AI in Educational Design: What It Does Well, What It Still Fails At, and Why the Teacher Remains Unreplaceable

For the past four years, one question has generated waves of anxiety in university corridors around the world: Can artificial intelligence plan—or even teach—a course better than a human educator? The short answer, according to the available evidence so far, is no. The long answer, however, requires nuances worth exploring carefully, because between technological enthusiasm and corporate fear lies a more sober and productive territory: that of critical collaboration.

---

What AI Really Delivers in Educational Planning

Before any criticism, it is fair to acknowledge what generative artificial intelligence already does in a concrete and useful way within instructional design.

1. Research and Content Mapping at Unprecedented Speed

When an educator starts building a course, one of the first tasks is to track bibliographies, extract key concepts, and map the available knowledge on a given topic. This exploratory process, which could previously consume entire days of work, can be compressed into minutes with the help of tools like ChatGPT, Claude, or Gemini.

AI scours, organizes, and presents an initial overview of the field of knowledge at a speed no human can match. This is not trivial: saving time in exploratory stages frees cognitive energy for the decisions that truly matter.

2. Structuring Content and Didactic Sequences

In addition to research, AI is competent in the task of organizing information into hierarchical frameworks, grouping themes and subthemes coherently. It can generate, in seconds, a proposed syllabus, a list of topics for each lesson, or a sequence of activities that serve as a starting point for the educator's work.

Drafting a first draft of learning objectives, creating preliminary assessment rubrics, or suggesting classroom dynamics—all of this falls within the current capabilities of AI, and denying these advantages would be, at the very least, naive.

3. Expanding the Field of Vision

There is a less obvious but equally important benefit: AI reduces the blind spots caused by biases or time limitations of the educator. An experienced teacher tends to organize a course according to their own theoretical references and past experiences. AI, trained on massive volumes of text, can suggest approaches, authors, or perspectives that the human expert might not consider—not because they are less competent, but simply because they haven't had the time or access.

---

Where AI Stumbles: Limitations That Cannot Be Ignored

If AI were simply "good at all this," the discussion would be over. It is not. Understanding where AI's criteria come from—and where it fails—is crucial for using it responsibly.

AI's Criteria Come from Patterns, Not Understanding

The apparent "common sense" that AI exhibits when putting together a lesson plan essentially derives from statistical patterns extracted from texts produced by humans. It recognizes that certain elements tend to appear together in instructional design documents and reproduces them fluently. But this is fundamentally different from understanding why those elements work pedagogically.

This means that:

• AI can generate learning objectives that are grammatically correct, but pedagogically empty
• It can suggest activities that seem coherent on the surface but ignore the specific realities of students, their cultural contexts, and their concrete difficulties
• It doesn't know when a didactic sequence is mistaken because it cannot test hypotheses in the classroom

The Lack of Situational Context is Critical

One of the pillars of effective instructional design is contextual diagnosis: who are the students? What is their entry level? What resources does the institution provide? What are the job market expectations for this profile? What is the institutional culture?

AI has no access to any of these variables—unless the educator explicitly provides them. Even when provided, AI's ability to interpret contextual nuances is limited. It works with what is in the prompt, not with what is implied in the lived reality of a classroom.

Genuine Pedagogical Creativity and Assessment

Another relevant limitation is that AI tends towards the commonplace. Ask any generative tool to suggest assessment activities for a philosophy course, and it will likely return: debates, argumentative essays, seminars, portfolios. All correct, nothing surprising.

Genuine pedagogical creativity—that which arises from the intuition of a teacher who knows their class, who recognizes a moment of opportunity, and improvises a transformative dynamic—is, for now, far beyond AI's capabilities.

---

The Working Model: Critical Collaboration, Not Total Automation

The most common mistake in the debate about AI and education is to present it as a binary dispute: either AI replaces the teacher, or it is useless and should be ignored. Neither of these positions reflects reality.

What current evidence suggests is a model of critical collaboration, where:

• The educator uses AI to accelerate the operational stages of design (research, structuring, first drafts)
• The educator reviews, questions, and adapts the material generated in light of their pedagogical knowledge and the real context of the class
• AI broadens the repertoire of possibilities without making the final decision on what is pedagogically appropriate
• The teacher maintains epistemic and ethical authority over the teaching process

This division of labor is not a concession to technologism—it is an honest reading of the competencies of each agent.

---

What This Means for Educational Institutions

For educational managers and coordinators, the message is clear: investing in AI for educational design makes sense, provided that this investment includes teacher training for the critical use of these tools.

Automating pedagogical planning without this training is a path to generic curricula, superficial assessments, and learning experiences that seem modern but are pedagogically impoverished.

Some practical recommendations for institutions:

• Train educators in the critical use of generative AI tools
• Establish protocols for mandatory human review for all AI-generated material
• Explicitly value the contextual and relational knowledge of the teacher, which AI cannot replicate
• Experiment incrementally, evaluating real impacts on student learning

---

Conclusion: A More Sober and Useful Territory

Artificial intelligence has arrived in educational design to stay. It is already useful and is likely to become even more so in the coming years. But the idea that it can replace the educator in planning and conducting courses—an idea that excites some while terrifying others—does not find support in current evidence.

What exists, and what is worth cultivating, is a territory of collaboration in which the speed and breadth of AI combine with the criteria, contextual sensitivity, and pedagogical creativity of the human teacher. This territory is less glamorous than the promise of total automation, but it is infinitely more honest—and much more effective for those who truly matter in this equation: the student.