Building an organisation that keeps learning.
Terakoya is a desktop app that connects your company’s records with AI. Everyone works from the same context, and what the work teaches stays with the company.
- 1Ask
Ask for work in plain words in the chat.
- 2Read
AI finds the relevant records in the company’s notes, listed on the right.
- 3Draft
A draft with sources. A person checks it at the end.
- 4Keep
What you checked goes back into the notes, ready for the next person.
Shared intelligence at the centre of the organisation.
Inside a company, information travels through the hierarchy. Reports go up, decisions come down, and the same explanation is repeated again and again.
With Terakoya, principles, decisions, ways of working and lessons go into an organisation brain, and executives and staff work from the same context. People, roles and AI models change; the knowledge stays with the company.
The next job starts from what the organisation has learned.
Start with one team and one task.
Connect
Install the app and connect your drive and AI. Choose the first task.
Ask
Ask for work in the chat. AI reads the notes and records each finished job.
Grow
The notes grow, and repeating the same explanations fades away.
Spread
Add tasks one by one, then other teams.
We work this way ourselves, every day.
Three of us, with no dedicated accounting or legal staff. Real jobs where a person asked, AI did the work, and a person decided.
Receipts scattered across paper, web and email, turned into a ledger that balances
- What we asked
- “Join the passbook and the web statements, collect this year’s receipts, and hand it all to accounting.”
- What AI did
- Rebuilt 34 transactions and checked the balance; collected invoices from suppliers’ sites and verified 477 after removing duplicates.
- What a person decided
- Explained unclear deposits; the owner moved it into the master file and the tax accountant decided the treatment.
- Result
- Three statement sheets and 477 PDFs handed over with sources, reconciliation records and open items.
Experience and evidence, turned into a Word manuscript ready to submit
- What we asked
- “Turn our experience running AI staff into an article for the journal.”
- What AI did
- Researched the related records, drafted the structure, text and figures, and revised to the journal’s guidelines.
- What a person decided
- Checked the experience and argument, and decided what to cut and what to keep.
- Result
- Manuscript sent for review as Word and PDF.
A lesson plan, turned into an app teachers can try
- What we asked
- “Turn the 90-minute class into screens teachers and students can try.”
- What AI did
- Turned the lesson flow into screens and code, with the teacher’s and the students’ steps.
- What a person decided
- Set the learning goals and the flow, tried it and gave corrections.
- Result
- A working prototype (not yet in use with real student data).
Five years of chat, sorted into what can be shared and what stays private
- What we asked
- “Preserve the whole Slack history and separate what can go into the company notes from what must stay private.”
- What AI did
- Read 4,492 records and 397 files one by one, copied the key points into notes, kept confidential items closed and marked deleted ones.
- What a person decided
- Approved the export and automation, decided what could be shared, and kept the original data.
- Result
- 428 records into the notes, 380 kept as confidential.
Real work we did ourselves with the predecessor of Terakoya. Screens are displays made from the real files, with accounts, counterparties and amounts hidden.
Price¥10,000 per person, per year.Includes a year of updates. Setup support and training are free for now.
Pricing and questions →Why not try it on one task?
We’ll walk you through it with the app on screen.