A practical programme for getting control of how your company uses artificial intelligence: protect data, set common rules and train your people. Three paths: for employees who have never tried AI, for departments, and for the company's AI coordinator. Learn at your own pace, free of charge.
Learning time depends on you. Implementation time depends on your company: colleagues fill in questionnaires, management approves documents. That is why every module shows both.
Review all materials and templates without implementing anything yet. Suits a manager deciding whether to start.
Start the next module once your company has completed the tasks of the previous one. Nobody rushes and nobody falls behind.
All durations are preliminary and will be refined once the materials are final.
For people who have not tried AI yet, or have tried it hesitantly. The “Safe AI User” level (A0–A2, about 1.5 hours) is for every employee, even those who will not use AI. The “AI User” level (A0–A5 and the final test, about 3 hours of material and 2 hours of practice) is for those who will use AI at work.
“Safe AI User”: A0, A1 and A2 tests passed. “AI User”: all A0–A5 tests and the final test passed. A test counts as passed at 80% or more. Enter your details, download the certificate and give it to your company's AI coordinator.
Test results are stored only in your browser. The certificate confirms self-paced training and is not a state-recognised qualification.
After path A. Four general modules fit any company: each has ready-made prompt templates, data-level rules, checks and a test. Three sector modules cover manufacturing, construction and logistics; we refine them together with companies from those sectors.
For the company's AI coordinator and management. Each module ends with a document or tool that you put in place in your company.
EU law and the law of the country you selected, relevant to an AI coordinator: what already applies, what is still under discussion, and what your company should do now. Status is changed only on the basis of official sources; media and law-firm reviews are listed as additional information.
Need a lawyer to help apply these requirements? Our partners:
Cards are reviewed weekly. A newsletter will be available once the platform runs on its own domain. Cards are for information only and are not legal advice.
Drafts for your company to adapt. Fill in the blanks marked with underscores and the national-law placeholders in square brackets. Your lawyer must review the documents before they are approved. In many countries internal company documents must be in the national language: use these English versions as a reference or for international companies.
Our partners can adapt these templates for your company:
Department prompt templates are in the path B modules and can be copied directly from the lessons.
Short screen recordings with voice-over: where to click, what to upload and how to word a prompt. Links will appear here once the videos are published.
“AI Coordinator” is a free self-learning platform for companies that want to use artificial intelligence safely and in an orderly way: train employees, set rules, protect data and meet the AI literacy obligation.
The platform is operated by MB Sol Maris, company code 306818822, Klaipėda, Lithuania.
We prepare the training materials, templates and law cards independently. Partners and sponsors have no influence on the platform's content. Where an expert has reviewed specific material, this is stated next to that material.
Users adapting the AI policy and other documents for their company often need a lawyer. Law firms can become platform partners: their details are shown next to the law cards, the templates and in the footer, according to the language version and the visitor's selected country.
To discuss a partnership, write to info@aikoordinator.eu
Yes. All training materials, tests, certificates and templates are free and may be used for internal company training.
No. Test results are stored only in your browser and the certificate is generated on your computer. You do not send us any data.
No. It confirms that the employee completed self-paced training and passed the self-check tests. For a company it is a simple way to document AI literacy measures, but it is not a state-recognised qualification.
The templates are drafts. Before approving them in your company, adapt them and consult a lawyer, especially on employment-law provisions and the role of employee representatives in your country.
Law cards are reviewed weekly against official sources. Each card shows its last update date.
EU-level law applies to all EU countries. National information is currently available for Lithuania; other countries will be added together with local legal partners.
