2025-01-24

Generative AI for educators: a practical guide

The EU's AI Act has done what many major regulations do: it has created both opportunities and concern. For educators and public administration, this means that AI's potential must now be balanced against clear rules and areas of responsibility. When IMY and DIGG were tasked with drawing up national guidelines for generative AI, it was not merely a technical question – it was a strategic one. With 18 parts across seven areas, the documentation can seem as impenetrable as a legal labyrinth. But at its heart, it all boils down to one core message: have a plan, protect personal data, analyse risks and procure consciously. Generative AI is not a solution but a toolbox, and as with all tools, it is how you use them that makes the difference.

Does this matter to me?

Why does generative AI matter to you? The answer is simple: because it is already changing how we work and learn. A survey by the Swedish National Agency for Education (Skolverket) showed that almost half of all compulsory school teachers use AI services, and usage is higher among older students and their teachers [1]. It is not just a trend, it is a movement. The public sector (including schools) has been identified as one of the areas that stands to gain the most from generative AI, with an estimated economic potential of no less than SEK 125 billion [2]. For the world of education, this means a wave of adaptation from system providers competing to offer new solutions for teaching, administration and learning. AI is no longer a question of 'should we?' but of 'how should we?'. And even if it can feel like a technological imperative, the benefits are tangible. Forecasts show that Sweden's GDP could grow by 9% through broad implementation of generative AI. So whether you are a teacher, school leader, student or administrator – AI is no longer something that happens in the future. It is happening here and now. The question is not whether we are ready, but how we adapt.

Taking the first step: Start with the needs

How do you begin the journey towards implementing generative AI? The answer is to take a step back and reflect on a simple but crucial question: What does the organisation need? The need is your compass. For municipalities just stepping into the AI landscape, the first step may simply be to explore the possibilities (and limitations) of today's AI. On the other hand, a more AI-experienced municipality may already have a clear vision and have identified specific areas where AI can support staff. But once the needs have been mapped, what comes next? The next step is to understand the market and the providers available. This is not just practical advice ahead of a procurement, it is also an investment in building a robust AI strategy. Understanding which systems and solutions exist helps you make informed decisions, from formulating AI policies to identifying the right tools for your specific needs.

To get started, you can begin with resources such as AI Sweden and their networks (don't forget their startup network) [3], or mappings such as the Sweden AI Startup Landscape [4] and Swedish Edtech Industry's edtech map [5]. Do you already have framework agreements in place? Then you can review which providers are already registered and available to call off from (it saves both time and headaches). In other words: starting right is not just about finding the right solution, but also about making the journey simpler and more strategic.

A plan for AI: Setting the direction for the future

Do we have an AI policy? That is a question every organisation should ask itself as generative AI steps onto the stage. Whatever your role in the world of education, an AI policy can be crucial for setting a clear direction and ensuring the technology is used responsibly and effectively. For most, this work will take place at municipal level, but if you have the opportunity to influence the process, here are some crucial steps to consider.

Define the goal and scope: What do you want to achieve with the policy? Which types of AI should it cover? Setting the boundaries for what the policy should address is the first step towards clarity. Think long term!
Include school operations: Should school operations be covered by the policy? The answer is almost always yes. AI affects education directly, and guidelines are needed to manage both opportunities and risks in the classroom.
Application and division of responsibility: How should the policy be implemented across administrations and committees? A clear plan for application ensures the guidelines do not just become a paper product but also a tool for practical change.
Rules of conduct and ethical principles: Which rules should govern AI use? Here it is crucial to ensure fairness, transparency, accountability and privacy. The UN Convention on the Rights of the Child is an excellent aspect to include in order to build a policy that prioritises the best interests of children and protects against bias and discrimination.
Data protection: Data is one of the most critical issues when it comes to AI. Here you can lean on IT/system management to ensure that data protection and GDPR compliance are handled correctly.
Limitations with the future in mind: If you are going to introduce limitations, be careful. AI develops rapidly, and it is important that the policy does not restrict future innovations or solutions that may become relevant. Keep the policy flexible.
Training initiatives: How much training on generative AI should staff receive? Equipping employees with knowledge and skills is crucial both for the policy to be followed and for building confidence around AI use.

By following these steps, you are not only helping to create a policy, but also a strategic foundation for how AI should be used to improve the school's operations. The policy is more than just a document, it is a promise to use the technology in a way that benefits society and future generations.

