My AI Toolkit: What I Use, What I Don’t, and Why

A transparent look at which AI tools I really use, where I use them and why each one has earned a place in my practice.

Spend any amount of time reading about AI in education and it can quickly feel as though there is a new tool to learn every week. One platform promises to plan lessons, another to differentiate them, another to mark pupils’ work, and another to turn everything into a quiz, video or presentation.

It is easy to feel behind before you have even begun.

The reality of my own practice is much less dramatic. I do not use dozens of AI tools, and I do not use one tool for everything. My choices are influenced not only by what each tool can do, but also by where I am, which device I am using and how demanding the task is.

If I am being completely transparent, this is what my current toolkit really looks like:

• Gemini is on the home screen of my iPhone for quick, everyday use

• Copilot is the tool I use in school through Microsoft 365 on my C2k Surface Pro

• I pay for a ChatGPT subscription and use it for the heavy lifting

• NotebookLM was my first real AI “love” and changed how I worked with source material

• Wayground, formerly Quizizz, has helped me build a bank of quizzes and flashcards for retrieval and revision

I have also experimented with Adobe Express for Education, Canva and AI-generated visuals, and have begun exploring more specialised assistants, but they are not all part of my everyday teaching.

That is deliberate. I would rather understand a few useful tools well than open accounts on twenty platforms and never get beyond the surface of any of them.

This is not a definitive ranking of the “best” AI tools for teachers. Tools change, school access varies and what works in one subject or classroom may be unsuitable in another. It is simply an honest snapshot of what I currently use, what I have not adopted and why.

Gemini: the AI on my home screen

Gemini is probably the clearest example of how convenience shapes the technology we actually use.

It sits on the home screen of my iPhone, so it is often the AI I reach for during everyday life. If I have a quick question, want an explanation or need to explore something while my phone is already in my hand, Gemini is readily available.

I do not generally use it for work or for my largest and most complex teaching projects. Its role is lighter and more immediate.

That does not make it unimportant. A tool earns a place in a routine partly because it is accessible at the moment it is needed.

It is also a reminder that asking which AI is “best” may be the wrong question. The AI I use for a quick everyday query on my phone is not necessarily the one I choose for sustained work on an A-level workbook or a detailed analysis of teaching material.

ChatGPT: the tool I pay to do the heavy lifting

I pay for a ChatGPT subscription. I think that is worth stating because it is part of an honest account of my experience.

I use it frequently and intensively enough that I have chosen to pay for it, although that does not mean every teacher needs to do the same.

When a task is large, complex or likely to involve several rounds of development, ChatGPT is usually where I go. Its main value for me is flexibility. I can begin with a rough idea, continue the conversation, question the response and refine it over several stages.

The Projects feature also helps me keep different areas of work organised and return to longer pieces of development without beginning again in a general chat.

I do not use ChatGPT only because it may save time. Some of its greatest value has come from helping me explore possibilities I might not otherwise have considered. AI can be a platform for innovation, not simply a shortcut.

Its versatility is also one of its weaknesses. ChatGPT can produce an answer that sounds confident, complete and professional while being inaccurate, too generic or poorly matched to the course I am teaching. It may combine material from different contexts, include content pupils do not need or create a question that does not reflect how a topic is actually assessed.

The more specialised the task, the more important my own subject and assessment knowledge becomes.

ChatGPT gives me a starting point. It does not give me permission to stop thinking.

Microsoft Copilot: the AI I use in school

In school, I use Microsoft Copilot through Microsoft 365 on my Surface Pro. It sits within the managed digital environment available to me through C2k, so it is the most natural AI for me to access during the school day.

For many teachers, familiarity matters. A tool that sits within an environment they already know may feel more manageable than an entirely new platform with another account, layout and workflow.

At present, Copilot is the AI that fits most naturally into my school environment.

I do not assume, however, that Copilot and ChatGPT will always give the same quality of response, or that every teacher will see the same features. The version available, the account being used and the school’s digital arrangements can all affect the experience.