PRIVACY POLICY Last updated: __ ____________ 20__ 1. Who processes the data 1.1. The data controller is MB Sol Maris, company code 306818822, registered office ____________________, email info@aikoordinator.eu (“we”). 1.2. Our activities do not require us to appoint a data protection officer. For any questions about personal data, contact us at the email address in point 1.1. 2. Basic principle 2.1. The platform is designed to process as little personal data as possible. No registration is required, there are no accounts, and we use no advertising or tracking cookies. 3. Data that stays on your device 3.1. Test results are stored only in your browser's local storage (localStorage). They are not sent to us. This storage is necessary for the function you yourself have chosen – showing progress and generating a certificate – so no consent is required for it. You can delete the data at any time by clearing your browser data for this website. 3.2. The name and company you enter for the certificate are used only to generate the certificate in your browser. They are not sent to us and we do not store them. 3.3. Your choice of language and country is stored in your browser so that the website opens the same way next time. 4. Technical data for operating and securing the website 4.1. The website and email are hosted by UAB Interneto vizija (Lithuania). When you open the website, the server processes technical request data: IP address, browser and device type, time of the request and the page opened. 4.2. Purposes: to deliver the website and protect it against attacks and abuse. 4.3. Legal basis: our legitimate interest in ensuring the operation and security of the website (Article 6(1)(f) GDPR). 4.4. UAB Interneto vizija processes this data as our processor under the service agreement. Retention period: ____________________ (according to the hosting provider's server log retention). 5. Partner link statistics 5.1. The platform may show partner links labelled “Partner”. When you click such a link, we count the click and you are immediately redirected to the partner's website. 5.2. Only aggregated totals are counted: date, partner, place on the website and the country you selected. We do not store your IP address or any other identifying data with the click. We do not link the click to you and do not pass personal data to partners. 5.3. Once you are on the partner's website, the partner processes your data under its own privacy policy. 6. Newsletter 6.1. When you subscribe to news on legal changes, you provide your email address. We process it on the basis of your consent (Article 6(1)(a) GDPR), solely to send the news. 6.2. The subscription is confirmed by clicking a link in a confirmation email. You may withdraw consent at any time by clicking the unsubscribe link in any email or by writing to us. 6.3. We keep your email address while you are subscribed. After you withdraw consent, the address is deleted, except for the record of consent and withdrawal, which we keep for ____________________ to be able to demonstrate compliance. 6.4. Newsletters are sent using ____________________; this provider acts as our processor. 7. Correspondence with us 7.1. When you email us, you provide your name, email address and the content of your message. We process this data to answer your enquiry and, for partnerships, to conclude and perform a contract (Article 6(1)(b) and (f) GDPR). 7.2. We keep correspondence for ____________________ after the last message, and partnership contract documents for as long as required by law. 8. Recipients of data 8.1. We do not sell data or pass it to partners. 8.2. Data may be processed only by our service providers acting as processors: the hosting and email provider UAB Interneto vizija and the newsletter provider ____________________. 8.3. We may disclose data to public authorities where required by law. 9. Transfers outside the European Economic Area 9.1. Website and email data are processed in the European Union; we do not transfer them outside the European Economic Area. If the chosen newsletter provider processes data outside the EEA, this point will say so, together with the legal basis for the transfer. 10. Your rights 10.1. You have the right to access your data; to have it rectified; to have it erased; to restrict its processing; to object to processing based on our legitimate interest; to data portability; and to withdraw consent at any time (Articles 15–22 GDPR). 10.2. Contact us at the email address in point 1.1. We will reply within one month of receiving your request; in complex cases this period may be extended, and we will let you know. 10.3. Much of the data (test results, certificate details, language and country choice) is stored only on your device, so we do not have it and cannot provide or delete it. You control it yourself. 10.4. If you believe your data is being processed unlawfully, you have the right to lodge a complaint with the State Data Protection Inspectorate of Lithuania (vdai.lrv.lt) or with the data protection authority of your country of residence. 11. Security 11.1. The website is accessible only over a secure connection (HTTPS). We apply organisational and technical measures to protect data against unauthorised access and process only as much data as is necessary. 12. Children 12.1. The platform is intended for companies and their employees. We do not aim to process children's data. 13. Changes to this policy 13.1. This policy may be updated. The current version is always published on this website with the date of the last update. Newsletter subscribers will be informed of significant changes by email.
