AI in the classroom is not magic but methodology – why P·A·R·T·S is the teacher’s new best friend
By Boris Mihailov (Teacher, lecturer at PU and AI trainer) | 12 November 2025, 10:59 am
Every educator who has tried to use artificial intelligence – whether Gemini, ChatGPT or another platform – knows this feeling: sometimes AI is a brilliant assistant and sometimes … it is simply a waste of time. In one moment it produces the perfect analogy for a complex scientific concept – an analogy that would take us, as teachers, hours to devise and polish. The next moment, on a seemingly similar task, it generates a text so superficial, factually wrong or just wooden that we wonder why we bothered at all.
Why is there such a drastic inconsistency? Why does this powerful tool behave so unpredictably?
From my experience as a teacher in school, at the university (PU) and as a trainer of hundreds of teachers, I see the same repeating story: a teacher, inspired by the news, experiments with AI. They receive a mediocre, unusable result. They become disappointed, decide that “this isn’t for me” or “it’s not ready yet” and give up. The technology, instead of becoming an ally, is abandoned and its enormous potential wasted.
The problem, however, is not the technology itself. The problem is that we still try to chat with AI instead of assigning it tasks. That is the battle for methodology.
At the moment all the major companies – Google, OpenAI, Microsoft and Anthropic – are in an intensive technological race. That is the visible part of the iceberg: who will build the “smarter,” faster and more capable model. Parallel to this, and perhaps more important for us, they are actively searching for and offering practical methodological models. These are guides, best practices and structured frameworks that show professionals how to use these complex tools effectively.
These frameworks, often called prompt models, are numerous. Acronyms such as R.I.C.E., C.A.R.E., B.R.O.K.E.R. and many others each try to “optimise” the system in their own way. Sites such as CPO Creativity even offer prompt generators to help structure these instructions. This diversity proves that the battle for methodology is real and fought on many fronts. Ultimately, a model is useless if no one knows how to extract value from it.
My role as a teacher and trainer is to be on this front line – not the technological one, but the methodological one. My job is to test, review and evaluate these experimental models in real conditions and to sift out those that actually work, bring value and can be applied directly in the Bulgarian classroom with its specific needs, curricula and constraints.
After testing dozens of approaches, one model stood out in my trials and trainings. It is Google’s model called P·A·R·T·S.
The problem: “Write me a lesson plan…”
Let’s be honest: most teachers start this way. They open the chat window and type their intuitive first request:
“Write me a lesson plan for photosynthesis for 9th grade.”
What they receive in response is a perfect example of pedagogical failure. It is a generic, often sterile and factually incorrect text. It doesn’t know my students. It doesn’t know that in 9 B I have three students with special educational needs and two for whom Bulgarian isn’t their native language. It doesn’t know that I have only 30 minutes of actual teaching time, not the full 45. It doesn’t know which topics we covered last week and which concepts are already clear or still difficult. The result is absolutely unusable in its original form and requires more time to edit than to write from scratch.
This happens because we treat AI as a conversation partner in a café rather than as a professional (albeit non‑thinking) assistant. That is where prompt engineering comes to help. I want to emphasise – this is not some elite computer science; you don’t need to be a programmer. It is simply the methodology of translating your pedagogical expertise into a language the machine understands: the skill of giving clear, precise, complete and structured instructions.
P·A·R·T·S is exactly such a methodology. Although it is only one of dozens of models, in my practice it has proven the most effective for pedagogical purposes. Its strength is not in its magical acronym but in the fact that it forces the teacher to do their own work first: to define the Persona (expertise), Action (the Bloom’s taxonomy verb), Requirements (constraints), Tone (emotional response) and Specifics (context).
From my expert experience, accumulated over hundreds of hours of training and real work at PU, I have seen that when educators move from intuitive, short prompts to a structured approach like P·A·R·T·S, the results transform fundamentally.
I’ve seen it dozens of times: at the beginning of the training the teacher is sceptical. After applying the structure, the same teacher sees how AI generates material that is 90% ready to use. AI stops being a lottery – “will it give me a meaningful answer today?” It becomes a predictable, powerful and above all effective assistant.
That is the difference between a toy that entertains us and a tool that helps us build. And in the battle for methodology I am betting on the tools that have proven their value in practice.
