AI in accounting

Key information

A training for accountants working in the Microsoft 365 environment who want to make real use of Microsoft Copilot in everyday tasks: client correspondence, drafting accounting explanations, analysing financial reports and preparing reports. Participants learn to write prompts that deliver specific, ready-to-use results, and understand the limits of safe AI use when working with clients’ financial and tax data.

During the training, participants will find out, among other things:

• How AI thinks, where its limitations lie and what hallucinations are (crucial when working with numerical data)
• How to use Microsoft Copilot in Microsoft Outlook for correspondence with the accounting firm’s clients
• How to use Microsoft Copilot in Microsoft Word to draft explanations, letters to public authorities and internal documents
• How to use Microsoft Copilot in Microsoft Excel to analyse reports, categorise costs and spot anomalies
• How to write prompts with the KREDO method to get precise and specific answers
• What the safety rules are for working with AI on clients’ tax, financial and personal data

The training combines short introductions with hands-on workshop exercises based on real scenarios from an accountant’s work.

Objective

The objective of the training is to familiarise participants with the possibilities of Microsoft Copilot in everyday accounting work, covering the use of Microsoft Copilot in Microsoft Outlook, Microsoft Word and Microsoft Excel, prompting with the KREDO method, and the safety rules for working with AI on financial and tax data. The training emphasises specific, repeatable workflows and critical evaluation of AI results, with particular attention to verifying numerical data.

Programme

Module 1: How AI works and what it means for working with financial data

• Language models as predicting the next words
• AI hallucinations: what they are and why they are especially dangerous when working with numbers
• The difference between public models (ChatGPT, Gemini) and Microsoft Copilot in the Microsoft 365 environment

Module 2: Prompting with the KREDO method

• Context: description of the situation and the recipient (individual client, company, public authority)
• Role: who AI should be in this task (accountant, editor, adviser)
• Edit: response format, length, structure, tone (formal letter, email to a client, internal note)
• Action: a precise instruction on what AI is to do
• Reference: a template or example that AI should refer to (previous letter, template, excerpt of a document)

Module 4: Microsoft Copilot in Microsoft Excel

• Analysing invoice reports
• Categorising costs
• Checking for basic anomalies in entries
• Generating formulas from a description in natural language
• Working with data
• Critical verification of results: when AI is right, when it is wrong, when it makes things up

Module 3: Microsoft Copilot in client communication (Microsoft Word, Microsoft Outlook and Microsoft Teams)

• Microsoft Word: drafting accounting explanations for clients, letters to public authorities (tax office, ZUS), templates for internal documents
• Answering the accounting firm’s clients’ questions, summaries of long email threads
• Draft replies to typical enquiries (deadlines, documents, explanations)
• Notes from client meetings in Microsoft Teams and searching for earlier arrangements
• Preparing for a conversation with a client based on correspondence history
• A short exercise on scenarios prepared by the trainer

Module 5: Security and responsible work with AI

• What can be put into public AI and what stays within Microsoft Copilot with company data protection
• Working with clients’ personal data and the GDPR: what is and is not permitted
• Hallucinations when working with numbers and dates: how to recognise and verify them
• Ethics of using AI in communication with clients and public authorities
• Questions and answers

Methodological note

The training works with a single main tool only (Microsoft Copilot from a Microsoft 365 subscription), so that within 6 hours the group can genuinely master how to use it in the context of accounting work, rather than merely see a few demonstrations. Elements of working with public AI models (ChatGPT, Gemini) appear in the security module in a comparative form, but not as a separate workshop track.

The workshop examples in modules 3 and 4 are built around typical scenarios from an accountant’s work: client correspondence, drafting letters, analysing reports.

Organisational requirements on the client’s side

• Every participant has an active Microsoft 365 subscription (Business or Enterprise) and works on a logged-in work account
• Every participant works on their own computer (laptop), not on a shared computer
• Stable Wi-Fi in the room, sufficient for the whole group working at the same time
• A projector or large screen for the trainer’s presentation
• A room suitable for workshop-style work (tables, space for notes)
• A flipchart or whiteboard with markers

Full integration of Microsoft Copilot with Microsoft Word and Microsoft Excel additionally requires the Microsoft 365 Copilot add-on. Without the add-on, participants use Microsoft Copilot Chat (available in the browser with a logged-in M365 account), which is enough for most exercises, but some scenarios in Module 4 are then carried out in a “copy and paste into Microsoft Copilot Chat” mode instead of directly in the applications. The option to be decided before the date is agreed.