TERMS OF USE 1. The platform is operated by MB Sol Maris, company code 306818822 (the “Operator”). By using the platform you agree to these terms. 2. Informational nature 2.1. The content of the platform is introductory and informational. It is not legal, tax or other professional advice. 2.2. The templates are drafts that users must adapt to their own situation. Before approving documents and introducing tools, consult a lawyer and a data protection officer. 2.3. Laws and AI tool settings change. The Operator endeavours to keep information up to date but does not guarantee that it is always complete and current. Check primary sources. 3. Use of content 3.1. The content of the platform belongs to the Operator unless stated otherwise. 3.2. Companies and organisations may use, copy and modify the training materials and templates free of charge for their internal purposes, including employee training. 3.3. Without the Operator's written consent, it is prohibited to sell the content, republish it as your own product or offer it as a paid service. 4. Links and partners 4.1. The platform may contain links to third-party websites, including partners. The Operator is not responsible for the content of those websites or the services provided by partners. 4.2. Partners have no influence on the platform's content. 5. Liability The Operator is not liable for damage arising from use of the platform's content without professional advice, except where liability cannot be limited by law. 6. Governing law These terms are governed by the law of the Republic of Lithuania. 7. Changes The terms may be amended. The current version is published on this website. Last updated: __ ____________ 20__
This material is introductory and intended to help you get acquainted with the topic. It is not legal, tax or other professional advice. The templates are drafts that must be adapted to your company. Before approving documents and introducing tools, agree the final decisions with a lawyer and your data protection officer. Laws and tool settings change, so always check primary sources.
In this lesson you will learn what an AI chat tool is, why there is no need to be afraid of it, and you will do your first three exercises yourself.
It is a program that creates text from a request you type (called a prompt): an email, a summary, a translation, a list or an explanation. It does not understand the world the way a person does and it does not check facts: it produces text that looks likely. That makes AI a fast assistant for drafts, but not a source of knowledge you can trust blindly.
The exercises use made-up data. Copy the prompt into the AI tool and see what you get.
1. A more polite, shorter email
Rewrite this email politely and briefly, in no more than 5 sentences: “Hi, the goods are late again for the third time, this does not work for us, please do something, our customers are angry.”
What to notice: that no commitments you did not want to make have crept in, and that the tone suits your company.
2. A meeting agenda
Draft a 30-minute meeting agenda. Topic: the new procedure for receiving goods in the warehouse. Participants: warehouse manager, two warehouse workers, purchasing specialist. Leave 5 minutes for questions at the end.
What to notice: whether the times add up and the topics fit the participants.
3. Explaining to a new colleague
In 5 simple sentences, explain what a VAT invoice is to someone on their first day in an office job.
What to notice: whether the explanation is correct. If in doubt, ask your accountant. That is your first check.
In this lesson you will learn why fluent AI text can still be wrong, and where to look for mistakes first.
An AI language model was trained on a huge amount of text and learned which words usually follow each other. When you ask a question, it produces the most likely continuation. That is very good for language, structure and style, but it is not fact-checking.
Most models write less well in smaller languages than in English: you will see grammar errors, awkward literal translations and imprecise terms. Always read a text in your own language before sending it.
The most important lesson for every employee: what you may put into AI tools, what you may not, and what to do if you make a mistake.
| Level | What it covers | What you may do |
|---|---|---|
| Red | Personal identification numbers, health data, employee salaries and personnel files, customer lists with contact details, bank and payment data, passwords, trade secrets, terms of unsigned contracts, litigation material | Never upload to any public AI tool. Only in a local offline model or in an environment approved by the company, if the policy allows it. |
| Yellow | Internal procedures, anonymised document extracts, summaries without customer names, internal reports without personal data | Only in an approved tool, logged in with a company account. Anonymise before uploading. |
| Green | Public information, general texts, ideas, translations of public texts, templates without specific data | You may use approved tools. Still check the result. |
If in doubt, treat the data as red and ask the AI coordinator.
AI can be built into places you do not expect: browser extensions, phone apps, transcription bots in video calls, free PDF and translation tools. Do not install them or use them for work documents without the AI coordinator's approval.
In this lesson you will learn to turn a short request into a clear task, so that the first AI answer is usable and you do not have to correct it ten times.
Weak prompt
Write an email to a customer about a delay.
AI does not know the customer, the reason or the new date, so it will make them up.
Good prompt (made-up data)
Draft an email to a customer. Context: order No. 1458 (20 office chairs) is late because the supplier delivered parts late. New delivery date: 12 December. The customer is important to us and orders regularly. Audience: the customer's purchasing manager. Format: no more than 6 sentences, businesslike and apologetic tone. Constraints: do not promise discounts or compensation, do not invent other facts. At the end, offer a call if they want to discuss.
