Costs split into build and usage. Build depends on the number of tools and your security requirements, so we quote per project. Usage is billed by the AI provider and can be paid by you directly. Two months of pilot data shows the real cost per person, so the decision to expand rests on actual numbers.
Related Services
Pricing guide
Prices are pre-tax guides and vary with scope, integrations, and post-launch operations. Formal quotes are free.
Related work
Projects close to this service. Those with challenge, scope, and stack described come first.
Case study株式会社Nature Innovation Group
傘レンタルアプリ「アイカサ」
A LINE app, rental-spot hardware, and payments — launched as one system
Scope|デザイン / コーディング / システム開発
Case study学校法人松本学園（広島桜が丘高等学校・広島県瀬戸内高等学校）
学食モバイルオーダー
Students pre-order and pay on their phones, easing cafeteria queues and cash handling
Scope|デザイン / コーディング / システム開発
CHIeru株式会社
英語学習サービス「トイテイク」
Scope|デザイン / コーディング / システム開発
Private AI Workspace
How is this different from Microsoft Copilot or ChatGPT business plans?
Business plans charge per seat, so heavy and light users cost the same, and reaching your CRM or custom systems often needs a premium tier or another tool. A private workspace is paid by usage and connects to whatever you choose. If you already have a business plan, we can design around using both.
Does our data leave the company?
The screen and logs run in an environment you control. Prompts are sent to the AI provider to generate answers, under business terms where inputs are not used for training. Which data may be used is agreed against your guidelines before the pilot.
What does it cost?
Build cost depends on the number of tools and your security requirements, so we quote per project. Most clients start with a pilot of about ten people, which keeps the upfront cost modest. Usage fees go to the AI provider and can be paid by you directly.
Which tools can it connect to?
Tools with official MCP connectors — many CRMs, groupware and task tools — usually work as they are. Where there is no connector but a public API, we can build one. Tell us what you use.
Could the AI change our data on its own?
The pilot runs with read-only permissions. If writes are needed later, we create a narrowly scoped account and add limits on the management side.
Will our staff actually use it?
During the pilot we share techniques weekly and spread what your best users do. Role-specific training is also available separately.
You can talk to us before requirements are finished.
We can start by sorting what to do first. We run businesses ourselves, so the conversation stays about labor and ops. Consultation and estimates are free.
Online meetings / reply by next business day / free consult & estimate