AI for accounting offices — with KSeF and no data leaks

Automatic reading of invoices and documents

Process automation

Problem: Data from invoices and paper documents is retyped by hand

How it works: OCR reads data from scans and photos, stores it in a structured form and passes it to a process integrated with KSeF 2.0

What you gain: No more retyping — the accountant checks the result instead of producing it, and retyping errors disappear

Client assistant for accounting questions

Process automation

Problem: Answering clients’ repetitive questions about their accounts takes accountants half a day

How it works: Each client has a separate workspace with data isolation and WRITE/READ roles; the assistant uses only that client’s documents and regulations

What you gain: Answers to typical questions are prepared as drafts for approval, and the accountant regains time for matters that need their expertise

Automatic tracking of deadlines and obligations

Process automation

Problem: Tax and reporting deadlines for dozens of clients are tracked by hand

How it works: Schedules (RRULE) trigger reminders and scripts, a “Morning Triage” sets the day’s priorities, and a Kanban board shows the status of tasks

What you gain: No deadline depends on one person’s memory — the system reminds, organises and shows what is at risk

AI cost settlement per client

Cost and efficiency

Problem: The cost of using AI cannot be assigned to a specific client

How it works: The cost of every query is measured per user, team and project; the internal Twin:Coin currency hides token price lists and lets you set limits

What you gain: AI costs go on the client’s account like any other service — with a limit and an alert before it is exceeded

Checking that regulations and procedures are up to date

Knowledge and documents

Problem: Regulations change faster than internal procedures are updated

How it works: CogOps (CDF-F6) measures how current the knowledge is and flags outdated sources in the quarterly review

What you gain: The firm works on regulations in force and knows which procedures need review before a client gets a wrong answer

A secure AI environment instead of public chatbots

Security and sovereignty

Problem: The risk that an employee pastes a client’s documents into a public chatbot

How it works: Documents marked as confidential are permanently blocked from use in external models, and an ABAC policy enforces this at the gateway

What you gain: The block works technically, not on trust — the firm can show this to the client and during an inspection

What client data must not be put into ChatGPT?

Tools without a data processing agreement and control over data location should not receive personal data, identification numbers, clients’ financial data or documents covered by professional secrecy. The AI use policy template includes a catalog of such data

Who is liable for an AI error, e.g. in a VAT rate?

The accounting firm, toward its client. AI prepares a proposal citing the relevant provision, and the accountant approves it

How do you write an AI policy for an accounting firm?

Based on the AI use policy template: approved tools, a data catalog, responsibility for the output and training