AI training for officials, managers, and IT and DPOs
Three online courses for the roles an AI rollout in a public office depends on. Each lasts 3 hours, shows five AI use cases live in SAIE and ends with a document for the office and a named, numbered certificate that opens the LIDER:AI Academy
One office, three roles, one week: first the manager decides, then IT and the DPO prepare the conditions, and finally officials work with the designated tool
10 Manager - Implementing AI in a public office without chaos: from the first decision to evidence for an inspection
Role-based course • 1 × 3 hours
Who it is for
the head of the municipality, mayor or city mayor as head of the institution, the secretary, the treasurer and heads of departments: the people who issue the order, allocate the budget and answer to inspections. Managers use generative AI more often than other staff (53%), while only 8% of cities have an AI strategy (NASK 2026, Association of Polish Cities 2025)
From 28 October 2026 the Commission for the Development and Security of Artificial Intelligence (KRiBSI) can inspect public offices and impose fines, and under the amended National Cybersecurity System Act (KSC) the head of the institution is responsible for its cybersecurity. The course answers the question a manager comes with: where to start, how much it costs and how to show an inspection that the office has AI under control. It guides you through implementation according to the CDF methodology, from the Initial Assessment to scaling, shows how to carry out the change without frustrating the team, what functions the system must have and how to keep costs in check from day one. On selected dates the legal part is led by guest trainer Aleksandra Albera, a lawyer specialising in artificial intelligence law
Course programme
Part A
A. The manager's risks and what an inspection will check
The Polish AI Systems Act: inspections and fines from 28.10.2026; the AI Act after the July 2026 amendments: AI literacy, transparency from 2.08.2026, high-risk systems from 2.12.2027
GDPR, tax secrecy and KSC: the responsibility of the head of the institution; three levels of responsibility: supplier, office and employee; warning signs when AI is used in public procurement
Institutional shadow AI: a ban does not change who is responsible, it only changes whether the office sees what is happening
Result:A list of documents and evidence the office will show an inspection
Part B
B. Implementation according to CDF and change without frustration
Initial Assessment, overarching goal, compliance before building, oversight, people, a 2-4 week pilot and scaling in waves
An announced temporary drop in productivity, AI leaders in departments and a before-and-after work balance
Nine documents: from the order and the tool register to the incident procedure and the security annex to contracts
Result:The overarching goal and the first use case with an owner and metrics
Part C
C. System functions, five use cases and costs
The manager's ten questions for any supplier, including us, and the answers shown live
Five use cases from the manager's perspective: metric, cost, evidence and real human oversight
Costs under control: department budgets, limits, forecasting and plan recommendation, buying in stages, a shared environment for several institutions
11 IT and DPO - Gateway, register, evidence: the conditions for safe AI in a public office
Role-based course • 1 × 3 hours
Who it is for
the office's IT staff and companies providing IT services to the institution, administrators, data protection officers and information security officers. 60% of technical staff in public offices use generative AI, and 61% of officials see AI as a threat to the privacy of the office's clients (NASK, 2026)
The IT specialist and the DPO come with the question: what must the system have for me to sign off on it. The course goes through the threats the office already faces today: personal data leaking to public models, prompt injection, agents with excessive permissions, runaway costs and dependence on a supplier, as well as the obligations under GDPR, KSC and the AI Act. It then turns them into twenty security requirements and shows the PROXY:AI gateway live: the path of a single request in eight steps, data masking, a register that cannot be altered and a model running in a network without internet access
Course programme
Part A
A. Threats and obligations
Personal data leaking to public models, passwords, keys and logs pasted into a chat, prompt injection and bypassing safeguards, agents with excessive permissions, runaway costs
GDPR (Art. 25, 28, 32 and 33-35) and KSC: the AI supplier as a processor, the office's status, reporting breaches and incidents, the information security management system
AI Act: deployer obligations today and from 2.12.2027 (Art. 26)
Result:A map of threats and obligations for the office's AI environment
Part B
B. Twenty requirements and documentation
Identity, data, models, attacks, agents, trail, costs and deployment option
Documents kept by IT and the DPO: the AI systems register, DPIA, the agent register, the incident procedure
Data classification: what may be sent to which model, trade secrets in bids, pseudonymisation versus anonymisation
Result:Security requirements for the office's AI environment
Part C
C. PROXY:AI and five use cases live
The path of a single request in eight steps: from authentication to an entry in an immutable register
Testing a rule in trial mode and installing a model in a network without internet access
Five use cases from the inspection side and an evidence pack for an inspection in one click
Result:A draft DPIA for one use case and a template entry for the AI systems register
09 Official - AI in my work: safely, lawfully and better than in a private chat
Role-based course • 1 × 3 hours
Who it is for
specialist and citizen service staff: case officers, inspectors, secretariats and council offices who already use AI or are about to start; no technical background required. In local government offices half of the staff have already used generative AI at work, often on private accounts (NASK, 2026)
An official comes with the question: may I use AI, and if so, how, so that I am not responsible for the machine's mistake. The course starts with the concerns that hold back work with AI: an invented legal basis, a resident's data in a private chat and responsibility for the letter. It then gives a five-step working method and the limits that follow from the law, and finally shows live how the same task looks in a private chat and in TWIN:DESK, on five cases from the life of an office: from replying to a resident's letter to assigning incoming correspondence
Course programme
Part A
A. The official's concerns and the answers to them