Turn the request “Write up the meeting minutes” into a complete prompt. Then check it against this list:
AI-generated text must be checked before use. In this lesson you will learn to do that quickly and consistently.
Source material (made-up contract extract)
Goods are delivered within 14 calendar days from the date of payment. For each day of delay the supplier pays a penalty of 0.05% of the contract value, but no more than 10% of the contract value. Contract value: EUR 4,000.
AI summary
Delivery within 14 working days from signing the contract. Penalty: 0.5% per day, with no limit. A 10-day delay would mean a penalty of EUR 200.
1. “Working days” instead of “calendar days” – the deadline becomes longer. 2. “From signing the contract” instead of “from the date of payment” – the start of the deadline changes. 3. “0.5%” instead of “0.05%” – ten times higher. 4. The cap “no more than 10% of the contract value” is dropped and replaced with “no limit”. 5. The calculation uses the wrong rate. Correct: EUR 4,000 × 0.05% × 10 days = EUR 20. Conclusion: the summary sounds convincing, but all five important facts are distorted. That is why a summary is always compared with the original.
In the last lesson you will connect what you have learned with your own work: choose one activity and honestly check whether AI really makes it easier.
List about 10 tasks you do every week. For each one, note:
Best candidates for a first trial: frequent tasks that take a lot of time, with green or yellow data and a low cost of mistakes. For example: answers to standard questions, meeting agendas, drafts of internal announcements, shortening long texts.
Collecting material, structure, draft, checking, final approval. AI usually helps most with structure and drafting. Checking and approval remain yours.
Copy and fill in:
Activity: ________
Data level: ________
Tool: ________ (approved only)
Prompt template: ________
Check criteria: ________
Time before: ____ min per week
Time with AI (including checks): ____ min per week
Conclusion: continue / change the prompt / drop it
Show your completed plan to your manager or the AI coordinator. Good plans become department templates (path B).
The final test covers all path A lessons. Pass it (80% or more) together with the A0–A5 tests to receive the “AI User” certificate.
In sales, AI helps most in two places: quickly understanding what the customer is really asking for, and preparing the text of an offer. Prices and totals are always calculated by a person or a spreadsheet.
1. Analysing an incoming enquiry
You are an assistant to our sales manager. Analyse the customer enquiry below and give me a table with three parts: 1) what the customer is asking for (products, quantities, deadlines); 2) what information is missing to prepare an offer; 3) questions to ask the customer. Do not invent quantities, deadlines or other data that are not in the enquiry. Enquiry: [PASTE THE ANONYMISED ENQUIRY]
2. Cover letter for a commercial offer
Write a cover letter for a commercial offer. Products and prices excluding VAT: [LIST FROM PRICE LIST]. Delivery time: [DEADLINE]. Payment terms: [TERMS]. Addressee: [JOB TITLE]. Format: a 2-sentence introduction, a table “product, quantity, price, total”, terms as bullet points, and an invitation to get in touch at the end. Do not change the prices, do not calculate discounts and do not add terms that are not in the data provided.
3. Follow-up after an offer
Write a short follow-up email, no more than 4 sentences, to a customer to whom we sent an offer on [TOPIC] [NUMBER] days ago. Tone: polite, no pressure. Offer to answer questions or to meet.
Take your three most recent customer enquiries, anonymise them and try template 1. Compare the AI's list of questions with what you would have asked yourself.
In finance and administration, AI helps organise contract terms, extract data from non-standard invoices and draft administrative letters. This is where most sensitive data sits, so it matters greatly where and how you work.
1. Contract terms checklist
Analyse the anonymised contract extract below. Give me a table with the columns: clause, what it means in plain words, possible risk for our company, question for our lawyer. Pay particular attention to penalties and their caps, deadlines and when they start, unilateral termination rights, limitation of liability and automatic renewal. Do not quote anything that is not in the text and do not assess whether the contract is lawful. Extract: [PASTE]
2. Invoice data into a table (approved environment or local model only)
From the invoice text provided, extract into a table: invoice number, date, supplier, goods or service, quantity, price excluding VAT, VAT rate, VAT amount, total, payment due date. If a field is missing, write “missing”. Do not calculate missing amounts. Invoice: [PASTE]
3. Payment reminder
Write a polite reminder to a company about an overdue invoice. Invoice No. [NUMBER], amount EUR [AMOUNT], due date was [DATE]. This is the first reminder. Tone: businesslike, no threats. No more than 5 sentences.