Trainer

Franciszek Kołpanowicz, entrepreneur and certified AI trainer. He has been published in Rzeczpospolita, among others. He has over 2,000 hours of experience working with generative AI and has shared his knowledge at public events, including the “Manager in the Age of AI” training organised by Rzeczpospolita and at the UKEN and PASE conferences. His background in philosophy (Jagiellonian University) and his experience in IT quality control allow him to combine analytical thinking with asking the right questions. In training sessions he pays attention above all to the participants, drawing on their experience and needs as the work unfolds.

Key information

A training for accounting firm staff who want to make real use of generative AI in everyday tasks: client correspondence, drafting explanations and letters to public authorities, analysing reports and preparing reports. Participants work with the free version of Gemini, learn to write prompts that deliver specific, ready-to-use results, and understand the limits of safe AI use when working with clients’ financial, tax and personal data.

During the training, participants will find out, among other things:

• How generative AI works, where its limitations lie and what hallucinations are (crucial when working with numbers)
• How to write prompts with the KREDO method to get precise and specific answers
• How to use Gemini for correspondence with the firm’s clients and for letters to public authorities (tax office, ZUS)
• How to analyse reports, categorise costs and spot anomalies in Gemini
• How to generate spreadsheet formulas from a description in natural language
• What the safety rules are for working with free AI on clients’ financial and personal data

The training combines short introductions with hands-on workshop exercises based on real scenarios from an accounting firm’s work, carried out on demonstration or anonymised data.

Objective

The objective of the training is to familiarise participants with the possibilities of Gemini in the everyday work of an accounting firm, covering the drafting of letters and correspondence, analysis of reports and documents, prompting with the KREDO method, and the safety rules for working with free AI on financial, tax and personal data. The training emphasises specific, repeatable workflows and critical evaluation of AI results, with particular attention to verifying numerical data and protecting client data.

Programme

Module 1: Fundamentals of generative AI in working with financial data

• Participants share their experience of working with AI
• Key concepts related to how AI works: prompt, token, hallucination, context
• AI hallucinations: what they are and why they are particularly dangerous when working with numbers and deadlines
• Comparison of the key AI models (ChatGPT, Gemini, Copilot) and why we work with Gemini in the training
• A public model and client data: what it means when data goes outside

Module 2: Prompting with the KREDO method

• Context: description of the situation and the recipient (individual client, company, public authority)
• Role: who AI should be in this task (accountant, editor, adviser)
• Edit: response format, length, structure, tone (formal letter, email to a client, internal note)
• Action: a precise instruction on what AI is to do
• Reference: a template or example that AI should refer to (previous letter, template, excerpt of a document)

Module 3: Gemini in the accounting firm’s client communication

• Drafting accounting explanations for clients and letters to public authorities (tax office, ZUS)
• Draft replies to typical client enquiries (deadlines, missing documents, explanations)
• Summaries of long email threads and searching for earlier arrangements
• Preparing for a conversation with a client based on correspondence history
• Transferring results from Gemini to the word processor, spreadsheet and email (copy-paste workflow)
• A short exercise on scenarios prepared by the trainer

Module 4: Gemini in analysing reports and documents

• Analysing invoice reports and categorising costs using demonstration data
• Checking for basic anomalies and unusual entries in reports
• Generating spreadsheet formulas from a description in natural language
• Reading and summarising documents (PDFs, scans, contracts) uploaded to Gemini
• Critical verification of results: when AI is right, when it is wrong, when it makes things up

Module 5: Security, GDPR and responsible work with free AI

• What may and may not be entered into public (free) Gemini
• Anonymising client data before passing it to AI (tax ID (NIP), surnames, amounts, personal data)
• Working with the personal data of clients and their employees and the GDPR: what is and is not permitted
• Hallucinations when working with numbers and dates: how to recognise and verify them
• Ethics of using AI in communication with clients and public authorities
• Questions and answers

Methodological note

The training works with a single main tool only (the free version of Gemini), so that within 6 hours the group can genuinely master how to use it in the context of an accounting firm’s work, rather than merely see a few demonstrations.

