As schools face growing demands on time and resources, AI is no longer a futuristic concept but a practical partner in classroom management and content creation. One example is Kira Learning, founded by Google Brain pioneer Andrew Ng. Its AI agents handle time-consuming tasks such as grading, lesson planning and performance analysis, freeing educators to focus on mentoring and personalised instruction. What can Kira do and what can we learn from it in terms of writing your own GPT prompts to achieve some similar time saving results.

Streamlining Lesson Planning and Assessment

Kira’s primary agent drafts lesson outlines based on your syllabus, embedding Socratic prompts to encourage critical thinking. Rather than static slide decks, it generates interactive activities aligned with Bloom’s taxonomy. Meanwhile, a secondary agent uses natural language processing to assess student submissions, providing instant feedback and highlighting common misconceptions. Early adopters cite up to 40 % time savings on prep and marking.

Boosting Data-Driven Interventions

Beyond content, Kira integrates seamlessly with popular learning management systems. It compiles attendance, assignment and engagement data into unified knowledge maps. These maps visualise comprehension gaps, allowing you to launch targeted interventions before small issues escalate. Automated workflows generate intervention suggestions—such as peer mentoring or micro-lessons—with a single click, reducing administrative overhead by nearly a third.

GPT Prompts for Efficiency

To replicate these gains, try the following GPT prompt for your lesson planning:

You are an educational designer. Develop a 45-minute lesson plan on [topic] for [student age group], using Socratic questioning. Include objectives, key questions and interactive activities.

For automated summarisation of student essays:

You are an AI teaching assistant. Summarise the main arguments in this essay and identify two areas for improvement.

Using models like GPT-4 Turbo offers a balance between speed and depth, while specialised fine-tunes (e.g. OpenAI’s [education-optimised model]) can further refine outputs.

A Human-Centred Approach

Key to success is ensuring AI supports rather than supplants educators. Kira’s agents present their suggestions alongside confidence scores, inviting teacher revision. This human-in-the-loop design fosters trust and maintains pedagogical rigor. Institutions that adopt this model report higher teacher satisfaction and better student outcomes.

Looking Ahead

AI agents will evolve from assistants to collaborators, capable of multi-agent workflows that handle planning, evaluation and resource allocation in concert. By embracing these technologies today, schools can optimise workflow, enhance teaching quality and reclaim precious time for the human connections at the heart of education.