Take one standard supply contract that has already been signed, anonymise one section and try template 1. Compare the result with what you already know about that section.
In office work, AI saves time wherever there is a lot of reading and writing: finding answers in internal procedures, preparing minutes and turning complex instructions into clear ones.
1. Answer from an internal procedure
Answer the question using only the document below. Give the number of the clause you rely on. If the document does not contain the answer, say so and do not guess. Question: [QUESTION]. Document: [PASTE]
2. Minutes and task table
From the meeting notes, prepare minutes: 1) participants; 2) topics discussed; 3) decisions taken; 4) a task table with the columns “task, owner, deadline”. If the notes do not name an owner or a deadline, write “not specified”. Do not add anything of your own. Notes: [PASTE]
3. Instructions for a new employee
Rewrite these instructions for a new employee: short numbered steps, simple words, no more than 10 steps. Do not change requirements, deadlines or responsible persons. Instructions: [PASTE]
Take the notes of a recent meeting and try template 2. Send the minutes to the participants and ask whether anything is missing.
AI can quickly scan a table, find the biggest changes and draft a report. But analytics is full of numbers, and numbers are where AI makes mistakes most often.
1. Analysing a table
Below is a monthly sales summary by product group, without customer names. Identify: 1) the three biggest changes compared with last month, in euros and percent; 2) unusual values; 3) questions worth asking the department. Show how you calculated each percentage. Do not invent reasons for the changes: present them as questions. Data: [PASTE THE TABLE]
2. Report to management
Based on these verified findings, write a one-page draft report to management: the main conclusion in 2 sentences, the 3–5 most important figures, risks, proposed decisions. Use only the figures provided and do not add new ones. Findings: [PASTE]
3. A spreadsheet formula
Write an Excel formula that [DESCRIPTION, for example: calculates each product's sales change in percent between columns B and C]. Explain what each part of the formula does.
Take last month's summary without names and try template 1. Recalculate three of the AI's changes yourself in a spreadsheet. Did they all match?
In manufacturing, AI suits text work around the core activity: instructions, defect descriptions, comparing supplier specifications. Technical decisions and safety requirements are always approved by the responsible specialist.
1. Draft occupational safety instruction
Draft an occupational safety instruction for this job: [JOB, for example, laser cutting of sheet metal]. Structure: hazards, personal protective equipment, actions before, during and after work, actions in emergencies. Do not cite specific legal articles or numeric limits: where they are needed, write [CHECK]. The instruction will be approved by the safety specialist.
2. Defect description and possible causes
From the defect description provided, prepare: 1) a defect description in standard form (what, where, when found, how many units); 2) up to 5 possible causes; 3) for each cause, how to check it. Present the causes as hypotheses, not conclusions. Description: [PASTE WITHOUT CUSTOMER DATA]
3. Comparing supplier specifications
Compare the supplier specifications provided in a table with the columns: parameter, supplier A, supplier B, supplier C, meets our requirement (yes, no, not stated). Our requirements: [REQUIREMENTS]. Do not fill in missing values and do not convert units without explanation. Specifications: [PASTE]
Take a recently recorded defect and try template 2. Compare the AI's hypotheses with what the cause actually turned out to be.
In construction and design, AI helps manage the flow of documents: procurement requirements, site diary summaries, specifications. It does not perform engineering calculations and does not take on the responsibility of a certified specialist.
1. Procurement requirements checklist
From the procurement document provided, extract a table with the columns: requirement, document clause, document to be submitted, deadline, owner in our company (leave blank). Give the clause number for each requirement. Do not add anything of your own. Document: [PASTE]
2. Weekly site report to the client
From the anonymised site diary entries provided, prepare a weekly report to the client: work completed, deviations from the schedule, issues, plan for next week. Do not invent quantities or percentages: use only the figures in the entries. Entries: [PASTE]
3. Extracting a drawing specification (local model only)
From the specification table of the drawing provided, extract: item, name, material, quantity, dimensions. If a value is illegible, write “illegible”. Do not calculate or add anything.
