55-second overview: a private daily assistant, spreadsheet formulas written by cloud AI, and recurring reports that run in-house and wait for your sign-off.
Selective data sovereignty
Keep confidential files fully offline on your servers, or switch on cloud AI per task when you choose.
Build with leading cloud AI, run in-house
Let leading cloud AI build the workflow once. Run recurring invoice checks and sales analyses repeatably in real code and local models.
Verifiable work & human review
Real spreadsheet formulas instead of guessed numbers, full document change history, and human sign-off before anything is sent.
Everyone is building the proactive assistant. Ours runs where you decide, and at the end of the day you can see and verify everything it did.
08:00
Morning briefing
Before you open your inbox: the client inquiries that matter, today's scheduling conflicts, overdue milestones. Ask it to block prep time for your afternoon meetings, say "Book them", and it's done.
Your choice of model
Jump to:
Watch the assistant synthesize executive priorities and book time blocks directly into the calendar on a single confirmation.
11:00
Expense audit
"Find all unapproved expenses over €500 and show me the formulas." The assistant writes real formulas into the spreadsheet. You check the total yourself: €7,494.40.
Figures from formulas, not from the model
Jump to:
The assistant audits spreadsheet data: generating formulas, policy disallowance checks, and audit aggregates directly in the sheet.
16:00
Monthly sales report
The report arrives by email. The assistant analyses it, writes the board narrative and builds the slide deck, then stops and waits for your approval. Nothing goes out without you.
Waits for your approval
Jump to:
Workflow execution: data analysis, structured narrative, your sign-off, and presentation generation.
18:00
"What did the AI change?"
Open any document and see its history: every change, when it happened, and where it came from. Go back to any earlier version. Workflow runs and approvals are recorded the same way.
Recorded on your infrastructure
Jump to:
The assistant restructures a client proposal; the document history shows every change and where it came from.
Zokai = zōka (増加, Japanese for growth) + AI. Every document, workflow and approval adds context, so your assistant gets more useful the longer you work with it.
Client files and deal data stay inside your perimeter. Review contract archives, prepare compliance checklists and run routine due-diligence checks, with a record of every change.
Consultancies and SME leadership
Turn recurring client deliverables into workflows. Calculations run in code rather than on paid AI tokens, and the AI steps can run on local models.
Regulated European organisations
Built to help you meet GDPR, DORA and NIS2 requirements: you choose the hosting jurisdiction, confidential data can stay fully offline, and every change is recorded.
AI proposes. You decide. Every time.
You decide
Where it runs
Your own hardware, your data centre, the provider you trust, or dedicated servers we operate for you in Germany.
Which AI, for which task
Fully local models for confidential work, cloud models when you choose them. Switch at any time. In fully offline mode, nothing leaves your servers.
What it may do
The assistant only gets the tools you allow. High-stakes steps like sending, booking or approving wait for a person.
Your files stay yours
Documents, spreadsheets, notebooks and workflows are stored in open, standard formats. They're readable without Zokai Station, so you can leave whenever you want.
AI without the black box
What the assistant wrote
# Flag likely duplicate invoices
for a, b in pairs(invoices):
same_amount = a.amount == b.amount
similar_vendor = similarity(a.vendor, b.vendor) > 0.9
if same_amount and similar_vendor:
flag(b, reason=f"matches row {a.row}")
What you see
Read last month's invoices
Find same amount and a similar supplier name
Mark possible duplicates in the sheet
Wait for your approval
Approve
Under the hood, Zokai Station writes real code to do the work: formulas, scripts, workflow steps. You never have to read it. You see plain steps, test them on last month's data, and approve. The code stays there for anyone who wants to check it.
Correct steps in plain English.
"Skip invoices under €100." "Ask me before anything goes to the board." The assistant updates the workflow, checks it, and runs the test again.
Missing a function? The assistant builds it.
There's no feature wall. When the spreadsheet or a workflow lacks something, the assistant writes it as a script you can inspect, and keep.
Tested before it goes live.
Test runs replay a real past event. Reading is real; sending and writing are blocked.
Everything is recorded.
Every change, run and approval goes into a record that you can verify yourself.
Start with one process
You don't have to move your whole office. Start with one sensitive, recurring process, the one you'd never paste into a public chatbot. Run it in Zokai Station on your own data for four weeks. Keep everything else as it is.
It runs alongside the tools you already use. For the pilot process, upload the files it needs (Excel, CSV, Word, PDF). Data from your ERP or accounting system comes in as standard exports. Mail and calendar connect directly to Gmail and Google Calendar today. Microsoft 365 is in development. Until it's ready, your team keeps its current email and office tools, and Zokai handles the sensitive process in parallel.
Early Access: Zokai Station is in Early Access and available to selected organisations through commercial pilots. Scope, hosting, support and pricing are agreed individually for each pilot. Local models have no per-use AI fees.