This is one reason training sessions can sometimes be confusing: the presenter clicks a button that does not appear on everybody else’s screen.

My choice between ChatGPT and Copilot is therefore often practical rather than ideological. Which tool is available? Which one suits the task? Where are the materials I am working with? What does my school permit me to use?

Familiarity does not remove the need for care. I do not enter identifiable pupil information simply because a tool appears inside a familiar system. School policy, safeguarding, data protection and the distinction between personal and school accounts still matter.

Looking ahead: Copilot Agents as my AI Swiss Army knife

Microsoft Copilot forms part of a wider GenAI programme being rolled out to teachers across Northern Ireland.

As that rollout continues, I can see Copilot becoming a bigger part of my workload-reduction toolkit.

Copilot Agents, in particular, could become my AI Swiss Army knife: adaptable tools with different uses depending on the subject, class or task.

I have begun experimenting with specialised assistants, and the more I explore Copilot Agents, the more possibilities I can see. Not as one tool expected to do everything, but as a collection of focused tools configured for particular purposes.

NotebookLM: the tool that rewired my workflow

If I am honest, NotebookLM (now Gemini Notebook) was my first real AI love.

It changed the way I worked with source material because it allowed me to build a collection of sources I had chosen myself and work within that defined body of material.

That was particularly useful when working with CCEA specifications and related teaching resources. I could decide which documents, specifications, notes and resources should shape the work rather than beginning with a broad, generic response that might reflect a different curriculum or assessment context.

I used NotebookLM extensively during the last school year to keep research and the production of student resources within a focused collection of relevant material.

That sense of containment was extremely valuable. It gave me a different way to explore, organise and reshape selected material without losing sight of the sources I had chosen.

NotebookLM does not eliminate checking. If the source collection is incomplete, the response may also be incomplete. It can overlook context, misinterpret wording or give more weight to one document than I would.

“Grounded in sources” does not mean “guaranteed to be correct”.

Copyright also matters. Being able to upload a document does not automatically give us the right to republish its contents. I still need to consider what I am entitled to use, how the output will be shared and whether it reproduces protected material too closely.

NotebookLM (Gemini Notebook) is not always my first choice for open-ended creativity. When I want to explore, organise and work with a carefully chosen body of material, however, it offers something different from a general chatbot.

That source-focused way of working is what made it such a significant part of my workflow.

Wayground: my retrieval and revision bank

I began using this platform when it was called Quizizz. It is now Wayground.

I have used it for retrieval practice and, over time, built a bank of quizzes and flashcards that pupils can use for revision.

That accumulated bank matters. This is not simply a platform I opened once during a training session. It has earned a continuing place in my practice.

What have I not tried, and why?

Some well-known names are missing from this post for a very simple reason: I have never used them.

I have not tried Claude, for example. It would therefore be unfair for me to review it, compare its quality or suggest that teachers should or should not use it.

Its absence is not a judgement. The tools already in my routine have done what I have needed them to do so far.

I think of it a little like owning a car. If you know it, love it and trust it, and it has not given you any bother, why would you change it without a good reason? Another model may have different features, but that does not automatically mean it would serve you better.

The same applies to AI.

Many tools overlap. A platform may advertise ten teacher functions, but I may already be able to do most of them using a tool I understand well. Learning another platform also takes time. I need to understand its settings, test its output, learn its limitations and decide whether it is safe and suitable. That time needs to lead to a genuine benefit.

Access matters too. Free and paid versions differ, features change, and something available through a personal account may not be approved for school use.

Most importantly, I want evidence from real practice. A five-minute demonstration can make almost any tool look impressive. The more useful questions come later: How much checking did it require? Did it work with pupils? Did it improve learning? Would I choose it again when the novelty had worn off?

I am open to trying something new when it solves a problem my existing tools cannot solve. I just do not need to change platforms for the sake of novelty.