Take a completed public procurement you took part in and try template 1. Did AI find all the requirements you found back then?
Logistics and warehousing involve many repetitive documents and emails. AI helps quickly extract order data, describe incidents and hand over information to the next shift.
1. Transport enquiry data
From the customer's email provided, extract a table: loading place, unloading place, requested dates, cargo type, weight, volume, number of pallets, special requirements (dangerous goods, temperature), missing information. If a value is missing, write “not specified”. Email: [PASTE WITHOUT CONTACT DETAILS]
2. Cargo damage report
Prepare a cargo damage report: what was found, when and where, number of damaged units, list of photos taken, who has been informed, next steps. Tone: factual, without assigning blame. Use only the facts provided: [FACTS]
3. Shift handover note
From these shift notes, prepare a handover note for the next shift: work completed, work not completed, issues, what must be done first. No more than 10 lines. Add nothing. Notes: [PASTE]
Take your five most recent transport enquiries, anonymise them and try template 1. How many times did AI correctly flag the missing information?
Shadow AI is any AI tool employees use without the company's knowledge or approval. It is usually not malicious: people simply want to write an email or translate a contract faster.
Beyond personal data, trade secrets are at risk too: pricing, supplier terms, unsigned contracts. The data separation standard (three colours) is in lesson A2: transfer it into the company's AI policy.
Article 4 of the EU AI Act requires companies that use AI systems to take measures to develop their staff's AI literacy. The obligation has applied since 2 February 2025; it was clarified by Regulation (EU) 2026/1744, in force since 27 July 2026. A specific level of literacy need not be guaranteed, but measures must be taken. Path A and the “Safe AI User” and “AI User” certificates are a simple, documented measure of this kind.
Hand out the questionnaire (in the templates section) to colleagues. Once you have the answers, build a table: tool, who uses it, for what, what data is uploaded, who pays. Then:
In the management order, set a deadline by which all employees complete the “Safe AI User” level, and collect the certificates. This gives you evidence that the company has taken AI literacy measures.
Red-level documents (contracts, invoices, personnel files, drawings) need an environment that sends nothing to the internet. In this module you will learn what can realistically run on an office computer and how to review the company's AI subscriptions.
| Computer | What can run | What to expect |
|---|---|---|
| 8 GB RAM | Only small models (1–4 billion parameters) | Slow, weak in smaller languages. Suitable for trials only. |
| 16 GB RAM | Medium models (about 8–12 billion parameters) | Suitable for summaries and data extraction, but slower than cloud tools. |
| 16 GB or more of graphics memory, or 32 GB or more of unified memory (e.g. Apple M series) | Larger models | Comfortable everyday work with documents. |
Practical tip: instead of installing a model for every employee, designate one more powerful computer for processing sensitive documents.
Download applications only from their official websites and check the licence terms for commercial use.
Open-weights model families are updated constantly, so choose the latest version that fits your computer. The most widely used are Gemma (Google), Qwen (Alibaba), gpt-oss (OpenAI) and Mistral (France). Check each model's licence before commercial use.
Quality in smaller languages varies a lot between models. Before choosing, test 2–3 models on the same samples of your own documents.
When every employee writes to AI in their own way, results are uneven and hard to check. Standardised prompts and tabular forms fix this: consistent quality, easy checks and quick onboarding of new colleagues.
For each approved template, fill in a card (in the templates section): name, purpose, data level, permitted tool, template text, check criteria, owner, version and date. Keep the library in one place all employees can see, for example a shared company folder.
A template that nobody uses, or whose results constantly have to be redone, must be revised or removed from the library.
In the final module you will present the new rules to the team, set up quality control and prepare an annual AI implementation and maintenance plan for management.
The full script is in the templates section. The basic structure:
| Frequency | What the coordinator does |
|---|---|
| Monthly | Reviews incidents and the error log, follows legal and tool changes. |
| Quarterly | Reviews subscription use and costs, updates templates, assesses time saved. |
| Annually | Reviews the AI policy, updates training, audits AI tools, prepares the new annual plan. |
For each template: (time without AI − time with AI, including checks) × number of times per month. Multiply the hours by the average hourly cost and subtract subscription costs. That is an honest figure you can present to management.