Deploying AI inside a company
The capability question is usually the easy one. What decides a deployment is the contract, the data path and whether anyone can show an auditor what happened. This is how we would approach it — including what changes when the company is in Japan.
A rollout that survives review
Decide what you are measuring
Pick one measurable outcome before choosing a tool — hours saved on documentation, time to first draft, tickets deflected. Without it, the pilot ends in opinions.
Clear the contract questions first
Training exclusion, data residency, retention period, subprocessors, certification. Tools that fail these do not belong in the evaluation, regardless of capability.
Pilot with twenty to thirty people
Large enough to produce signal, small enough to avoid procurement. One month is usually sufficient to know whether the tool survives contact with real work.
Write the internal policy before you scale
What may be entered, what may not, who approves exceptions, and where the audit trail lives. Deploying first and writing the policy later is how shadow AI starts.
Six questions that disqualify faster than a demo
- Is exclusion from training written into the agreement, or only stated on a marketing page?
- Where is data processed, and can you choose the region?
- Does SSO and user provisioning match how you already manage accounts?
- Are there audit logs that your internal controls can actually use?
- Is there a contracting entity and support in the language your team works in?
- What happens to your data when the contract ends?
A note on Japan
If the deploying entity is in Japan, three things come up that rarely appear in English-language guidance: whether the vendor has a Japanese contracting entity, whether invoice payment in yen is available, and whether a Tokyo region exists for data residency. Capability rarely decides these deployments. Those three do.
What changes by industry in Japan
Each of these sectors has a specific regulatory shape that determines which tools are even eligible. These guides cover the rules, how to weight selection, and what a realistic first deployment looks like.
AI in healthcare — what gets approved in Japan
Patient data confidentiality, Japan's 3-ministry guidelines, clinical documentation
Read the guideFinanceAI in financial services — clearing the bar in Japan
FISC security standards, data residency, audit trails
Read the guideManufacturingAI in manufacturing — technical secrets and export control
Technical document confidentiality, export control, OT separation
Read the guideSaaS and ITAI for SaaS and IT companies
Engineering productivity, customer data handling, SOC 2
Read the guide