Start with a free diagnosis

You want to use generative AI and AI agents, but you’ve held back because you’re worried about data leaking out. Or you simply don’t know what to decide first. This service is for small and mid-sized companies in exactly that spot.

You can start with the diagnosis alone. It’s free and needs no call: answer the 8-question self-check below, and we’ll reply in writing with the points to tackle first.

Request the free diagnosis (contact form)

Just tick the items that apply to you in the contact form.

1. You don’t know whether staff are using generative AI on personal accounts
2. There’s no written line between what may and may not be entered into AI
3. If a client asked how you use AI or how you protect their data, you couldn’t answer right away
4. You’d like to try AI agents such as Claude Code or Codex, but worry about what they might do on your computers
5. AI output has gone out to clients without being checked (or you worry it might)
6. Only one person in the company knows AI well, or no one does
7. There’s no agreed contact or procedure for when someone enters data by mistake
8. It’s unclear who manages the paid plans and accounts

• Staff can decide for themselves whether something is OK to enter, using a one-page sheet
• Risky actions by AI agents can be blocked by settings (not everything can be prevented)
• When a client asks, you can hand over a clear explanation of your rules and settings
• When in doubt, you have someone to ask by chat or email

A good fit if you

• Want to use generative AI but have held back over data-leak worries
• Have been asked by a client how you manage AI, or expect to be
• Hold a certification such as ISO/IEC 27001 or a privacy certification and want AI handled to the same standard
• Have no one in-house who knows AI well
• Prefer to work in writing rather than in meetings
• Have staff from other countries and want materials in their native languages
• Are a professional practice that handles sensitive personal data, such as a law firm or clinic (we add stricter rules for this)

Example: for an accounting firm, before staff use AI to summarize a client’s financial statements, we sort out what to check and how to configure the tools.

Not a good fit if you

• Want to be walked through things in live meetings
• Have decided to ban AI entirely
• Need same-day turnaround, such as “by tomorrow”

1. A team of AI agents does the work

Specialist agents for legal and privacy, engineering, security and design split the work, overseen by a responsible manager at our company. There are no meetings; we work by email, chat and recorded video.

2. Our own guardrails really stop our own AI

We run AI agents that share memory across four machines (computers and servers), and we hand them work the way we would to staff. From September 20 to 27, 2026, our guardrails stopped our own AI agents:

• 76 times: about to draw a conclusion without checking a tool’s full output
• 6 times: about to write “can’t do it” or “later” without trying
• 3 times: about to answer “I don’t know” without looking it up

Counted from our own operation logs as of September 27, 2026. The records are available for review with confidential details redacted.

3. We run a certified privacy program ourselves

We hold the Privacy Mark (Japan’s personal-data protection certification, granted by JIPDEC under JIS Q 15001), held since 2006, and run it in-house. The AI use policy we fill in for your company is written from that operating experience.

4. Written support in your staff’s language

We answer in your staff’s native language — English, Chinese, Korean, Vietnamese, Spanish, Portuguese and others (ask us about other languages). Everything is in writing, so there’s no need to match time zones.

1. Contact us
   Send a request through the form.
2. Diagnosis
   Based on the items you ticked, we may email a few follow-up questions, then reply in writing.
3. Quote and agreement
   We tell you the scope you need and the price.
4. Access
   Your administrator grants us an account with the permissions we need. We never take your passwords. If you don’t have an administrator account yet, we’ll walk you through creating one, with screenshots.
5. Rules, setup and explainer video
   We deliver each piece as soon as it’s ready.
6. One month of questions
   At the end, your administrator removes our access. You can also switch to the monthly advisory.

1. Purpose and scope
2. Definitions (generative AI, AI agents, input data)
3. Approved tools and how to request access
4. Information that must not be entered
5. Checks when handling personal data (purpose, outsourcing, sharing with third parties)
6. Checks before using AI output (facts, rights, publishing outside the company)
7. What AI agents may do, and what they must be stopped from doing
8. Account and permission management
9. Reporting mistaken input and incidents
10. Training and regular review

Appendices: one-page staff sheet / register of tools in use

Example: your question

Our staff want to enter customer names into AI to write up meeting minutes. Is that OK?

Example: our reply

Yes, as long as writing up minutes is within the purpose for which you collected that customer information. Use a business plan where your input is not used for training, and turn off feedback sending (such as rating buttons) in the settings. Under your rules, this falls into the “OK to enter as is” tier. We’ll add this check to your record. This reflects Japan’s APPI. If the GDPR applies to you, you will also need a data processing agreement with the AI provider and a lawful basis for the processing; please confirm with your legal counsel.

Start with the free diagnosis

Free, no call needed: answer the 8-question self-check and get a written reply.

Request the free diagnosis (contact form)