Why AI sounds confident when it is wrong: an invented legal basis, an outdated provision, an omission
What happens to data entered into a private chat and who is responsible for a letter prepared with AI
Double work and worries about one's skills: what changes in the work and what stays with the human
Result:A shared list of the group's concerns with answers based on the law and examples
Part B
B. Working method and limits from the law
Five steps of safe work: task, prompt, source, verification, decision and trail; five questions before pasting a document; demo: the same question asked in three ways gives three different conclusions
What not to do: residents' data in an unapproved tool (GDPR), tax secrecy, trade secrets in bids, AI deciding a case, hiding AI from residents (Art. 50 AI Act), prohibited practices (Art. 5 AI Act)
Exercise: twelve typical office tasks, including public procurement tasks, in the three fields of the map
Result:A task map for your own position: I use / I check the conditions / I do not use
Part C
C. A private chat versus TWIN:DESK live
The same task in a private chat and in the office's digital AI desk: a comparison in nine points
Five use cases: a reply to a resident's letter, a search engine for orders and resolutions, minutes of a council session, the treasurer's statements, assigning incoming correspondence
An answer with a quotation, the version of a provision in force, masking the PESEL number, a calculator instead of the model's “memory”, an agent plan for approval
Result:Safe AI Work Card: task map, five steps, five questions before pasting, a 10-question verification checklist and a “what I don't do” list
For more than 10 years he has been building IT solutions for the public sector, including reporting systems for supervisory institutions. He designs and implements IT systems: he turns the needs of public offices into concrete solutions, configures, tests and documents them. He runs courses and workshops on the CDF methodology, ISO/IEC 42001 and the AI Act
In the courses:how to move from a problem to a controlled pilot, how to plan an AI rollout in a public office and how to prepare documents that will stand up to an audit
TWIN:DESK development and implementation specialist
TWIN:DESKTWIN:DESK implementationsTechnical aspects of AI systemsBig Data
He develops and implements TWIN:DESK, the digital AI desk of the SAIE ecosystem. He knows the technical side of building AI systems and working with large data sets
In the courses:how the five use cases shown live work in TWIN:DESK on the technical side and how to roll them out in a public office
Lawyer and ISO/IEC 42001 lead auditor, compliance at allclouds.pl
AI ActPolish AI Systems ActISO/IEC 42001Liability for AI decisions
An AI lawyer: she builds AI systems, audits them (ISO/IEC 42001) and makes sure they comply with the law. At allclouds.pl she is responsible for compliance, and she is a law PhD candidate. She explains who is liable when AI makes a mistake, and what a public office cannot shift to the supplier in a contract
In the courses:what the Commission for the Development and Security of Artificial Intelligence will check, who is liable for an AI error and how the office shows it has AI under control
Select the course, date and number of participants in the order form. The quote takes account of places already used by your institution.
FAQ
How do we register for a course and how long does the whole process take?
You complete the form on the page: requester details, institution name and the selected courses with dates and number of participants. After submitting, you receive a copy of the registration and a proforma invoice (for paid training) at the e-mail address provided. Three days before the first day of training we send participants the link to the course
Can we change the date or a participant after the registration is confirmed?
Yes, free of charge. A participant can be replaced by another person from the same institution up to the start of the training. Moving to another date from the calendar is possible up to 5 business days before the training, provided there are free places on the new date. Just send an e-mail to office@allclouds.pl
Until when can we cancel without charge?
Cancellation is possible up to 5 business days before the first day of training; after that we offer a change to another date or another participant instead of charging a fee
How do role-based courses 09-11 differ from workshops 01-08?
Workshops 01-08 teach skills independent of any tool, in three blocks of 80 minutes. Role-based courses 09-11 are designed for three roles in a public office: the official, the manager, and IT and the DPO. Each is one 3-hour block with a break that leads from the concerns holding AI back, through ways of reducing them, to a live demo of how the Polish AI stack solves the same problems: TWIN:DESK, PROXY:AI and the CDF methodology. Each ends with a document for the office: the Safe AI Work Card, the 90-day AI implementation plan or security requirements with a draft DPIA
Do participants of role-based courses work in SAIE?
They do not create accounts. The trainer shows the environment live, on anonymised documents of a sample municipality. Participants work in the browser: they do short checks and games from our knowledge base, answer polls and in the chat, and receive links to materials during the course and by e-mail afterwards. The checks and games require no registration and send nothing. If the office network blocks a service, the trainer shows it on their own screen. AI adoption leaders work on their own SAIE accounts in the LIDER:AI Academy
In what order should the manager, IT and officials attend?
First 10 Manager, because the manager decides on the goal, the rules and the budget. Then 11 IT and DPO: the IT specialist and the data protection officer prepare the technical conditions and the documents. Finally 09 Official: staff learn to work with the tool the office has designated. Every week the courses take place in this order, and the five use cases shown live are shared, so after the courses all roles talk about the same cases, each from its own side
Do we have to buy SAIE after a role-based course?
No. The documents from the course, that is the Safe AI Work Card, the 90-day plan and the security requirements, work with any tool, and you can ask the ten supplier questions of any supplier, including us. The SAIE demo lets you see how the rules work in a system, not only in an employee's memory. If you want to go further, the natural next step is a demonstration on the office's anonymised documents or an Initial Assessment in the CDF methodology. An office whose staff hold a certificate from our course can also send two leaders to the free LIDER:AI Academy and receive 5 SAIE trial licences for 2027, with no obligation to buy