The data analysis exercises in module 4 are carried out on demonstration or anonymised data. No real client data is entered into the public, free AI model.

Trainer

Franciszek Kołpanowicz, entrepreneur and certified AI trainer. He has been published in Rzeczpospolita, among others. He has over 2,000 hours of work with generative AI and has shared his knowledge at public events, including the “Manager in the Age of AI” training organised by Rzeczpospolita and the UKEN and PASE conferences. His background in philosophy (Jagiellonian University) and his experience in IT quality control allow him to combine analytical thinking with asking the right questions. In training sessions he pays attention first and foremost to the participants, drawing on their experience and needs as the work unfolds.

Prerequisites

Good knowledge of using a computer and an office suite (email, word processor, spreadsheet). Prior experience with generative AI is not required.

Organisational requirements on the client’s side

• Before the training, each participant has a free Google account with access to Gemini (gemini.google.com) that has been set up and tested
• Login checked in advance on the device the participant will work on
• One workstation per person, a laptop for every participant, no sharing
• Stable Wi-Fi in the room (on-site format) or a stable connection for participants (remote format), sufficient for 10 people working at the same time
• Projector or large screen for the trainer’s presentation (on-site format) or a videoconferencing platform (remote format)
• A room suitable for workshop work, with tables and space for notes (on-site format)
• Flipchart or whiteboard with markers (on-site format)

The training works with a single tool, Gemini, so that within 6 hours participants can genuinely master how to use it in the context of their own work, rather than merely see a few demonstrations.

To be confirmed before the training

Licences and access:

• Does every participant have their OWN work account, or do they share one?
• Which Google Workspace plan does the organisation have?
• Gemini access to Google Workspace applications (Gmail, Drive, Docs, Sheets) switched on: a setting on the organisation’s administrator side

Participants’ technical readiness (CRITICAL):

• Every participant must arrive with a Google Workspace account already signed in on their own laptop: we do NOT set up accounts in the training room
• Please run a test before the training: each participant opens gemini.google.com on their work account and checks that they can ask a question and receive an answer
• If any participant does not have access to Gemini: please let us know in advance and we will discuss an alternative route

Infrastructure:

• Stable Wi-Fi

Key information

A training for accounting department staff who want to make real use of generative AI in everyday tasks: correspondence with counterparties, handling enquiries from other departments, drafting letters to public authorities, analysing reports and preparing reports for the management board. Participants work with the free version of Gemini, learn to write prompts that deliver specific, ready-to-use results, and understand the limits of safe AI use when working with the company’s financial, tax and personal data.

During the training, participants will find out, among other things:

• How generative AI works, where its limitations lie and what hallucinations are (crucial when working with numbers)
• How to write prompts with the KREDO method to get precise and specific answers
• How to use Gemini for correspondence with counterparties, for handling enquiries from other departments and for letters to public authorities (tax office, ZUS)
• How to analyse reports, categorise costs and spot anomalies in Gemini
• How to generate spreadsheet formulas from a description in natural language
• What the safety rules are for working with free AI on the financial and personal data of the company and its counterparties

The training combines short introductions with hands-on workshop exercises based on real scenarios from an accounting department’s work, carried out on demonstration or anonymised data.

Objective

The objective of the training is to familiarise participants with the possibilities of Gemini in the everyday work of an accounting department, covering the drafting of letters and correspondence, analysis of reports and documents, prompting with the KREDO method, and the safety rules for working with free AI on financial, tax and personal data. The training emphasises specific, repeatable workflows and critical evaluation of AI results, with particular attention to verifying numerical data and protecting the data of the company and its counterparties.