There are also creative tools I have dabbled with rather than fully adopted. I have received training in Adobe Express for Education and Canva and have some experience of using both. For now, they sit outside the small group of tools I use most regularly.

What am I not currently using AI to do?

The boundary is not only about particular brands. There are also tasks for which I remain cautious about using AI at all.

AI marking: where I still feel uneasy

At a conference last year, I watched a teacher demonstrate how they were using AI to mark pupils’ work.

I could understand the appeal. Marking takes a considerable amount of teachers’ time, and anything that promises faster, more thorough or more consistent feedback will attract attention.

At the same time, we frequently talk about pupils using AI to write essays as a form of cheating.

Full disclosure: I would feel somewhat guilty if I used AI to mark my pupils’ work primarily to save myself time. If I expect a pupil to submit work that reflects their own thinking, it feels reasonable that they might expect the response to come from my professional thinking too.

I would feel differently about using AI as a back-up or second pair of eyes. Once I had marked the work myself, it could potentially help me check that I had been thorough, applied the criteria consistently or not overlooked something important. In that role, AI would be supporting my judgement rather than replacing it.

In practice, AI marking is something I have deliberately steered away from with extended Biology responses. These require interpretation. Pupils may express a valid idea in an unexpected way, make a partially correct link or reveal a misconception that needs a carefully judged response. I still want to read that thinking for myself.

That does not mean I avoid all automated marking. I use Microsoft Forms for multiple-choice retrieval quizzes, where pupils’ answers can be checked automatically.

Strictly speaking, this is automated scoring rather than generative AI deciding the mark. I have set the questions and decided the correct answers in advance. The task has a clear right-or-wrong outcome.

For me, there is an important difference between automating an objective check and delegating professional judgement.

I decided to ask my A-level students what they thought about AI marking. The suggestion was not looked upon favourably. In fact, they were aghast that I would even consider it.

I then asked them, “What if the AI marked your work better than I could?”

That made them pause. I could almost see the cogs beginning to turn.

I still did not receive an enthusiastic response, but the question made the issue less straightforward.

Their reaction reminded me that the debate is not only about whether AI can mark accurately. It is also about trust, fairness and what pupils believe their teacher owes them. Feedback is part of the relationship between teacher and pupil.

I have not reached a final answer on AI marking. For now, I am more comfortable with the idea of AI checking my work than doing my work for me, and only where privacy, school policy and the handling of pupils’ work have been properly considered.

My simple rule for choosing an AI tool

Before using an AI tool, I now try to ask four questions:

1. What actual problem am I trying to solve?

2. Why is this tool a sensible fit for that problem?

3. What will I need to check, correct or adapt?

4. Will the final benefit justify the time and any additional risk or complexity?

If I cannot answer those questions, I probably do not need the tool yet.

A small toolkit can still be a useful one

Teachers do not need to become collectors of AI apps. Competence is not measured by the number of platforms we can name or the number of tabs open across the top of a screen.

For me, competence means knowing a few tools well enough to choose between them, question what they produce and recognise when not to use them.

It also means being willing to change my mind as the technology, the evidence and my own practice develop.

My current toolkit will not remain fixed. I will continue to explore new tools when they offer a clear possibility for better teaching, better learning or a genuinely useful new idea.

But I do not need to chase every new release to feel that I am keeping up.

The goal is not to use more AI. The goal is to use it more thoughtfully.

There is far more to say about how I use each of these tools in practice, including the resources I have created, what worked, what needed repairing and what I would do differently. I will explore those classroom applications individually rather than trying to squeeze them all into one post.

Which AI tool will become the go-to choice for teachers in Northern Ireland and beyond? I already have a strong contender in mind, but I will be putting it through its paces this term and sharing what I find.

If you have a standout tool in your own AI toolkit, share it in the comments. If you are interested in seeing how I use these tools in the classroom, subscribe below to follow future posts.