Services / AI
Generative AI with Claude
From finding where it pays off, through a proof of concept, to a system your team runs without us. Built on the quality control and security practice we have run as a contract developer since 2012.
How an engagement runs
Each step can be bought on its own, or run end to end.
Assessment
We map the work as it is done today, find where generative AI changes the cost, and say where it does not.
Proof of concept
One narrow task, built and measured, until you can decide whether it goes to production.
Production
Integration with existing systems, permissions, logging and audit trails, acceptance testing, then operation.
Enablement
Training and written rules so your own people keep it running and improve it after we leave.
Where it tends to pay off
Internal operations
Applications and enquiries, searching and summarising internal documents, routine reporting.
Customer support
Drafting first replies, searching past cases, analysing response quality.
Software development
Reading and migrating legacy code, generating tests, implementing from written specifications.
Document processing
Extracting fields from forms and contracts, structuring unstructured documents, working across languages.
How we handle your data
Not used for training
We configure engagements so that data you give us is not used to train models, and we secure that in the contract as well as in the settings.
Access control and logs
Permissions per user and per purpose, with operation logs retained. Managed under our ISO/IEC 27001 certified system.
A person checks the output
Generated output is not used as a decision on its own. Who reviews it, and what they are responsible for, is written into the workflow.
More detail on Security & Trust. Our full AI usage policy is published in Japanese at AI利用方針.
Questions we are asked
We want to start small. What is the smallest useful engagement?
An assessment on its own, or a proof of concept limited to a single task.
Our data cannot leave our environment.
We fix the boundary during requirements and show you the data flow as a diagram before any work starts.
Can this connect to systems we already run?
Yes. Contract development and maintenance of core systems is our main business, so we design around existing assets rather than replacing them.
Will our team be able to run it afterwards?
That is what the enablement step is for — written operating rules and training, not just a handover document.
Start with the problem, not the tool
You are welcome to talk to us before the problem is fully defined.
Contact form