Agentic service
An AI assistant connected to your tools through MCP: invoicing, bank statements, accounting paperwork, email and tasks. It works right inside Claude and Slack, which you already use. It proposes and prepares; decisions with consequences stay with you.
Running itashu's own back office since February 2026. 32 tools, no new app.
A familiar state
Without the assistant
With the assistant
Scope
Invoices from hours worked, including per-client rates and overhead. It watches due dates and prepares the follow-up.
Reads the bank statement and proposes matches against receipts. When a case is ambiguous it does not decide alone; it presents candidates.
Collects invoices from email, filters out duplicates, flags suspicious documents and assembles the folder for accounting.
Prepares the email with the work breakdown in your accountant's language. As a draft; sending is yours.
Searches the inbox, turns an email into a task with a link back to the original message, prepares a draft reply.
A monthly work report from the calendar and time entries as a PDF, stored where you expect it.
Under the hood
The assistant is an MCP server. You use it from Claude or Slack; no new application, no extra login.
Project tool, email, drive and calendar through OAuth grants you can revoke at any time. It has no database of its own; your tools remain the source of truth.
Payment matches come back with a confidence level; suspicious documents get flagged and left in place. It deletes nothing; at most it moves things to the bin.
The matched statement gains a receipt column, receipts gain the vendor name in the filename. The folder you hand to accounting explains itself.
Emails with consequences end as drafts, invoices wait for review. The assistant prepares everything; the last word is yours.
Case study
We built the assistant for our own company. Since February 2026 invoicing, the bank and the monthly accounting paperwork all flow through it. Not a demo prototype; the system our administration runs on.
The assistant is built to measure for one company and its tools. Deploying it for you means connecting to your stack: a different project tool, a different bank, a different accountant. It is not boxed software; it is a project with a clear scope. Which is exactly why we start with a process audit, not an installation.
Where it leads
Matched statements and receipts are not copied into a spreadsheet once a year. The assistant writes them back into the documents, and the same data continuously builds a financial view of the company and the family. No second data entry, no shadow database.
Engagement model
The AI assistant is not a box; it is fitted to your processes and tools. So the engagement has two parts: a one-off deployment and monthly operations, in which the assistant grows with you.
A fixed-scope project
Grows with you
Monthly subscription
The deployment price depends on how many tools and processes get connected; operations on the scope. You will get a quote with an exact number within two business days.
Questions
An assistant built on a large language model, in our case Claude, with your company tools connected as capabilities through MCP (Model Context Protocol). It is not another application: you work with it in Claude or Slack and it reads and prepares things inside the systems you already run.
It is set up to leave consequences to a person. Emails with consequences are prepared as drafts, suspicious documents get flagged and left in place, and deleting means moving to the bin. What it may do alone is defined at deployment, by your rules.
Today ClickUp, Gmail, Google Drive, Google Calendar and Slack. Any other system with an API is a matter of one connector: project tools, CRMs, invoicing services. During the audit we go through your stack and say what connects right away and what needs work.
In your systems, where it has been all along. The assistant has no database of its own: it reads data through grants you give it and can revoke at any time, and writes results back into your documents. It runs on its own instance, not in a shared environment.
Deployment is a project with a price agreed up front based on scope. Operations is a monthly subscription. You will get exact numbers in a quote within two business days after a short call about your processes.
The first tools, typically invoicing or email processing, can run within weeks, not months. The scope then grows step by step based on what proves itself. The exact plan is the output of the initial audit.
More products
Forty minutes looking at how things stand. It tells us how far we can take you towards being an AI-native company, and what that will cost you at the end of the day.
Jaro Rais
AI consultant and developer
Reach out:
jaro.rais@itashu.coGet in touch and we will come back to you within one working day.
By submitting you agree to your data being processed as described in the privacy policy.