Programme

Module 1: Fundamentals of generative AI in working with financial data

• Participants share their experience of working with AI
• Key concepts related to how AI works: prompt, token, hallucination, context
• AI hallucinations: what they are and why they are particularly dangerous when working with numbers and deadlines
• Comparison of the key AI models (ChatGPT, Gemini, Copilot) and why we work with Gemini in the training
• A public model and company data: what it means when data goes outside

Module 2: Prompting with the KREDO method

• Context: description of the situation and the recipient (counterparty, another department, management board, public authority)
• Role: who AI should be in this task (accountant, editor, adviser)
• Edit: response format, length, structure, tone (formal letter, email to a client, internal note)
• Action: a precise instruction on what AI is to do
• Reference: a template or example that AI should refer to (previous letter, template, excerpt of a document)

Module 3: Gemini in an accounting department’s communication

• Drafting accounting explanations for the management board and other departments, and letters to public authorities (tax office, ZUS)
• Draft replies to typical enquiries from counterparties and internal departments (payment deadlines, missing documents, explanations of settlements)
• Summaries of long email threads and searching for earlier arrangements
• Preparing for a conversation with a counterparty or the management board based on correspondence history
• Transferring results from Gemini to the word processor, spreadsheet and email (copy-paste workflow)
• A short exercise on scenarios prepared by the trainer

Module 4: Gemini in analysing reports and documents

• Analysing invoice reports and categorising costs using demonstration data, including by cost centre
• Checking for basic anomalies and unusual entries in reports
• Generating spreadsheet formulas from a description in natural language
• Reading and summarising documents (PDFs, scans, contracts with counterparties) uploaded to Gemini
• Critical verification of results: when AI is right, when it is wrong, when it makes things up

Module 5: Security, GDPR and responsible work with free AI

• What may and may not be entered into public (free) Gemini
• Anonymising company and counterparty data before passing it to AI (tax ID (NIP), surnames, amounts, personal data)
• Working with the personal data of employees and counterparties and the GDPR: what is and is not permitted
• Hallucinations when working with numbers and dates: how to recognise and verify them
• Ethics of using AI in internal communication, with counterparties and with public authorities
• Questions and answers

Methodological note

The training works with a single main tool only (the free version of Gemini), so that within 6 hours the group can genuinely master how to use it in the context of an accounting department’s work, rather than merely see a few demonstrations.

The data analysis exercises in module 4 are carried out on demonstration or anonymised data. No real company financial data is entered into the public, free AI model.

Trainer

Franciszek Kołpanowicz, entrepreneur and certified AI trainer. He has been published in Rzeczpospolita, among others. He has over 2,000 hours of work with generative AI and has shared his knowledge at public events, including the “Manager in the Age of AI” training organised by Rzeczpospolita and the UKEN and PASE conferences. His background in philosophy (Jagiellonian University) and his experience in IT quality control allow him to combine analytical thinking with asking the right questions. In training sessions he pays attention first and foremost to the participants, drawing on their experience and needs as the work unfolds.

Prerequisites

Good knowledge of using a computer and an office suite (email, word processor, spreadsheet). Prior experience with generative AI is not required.

Organisational requirements on the client’s side

• Before the training, each participant has a free Google account with access to Gemini (gemini.google.com) that has been set up and tested
• Login checked in advance on the device the participant will work on
• One workstation per person, a laptop for every participant, no sharing
• Stable Wi-Fi in the room (on-site format) or a stable connection for participants (remote format), sufficient for 10 people working at the same time
• Projector or large screen for the trainer’s presentation (on-site format) or a videoconferencing platform (remote format)
• A room suitable for workshop work, with tables and space for notes (on-site format)
• Flipchart or whiteboard with markers (on-site format)

The training works with a single tool, Gemini, so that within 6 hours participants can genuinely master how to use it in the context of their own work, rather than merely see a few demonstrations.

