Ez-RPH: AI-Powered Teacher Workflow Automation Ecosystem

About the Project

Keywords: Artificial Intelligence, Generative AI, Human-in-the-loop, daily lesson plans, Teacher workload Introduction Malaysia has more than 420,000 teachers in Ministry of Education (KPM) schools (Kementerian Pendidikan Malaysia, 2026). Every one of them writes daily lesson plans (Rancangan Pengajaran Harian, RPH) and keeps classroom-based assessment (Pentaksiran Bilik Darjah, PBD) records and much of this work repeats itself. Each time the school timetable changes, teachers have to key in or reset their lesson-plan templates again. Every new lesson plan starts from a blank page. PBD results are recorded once and then entered again into idMe, which comes to more than 120 entries for one class of 30 pupils. That entry takes time as well as when the server is slow, a session drops or facing unstable network, teachers are left unsure whether every record was actually saved. Studies link this kind of duplicated entry to systems that do not share data (Tsupari et al., 2024). Artificial Intelligence (AI) offers teachers support in planning, implementation and assessment (Çelik et al., 2022), and multimodal AI models now read the text in table images far more accurately than traditional Optical Character Recognition (OCR), although table layout remains harder for them (Nunes et al., 2025). Generative AI can produce a lesson plan aligned with a curriculum framework within 30 minutes, yet such drafts can contain missing details and a fabricated resource (Powell & Courchesne, 2024). Preservice teachers in Malaysia likewise found that ChatGPT saved time in lesson planning but gave factual errors and suggestions misaligned with the Malaysian curriculum (Chua et al., 2026). These findings point to hybrid human-AI designs in which AI augments, rather than replaces the teacher (Molenaar, 2022). So the question we started from was a practical one which is where in a teacher’s day can AI output be trusted? Objectives We set out to (1) use AI to turn a photo of a timetable into a checked teacher profile and weekly schedule; (2) use AI to draft daily lesson plans from the teacher’s chosen textbook pages, with the teacher still making the final call; (3) automate repeated PBD entry into idMe using fixed rules; and (4) find out whether practising teachers consider the system usable and useful. Methodology Ez-RPH was built in stages as one ecosystem on web, Android and iOS plus a browser extension for PBD entry. We kept to one rule: AI where the teacher can easily check the result, fixed rules where records must be exact, and the teacher verifies everything (Figure 1). In Step 1, the teacher uploads a timetable photo. A multimodal AI model turns the photo into a structured timetable, with each lesson’s day, time, subject, class and year, and flags unreadable cells instead of guessing. The teacher corrects the result, adds class sizes and saves. A new timetable just means a new photo. In Step 2, the teacher picks a learning standard and textbook pages as reference, and may type optional instructions for the AI. The AI drafts objectives, success criteria, activities and teaching aids that follow those instructions and stay with the chosen pages, which keeps the draft tied to the Malaysian textbook and curriculum. The teacher’s instructions take priority over the textbook text as well as every AI request is logged and rate-limited. The teacher reviews the draft before applying it, and always writes the reflection. In Step 3, Auto-PBD, the teacher chooses each pupil’s mastery level (Tahap Penguasaan, TP). Fixed rules fill in the matching remarks, intervention and effort level, and after one teacher login the extension enters them into idMe for checking and submission. The records stay in Ez-RPH, so a failed submission is resent, not retyped. We evaluated it at five hands-on workshops with a five-point Likert questionnaire. Results Six Ez-RPH modules are now in use. At five workshops, 117 teachers rated the system on ease of use, usefulness, time management, intention to use and recommendation. Item means ranged from 4.28 to 4.60 out of 5, with agreement between 84% and 95% (Table 1). These are teachers’ perceptions; we have not yet measured the time saved. In a recorded demonstration, Auto-PBD completed records for a class of at least 30 students within 30 seconds, although this was not a controlled comparison with manual entry. By 4 October 2026, Ez-RPH had 658 registered users, and the Auto-PBD extension had 138 weekly users. In practice, one photo of a timetable becomes a checked teacher profile and weekly schedule as well as a lesson plan begins from an editable draft based on the chosen textbook pages instead of an empty form. The 2026 release holds a MyIPO Certificate of Notification of Copyright (CRLY2026E08091). Implications and Potential Impact Ez-RPH shows one way to bring generative AI into schools without taking decisions away from teachers. AI handles the blank page, fixed rules handle the retyping, and the teacher still decides the TP, writes the reflection and submits the record. We expect this to give teachers more time for lesson preparation and for their pupils, though that still has to be measured. Running costs are low. There is no hardware to buy, because the system runs on the phones and laptops teachers already use, and AI use is paid for through credits. One teacher can start on their own, and a school or district can follow. The potential market is every KPM teacher, and the need comes back every reporting cycle. Ez-RPH is already on the App Store, Google Play and the web. New users get ten free AI timetable imports, then move to a monthly or yearly plan with AI credits. Seventy-seven teachers hold subscriptions through referral codes, and turning them into paying users is our next step. We also plan a timed comparison with manual entry, and further modules are in development. Acknowledgements We thank the teachers who joined the Ez-RPH workshops and gave us their feedback. Statement on Originality and Use of Generative AI This extended abstract is our original work. We used generative AI tools to help with drafting and language editing, and we checked every statement, figure and reference ourselves. We take full responsibility for the content. The AI features described in this abstract (AI timetable setup and AI-assisted RPH drafting) are functions of the Ez-RPH system itself. References Çelik, I., Dindar, M., Muukkonen, H., & Järvelä, S. (2022). The promises and challenges of artificial intelligence for teachers: A systematic review of research. TechTrends, 66(4), 616-630. https://doi.org/10.1007/s11528-022-00715-y Chua, K. H., Sathasivam, R. V., Mafarja, N., & Abdul Rahim, S. S. (2026). Exploring the benefits and challenges of AI-driven lesson planning among preservice science teachers. STEM Education, 6(1), 1-20. https://doi.org/10.3934/steme.2026001 Kementerian Pendidikan Malaysia. (2026). Maklumat guru. EMIS Online. Retrieved September 30, 2026, from https://emisonline.moe.gov.my/risalahmap/ Molenaar, I. (2022). Towards hybrid human-AI learning technologies. European Journal of Education, 57(4), 632-645. https://doi.org/10.1111/ejed.12527 Nunes, G., Rolla, V. G., Pereira, D., Alves, V., Carreiro, A., & Baptista, M. (2025). Benchmarking table extraction: Multimodal LLMs vs traditional OCR. In Proceedings of the 1st Joint Workshop on Large Language Models and Structure Modeling (XLLM 2025) (pp. 8-15). Association for Computational Linguistics. https://doi.org/10.18653/v1/2025.xllm-1.2 Powell, W. A., & Courchesne, S. (2024). Opportunities and risks involved in using ChatGPT to create first grade science lesson plans. PLOS ONE, 19(6), Article e0305337. https://doi.org/10.1371/journal.pone.0305337 Tsupari, K., Lagstedt, A., & Kauppinen, R. (2024). Revealing the digital transformation, uncovering hidden process dimensions of education. Business Process Management Journal, 30(8), 260-283. https://doi.org/10.1108/BPMJ-09-2023-0748