AI Agents in Education: What They Are and How They Transform Teaching

After the recent launch of IGNITE Copilot, as well as its adoption by more than 15,000 teachers, IGNITE’s management team is focused on helping educational centers in this mindset change, accompanying teachers who need to efficiently manage their knowledge base.

Without a doubt, we understand that process optimization in an educational center must have ‘top-down’ governance, meaning it must be on the agenda of the Director of the educational institution.

The pedagogical and development direction of IGNITE Copilot has defined an internal evolution of the application, based on an ‘intelligent educational process’.

Specifically, we identify 3 dimensions to achieve center efficiency with the help of AI:

Process automation (such as creating Learning Situations and their scheduling, as well as conversational assistants to help in the teaching process. [CURRENT]

Regarding process automation, we will be able to invoke autonomous agents, calling them from the process itself [roadmap end of 2025].

Orchestration of all educational processes of the teaching team, including automations and AI agents.

🔍 What is an artificial intelligence agent, in the context of teaching?

It is an evolution of generative AI: it not only generates content, but detects contexts, decides actions, executes them and learns from the results.

These agents work autonomously and independently, forming a “digital workforce” alongside humans.

SOURCE: The future of work is agentic – McKinsey 3 June 2025

🛠 How to design educational agents in academic centers

Next, we are going to share how IGNITE Copilot engineers are defining each process of standard didactic planning, in order to identify all the microsteps, and ultimately create well-thought-out engineering:

Selection of clear processes In organizations, the first use cases have been highly structured processes (help desks, tickets, customer service). In education, this involves identifying repetitive and well-defined tasks, such as lesson planning, rubric creation, or grade management.

Process engineering with microsteps It is necessary to meticulously document each step of the processes:

• What information is needed (objectives, standards, criteria).
• What decisions to make (type of activity, level of detail, didactic resources).
• What to deliver to the agent (templates, reference content).

Agent tuning: continuous training Requires:

• Content experts (teachers or pedagogues) to nourish the agent with updated knowledge.
• Prompt engineers, who clearly define how the agent should react to each situation

Implementation with improvement loop

Similar to customer service: teach the agent to plan, receive feedback (is the plan coherent, adequate for the objectives?), and adjust.

Didactic planning can be simulated and its quality evaluated, e.g., generating a teachable plan and comparing it with standards.

Adoption and change management

It is advisable for younger instructors to experiment first, adapting traditional processes.

The role of educational leadership and the management team will be key to provide accompaniment, show results and create confidence

🎯 Strategic benefits in the classroom

Reduction of repetitive tasks: the agent can quickly generate a first draft of classroom planning, with activities and materials.

Access to best practices: loading rubrics, previous experiences, curricular bibliography, the agent suggests coherent alignments.

Continuous learning: after each use, the agent improves its proposal via feedback analysis.

Teaching focus: frees the teacher from bureaucracy and redirects them to pedagogical design, personalized attention and methodological improvement.

🧩 Suggested model for educational agent design:

Example of Didactic Planning with an AI Agent:

Now we are going to create an AI educational agent capable of generating a complete quarterly didactic planning for the area of Biology and Geology in 2nd ESO, first quarter, in accordance with LOMLOE, and including various learning situations. The agent will take into account classroom diversity: 1 student with high abilities and 3 with dyslexia.

🧠 SYSTEM PROMPT (configures the agent as a quarterly planner)

You are an educational agent specialized in quarterly planning of the Biology and Geology area in ESO, in accordance with the LOMLOE curriculum (RD 217/2022). Your function is to generate complete didactic plans for each quarter of the school year. You follow this structure:

1. General description of the quarter (specific competencies, basic knowledge, methodological approach).
2. Planning table with the learning situations of the quarter. Each row includes:
   • Title
   • Duration (weeks and estimated sessions)
   • Basic knowledge addressed
   • Final product
   • Main assessment
3. Development of each learning situation (complete structure: objectives, description, justification, didactic sequence, resources, grouping, didactic strategy, UDL and rubric).
4. Adaptations for diversity (dyslexia, high abilities)
5. Overall quarter assessment (assessable criteria, rubrics and instruments)

You are precise, rigorous, clear and creative. You use realistic examples, adapted to the 2nd ESO level. Your proposals help teachers save time, improve pedagogical quality and attend to classroom diversity.