To be confirmed before the training

Licences and access:

• Does every participant have their OWN work account, or do they share one?
• Which Google Workspace plan does the organisation have?
• Gemini access to Google Workspace applications (Gmail, Drive, Docs, Sheets) switched on: a setting on the organisation’s administrator side

Participants’ technical readiness (CRITICAL):

• Every participant must arrive with a Google Workspace account already signed in on their own laptop: we do NOT set up accounts in the training room
• Please run a test before the training: each participant opens gemini.google.com on their work account and checks that they can ask a question and receive an answer
• If any participant does not have access to Gemini: please let us know in advance and we will discuss an alternative route

Infrastructure:

• Stable Wi-Fi

Key information

A training for accounting firms that want to move from asking AI individual questions to automating a specific process in the firm. The first day covers working with Claude on documents and in client contacts, and the rules for keeping client data secure. The second day shows how to build, in Claude Code, an automation for reconciling costs, that is, matching transactions from a bank statement against invoices from KSeF and from email.

During the training, participants will find out, among other things:

• How generative AI works and why every answer involving numbers and deadlines needs checking
• How to write prompts using the KREDO method for work on accounting documents
• How to read foreign invoices, letters from authorities and contracts with Claude
• How to check counterparties in public registers while working with AI
• Which client data an accounting office may pass to AI and how to describe this in the office’s AI policy
• How skills and connectors work in Claude Code
• How to build a cost reconciliation from the bank statement, KSeF and email
• How to control automation through approval, an audit trail and a confidence level
• How to plan and measure a pilot of a single process in the office

After the training, the office can move on to a pilot: over 4-6 weeks, the trainer and the office team jointly implement cost reconciliation in the office’s environment and measure its results. The training covers work with Claude and Claude Code, and the exercises are run on demonstration data prepared by the trainer.

Objective

The objective of the training is to familiarise participants with the use of Claude and Claude Code in the work of an accounting office, from working with documents and client correspondence to automating the reconciliation of costs against the bank statement and invoices from KSeF. The training covers prompting with the KREDO method, client data security rules, building skills and connectors, and human control of automation. The emphasis is on checking AI outputs and on preparing a pilot of a single process in the office.

Learning outcomes

After completing the training, participants:

• use Claude in their work on accounting documents and in correspondence with clients
• verify AI answers against the source document
• know the rules for passing client data to AI tools
• understand how automation in Claude Code is built on skills and connectors
• know the course of a cost reconciliation from the bank statement, KSeF and email
• have a plan for piloting a single process in the office

Programme

The programme comprises two training days and an implementation pilot carried out after the training.

Day 1: AI on the documents of an accounting office

Module 1: The basics of generative AI

• Sharing participants’ experience
• Key concepts related to how AI works (prompt, token, hallucination, context)
• How to act in the face of false content created by AI
• Comparison of the key AI models

Module 2: Prompting with the KREDO method in work on accounting documents

• Building a prompt using the KREDO method
• Projects in Claude as a place for the instructions and documents of a single task
• Extracting deadlines, amounts and the parties to a contract from a document
• Comparing AI answers with the source document
• The most common AI errors with numbers, dates and regulations

Module 3: Claude in handling documents and clients of the office

• Reading foreign invoices and preparing data for posting
• Summarising a letter from the tax office or ZUS and drafting a reply
• Verifying a counterparty in public registers
• NBP exchange rate for converting a foreign-currency invoice
• Answers to typical questions from the office’s clients
• Reminders about missing documents

Module 4: Security, professional secrecy and the office’s AI policy

• Which client data may be passed to AI and which may not
• Personal account versus company account in an AI tool
• GDPR and data processing agreements for AI tools
• The obligation to ensure AI competence arising from the AI Act
• The accounting office’s AI policy on one page
• Responsibility for the result: AI prepares, the accountant approves

Day 2: Automating cost reconciliation

Module 5: From conversation to agent: Claude Code in an accounting office

• How Claude Code differs from a conversation with AI
• Skills as the office’s recorded procedures
• Connectors to KSeF, the accounting software and public registers
• The office repository as a single place for procedures, instructions and change history
• The test environment as the starting point of every automation

Module 6: Cost reconciliation step by step

• The bank statement as the starting point of the reconciliation
• Domestic invoices from KSeF and foreign invoices from email
• Matching transactions to invoices with a justification and a confidence level
• A summary of transactions with and without a document
• Organising documents in the client’s folder
• Running the reconciliation on a schedule