From market scan to action: The next step in the AI journey

So, you have done your research, created or read your AI policy and drawn up a list of potential providers – what happens now? If you are not responsible for purchasing yourself, you can simply hand over your list and interest to those who manage procurements and continue to the next chapter of this guide. But if the responsibility for purchasing rests with you, it is a matter of navigating carefully and avoiding pitfalls. This is where a risk assessment comes in.

The risk assessment can feel like a time-consuming process, but it is crucial for ensuring the right solution is purchased. To help you along the way, here is a simple checklist to start with when evaluating generative AI solutions:

How does the intended use relate to the AI Act?
Is the fee structure clear and transparent?
Are there limits on how much the system may be used?
What types of content (modality) can the system generate and handle?
Who owns the rights to what the system produces?
Can the provider use what is entered into or produced by the AI, and is there a risk of data leaks or GDPR breaches?
Is data processed and stored within the EU?
Does the user have influence over where and for how long data is stored?
What responsibility does the provider take, and how is this regulated in the agreement?
Is the agreement clear on termination, notice periods and other important terms?
Can the system be procured directly or is it covered by existing framework agreements?

This checklist is your guide to ensuring that the fundamental aspects of a potential solution have been carefully thought through. It can be a time-consuming process, so start right away and work systematically.

Once the system is in place – what happens then?

Once you have found and acquired solutions that meet your needs, the next step is about maximising their value. But that is an entirely new phase – and it requires its own plan. Let's dive into what happens after the purchase.

The system is in place: How to use it the right way

You now have generative AI in your toolbox and are ready to use it – but how do you make sure it is used responsibly? Here comes a keyword you must always keep in mind: GDPR. Working with AI and personal data carries a responsibility that cannot be compromised, but let's make it a little more approachable with an analogy.

Imagine you borrow books from a library to teach your children about history. How would you handle those books? Probably like this:

You use the books only for the purpose they are intended for – reading and learning.
You borrow only the books you really need.
You are careful with the books, so they are not damaged or lost.
You return the books when you are done, or when the loan period expires.

You should think the same way when handling personal data with the help of generative AI:

Use the data only for the specific purpose you collected it for.
Collect only the data necessary to achieve that purpose.
Protect the data from unauthorised access and make sure it does not ”disappear” or end up in the wrong hands.
Delete or return the data when it is no longer needed.

By thinking of personal data as something you 'borrow' – to be handled with care and returned on time – it becomes easier to understand and comply with the GDPR in practice. It is about respect for personal privacy, whether it concerns students, teachers or municipal residents.

Of course, there is a wealth of details and legal nuances that could fill pages best suited to a separate post, but as long as you already have basic GDPR guidelines in place, you can use these principles as practical guidance. With this in place, you are ready to move on – from having the systems to actually starting to create value with them.

The relationship between Skoolie and municipalities: Simplicity, security and clarity

What can you expect if you choose Skoolie as your provider? The answer is simple: a solution that delivers not only results but also confidence and clarity. By revisiting the checklist we went through earlier, we can make concrete what Skoolie offers:

Clear fee structure: Everything is transparent and presented when responding to call-offs via framework agreements or direct procurement. No surprises, just a clear picture of costs.
No usage limits: You can use the system as much as you need without worrying about restrictions.
Generates text: The platform's focus is on producing text-based content.
Ownership of what is produced: Everything you create via the platform is entirely yours. Skoolie retains no rights to the content.
No training on your data: Skoolie does not use entered or generated data to train AI, which eliminates the risk of unauthorised use of personal data via the AI.
Data processing within the EU: All data is processed and stored within the EU, with AI processing carried out specifically in Sweden – extra reassurance for GDPR compliance.
Flexibility over storage and retention periods: The customer, as data controller, decides this through the data processing agreement.
Division of responsibility: Clear roles between public administration, the IT distribution partner and Skoolie make it easy to know who is responsible for what.
Clear terms for termination: The agreement contains clear terms on termination and notice periods, so it is easy to end the collaboration if needed.

Skoolie makes it simple, secure and smooth for the school sector to use generative AI. Whether it is about supporting students or making life easier for staff, the goal is for AI to create real value without compromising security or compliance.

If your municipality is unsure whether you fall below the procurement threshold, you are welcome to contact us at Skoolie. We are happy to provide all the information and guidance you need. With Skoolie you get not just a provider – you get a partner who understands the needs of education and municipalities in a time of technological change.

In closing

If reading this has made you curious to dive deeper into DIGG and IMY's guidelines, you can read them on DIGG's website [6]. And if you have suggestions for improving this guide, please do get in touch – this is a living document that we at Skoolie will continue to update.

Rasmus Häggkvist - rasmus.haggkvist@skoolie.se