💬 USER PROMPT (introduced by the teacher):

Design a complete didactic planning for the first quarter of Biology and Geology in 2nd ESO. Include 3 learning situations that address the main contents of the quarter. In the classroom there is one student with high abilities and three with dyslexia. Methodological adaptations, UDL, competency assessment and alignment with the LOMLOE curriculum must be contemplated.

✅ RESULT GENERATED BY THE AGENT

🧭 1. General description of the quarter

Specific competencies worked: CE1. Understand the cell as a structural and functional unit of living beings. CE2. Analyze the organization of living beings and their vital functions. CE3. Develop scientific thinking and communicative skills in sciences.

Key basic knowledge:

• The cell: types, structure, functions.
• Cell theory.
• Observation and analysis under the microscope.
• Classification of living beings.
• Common characteristics and biological diversity.
• Scientific method.

Methodological approach: Project-based learning, guided inquiry, scientific gamification, cooperative work, attention to diversity with UDL principles.

3. Summary development of each learning situation

📌 Situation 1: “Journey to the interior of life”

• Objectives: Understand cellular structure, identify organelle functions, visually represent a cell.
• Final product: 3D cell model + group oral presentation.
• Activities: introductory videos, group work on organelles, guided research, model construction, exposition.
• Adaptations:
  • Dyslexia: visual templates, oral supports, screen readers.
  • High abilities: optional complex model design, expanded scientific exposition.
• Assessment: rubric by assessment criteria + individual reflection.

📌 Situation 2: “Microbe hunters”

• Objectives: Learn to use the microscope, observe plant and animal cells, record scientific observations.
• Final product: Illustrated digital laboratory notebook.
• Activities: real practices, use of microscope and mobile app, diagrams, visual diary.
• Adaptations:
  • Dyslexia: audio recording of observations, use of colors for diagrams.
  • High abilities: comparative observation between cell types, expanded reports.
• Assessment: control list + self-assessment with adapted rubric.

📌 Situation 3: “Tree of life: classify to understand”

• Objectives: Understand criteria for classifying living beings, build dichotomous keys, communicate scientific knowledge.
• Final product: Digital infographic or poster about classification.
• Activities: example analysis, classification games, key design, poster creation.
• Adaptations:
  • Dyslexia: use of images and pictograms, voice dictation.
  • High abilities: creation of their own alternative taxonomy.
• Assessment: competency rubric + group co-assessment.

🧩 4. Overall quarter assessment

Assessment criteria worked:

• CE1.1. Explains cellular structure with appropriate language.
• CE2.1. Analyzes similarities and differences between cells.
• CE3.2. Applies the scientific method in observations.
• CE3.3. Communicates scientific knowledge with rigor and clarity.

Assessment instruments:

• Rubrics per situation
• Direct observation
• Student diary
• Co-assessment + self-assessment
• Practical tests

♿ 5. Featured UDL strategies

Multiple forms of action and expression:

Physical models, infographics, oral presentations, interactive diagrams.

Tools: Canva, Tinkercad, voice dictation, PowerPoint.

Multiple forms of representation:

Videos, animations, 3D models, tactile resources.

Tools: Phet Simulations, Edpuzzle, Genially.

Multiple forms of engagement:

Gamification, challenge-based work, choice of final products.

Tools: Kahoot, educational Escape Room, visual tools.

✨ Conclusion

As we all know in the Education sector, both teachers and edtech solution providers, AI democratizes prediction: It reduces the time to make didactic planning and classroom programming.

A step further we have with this ‘digital twin’ of the teacher, which minimizes temporal inefficiencies, and knows the educational and regulatory context.

From IGNITE Copilot, we suggest 3 differentiated phases, for AI adoption in the learning process:

Customization of an AI Protocol and Policies for the educational center. We provide you with a base to start from: https://ignitecopilot.ai/#protoc_centros

Implement the didactic planning processes that I intend to optimize transversally, in the different educational stages (process reengineering). This ‘process mining’ prior to connection with IGNITE Copilot, is what will inform us why we go faster or slower, and from there we decide whether to automate.

Involve IGNITE Copilot’s AI, Enterprise mode, to adopt part of the ‘intelligent process’ designed in B): https://ignitecopilot.ai/planes-y-precios/

The use of well-designed educational agents allows:

• Faster and more consistent planning and programming.
• A teaching model focused on what humans do best: empathy, creativity and pedagogical reflection.
• An evolution of the teaching role towards design, accompaniment and continuous improvement.

There is no educational AI without prior process intelligence.

TIP: Teacher, it’s not the algorithm – it’s the process.