Module 7: Controlling automation

• Approval by the accountant before every entry
• An audit trail and a confidence level for every match
• Why amounts and taxes are calculated by software, not by a language model
• Access rights and passwords in automation
• What to do when the accounting software has no API

Module 8: The pilot plan for the office

• Choosing the process and the scope of data for the pilot
• Pilot metrics
• Roles in the pilot on the office’s side and the trainer’s side
• Timetable and criteria for the decision to extend
• Questions and answers

Implementation pilot after the training

Stage 1: Analysis of the process in the office

• How cost reconciliation runs in the office before implementation
• Data sources in the office
• Baseline measurement of working time
• Selecting the office’s clients covered by the pilot

Stage 2: Configuration and tuning

• Configuring skills and connectors in the office’s environment
• Matching rules adapted to the office’s clients
• Tests on data from a closed period
• Instructions for the office team

Stage 3: Parallel running and measurement

• Automated reconciliation running alongside the existing work
• Analysis of discrepancies and rule corrections
• Measuring working time and the match rate
• Weekly meetings with the office team

Stage 4: Report and decision

• Report with results before and after implementation
• Recommendation to extend, adjust or end the pilot
• Further office processes to automate

Trainer

Franciszek Kołpanowicz: entrepreneur and certified AI trainer who helps leaders and teams become more effective through intelligent technologies. He has been published in Rzeczpospolita, among other places. He has more than 2,000 hours of work with generative AI behind him and shares his knowledge at public events, including the “Manager in the Age of AI” training organised by Rzeczpospolita, as well as at the UKEN and PASE conferences.

His background in philosophy (Jagiellonian University) and his experience in IT quality control allow him to combine analytical thinking with asking the right questions. His own practical journey with AI showed him how this technology can become a real lever for growth.

In 2026 he delivered AI training for accounting and finance departments. He keeps his own accounts using automations built in Claude Code, including the posting of costs from KSeF and the monthly reconciliation of documents.

Drawing on hands-on experience, he shows leaders and their teams how to use AI not only to work more efficiently but also to think more clearly and make faster business decisions. His AI training has, among other results, cut the time needed for routine analysis of multi-page legal documentation by 80%.

In his training sessions he pays particular attention to the participants themselves, drawing on their experience and needs as the work progresses. As a result, his sessions are not repetitive routine but an enriching and individual experience.

His mission is to make artificial intelligence a practical tool for development: business, intellectual and cultural.

Prerequisites

A good knowledge of using a computer and an office suite. Previous experience with generative AI and programming skills are not required.

By Day 2, each participant has Git for Windows installed on their computer. The trainer sends the installation instructions before the training.

Organisational requirements on the client’s side

• One workstation per person, a laptop for every participant, no sharing
• Stable Wi-Fi in the room (on-site format) or a stable connection for participants (remote format)
• Projector or large screen for the trainer’s presentation (on-site format) or a videoconferencing platform (remote format)
• A room suitable for workshop work, with tables and space for notes (on-site format)
• Flipchart or whiteboard with markers (on-site format)

Each day of the training works with one tool from the same family, the first with Claude, the second with Claude Code, so that participants have time to genuinely master using it in the context of the office’s work, rather than merely see a few demonstrations.

For the pilot phase:

• A person in the office responsible for the pilot and a person who approves the results
• A personal data processing agreement covering the implementation work
• Access to the data needed in the pilot, granted according to the arrangements from stage 1
• The office team’s time for weekly meetings

To be confirmed before the training

Licences and access:

• Does every participant have an active Claude Pro subscription?
• Does every participant have their own Claude account, or do they share one?
• Does the organisation restrict the use of Claude connectors to Microsoft 365? (some organisations block OAuth to external tools)

Participants’ technical readiness:

• Every participant arrives with the Claude Desktop application installed and an account already signed in: we do not set up accounts in the training room
• Test one week before the training: each participant opens Claude Desktop and verifies that they can ask a question and receive an answer
• Test one week before the training: an IT specialist checks on one of the accounts whether a connector to Office suite tools can be set up
• If any participant does not have access: advance notice, and we will discuss an alternative way of working

Infrastructure:

• Stable Wi-Fi tested under load