AI for Large Class Sizes in Kenya
The Competency-Based Curriculum doesn't just ask a Kenyan classroom to cover content — it calls for ongoing observation, recording, and reporting on each learner's individual progress toward specific competencies, a model designed with a manageable class size in mind. In a Standard 3 room of 55 or more, that observation-based approach collides directly with the same enrollment surge Free Primary Education created two decades earlier.
Quick Answer: In a large Kenyan classroom under the Competency-Based Curriculum (CBC), AI's most realistic contribution is speeding up the individual observation, recording, and reporting CBC's formative model demands — generating structured observation checklists, competency-aligned task variations, and batch-drafted portfolio comments a teacher still verifies. It cannot add the teachers TSC's own data shows Kenya is short of, or shrink a class roll on its own.
This guide is part of a wider series on how AI's usefulness shifts by curriculum and region; see AI in Education Around the World: A 2026 Regional Guide for the fuller picture. What makes this country's large-class problem distinctive is what CBC specifically asks to be done for each individual learner, not just how many learners are in the room.
Why Kenya's Large Classes Hit CBC's Assessment Model Hardest
A large class strains any curriculum, but CBC's design assumes a level of individual observation that a national-average class size makes genuinely difficult to deliver with fidelity.
From 8-4-4 to CBC: A Different Assessment Philosophy Entirely
Kenya replaced the exam-heavy 8-4-4 system with the Competency-Based Curriculum (CBC), rolled out from 2017 under a restructured 2-6-3-3 education model. Where 8-4-4 leaned on end-of-term and end-of-cycle exams to measure learning, CBC is built around continuous, competency-based assessment — judging whether a learner can actually do something, not just recall it on a single test day.
- 8-4-4 concentrated assessment into infrequent, high-stakes exams a teacher could mark in a single sitting
- CBC spreads assessment across the term, requiring ongoing observation of specific competencies as they develop
- That shift trades exam-marking volume for a different, more continuous kind of individual-tracking workload
Free Primary Education's Legacy: More Learners Than the System Was Built For
Kenya's Free Primary Education (FPE) policy, introduced in 2003, drove a rapid national enrollment surge as school fees stopped being a barrier for millions of families. That enrollment growth was a genuine access success, but classroom and teacher capacity didn't expand at the same pace — a gap CBC's more individually demanding assessment model now has to operate inside.
The TSC Teacher Shortage Behind the Numbers
The Teachers Service Commission (TSC) — the national body employing and regulating Kenya's public-school workforce — has repeatedly reported a shortfall running into the tens of thousands of posts, concentrated hardest in public primary and junior secondary schools. UNESCO Institute for Statistics has recorded national pupil-to-teacher figures sitting above policy targets in many counties, and — as with most national averages — the real picture in a specific fast-growing urban or peri-urban school often runs considerably higher than what the reported average alone suggests.
Regional Variation: Nairobi's Classrooms Aren't Kenya's Only Classrooms
A single national figure hides real variation between a crowded Nairobi estate school, a fast-growing peri-urban county, and a sparsely staffed school in an arid or semi-arid county.
- Nairobi and other major urban centers often carry the highest raw enrollment pressure per classroom, driven by rapid population growth outpacing new school construction
- Peri-urban counties surrounding major cities frequently see the fastest year-on-year enrollment growth as families relocate for work and affordable housing
- Arid and semi-arid counties more often face a straightforward teacher-supply shortage rather than sheer classroom overcrowding, with some schools running multi-grade combined classes
- A workflow built for "large classes" generally should still flex for which of these patterns actually describes a specific school
What CBC Actually Asks a Teacher to Track Per Learner
Before reaching for any tool, it's worth being precise about what CBC's assessment design actually demands, since a generic "large class" strategy borrowed from an exam-based system won't map cleanly onto it.
Strands, Sub-Strands, and Individual Competency Records
CBC organizes each subject into strands and sub-strands, with specific learning outcomes a teacher assesses through observation, projects, and portfolio evidence rather than a single test score. The Kenya Institute of Curriculum Development (KICD), which designs CBC's curriculum framework, expects a teacher to track where each learner sits on these outcomes across the term, not just record one summary grade at its end.
- Each strand breaks into sub-strands with specific, observable learning outcomes
- A teacher records evidence of a learner meeting an outcome — a completed project, an observed skill, a portfolio piece — not just a percentage score
- That evidence feeds into termly reporting under Kenya's competency-based reporting formats, which describe a learner's standing per strand rather than a single overall mark
- Each learner accumulates this evidence across a school career, meaning a poorly documented term isn't just a missed report card — it's a gap in the longitudinal record CBC's design assumes will exist
This design makes sense for its stated purpose: giving a genuinely detailed, longitudinal picture of what a learner can actually do, rather than a single number that hides which specific skills are solid and which aren't. That purpose doesn't disappear just because a class is large — if anything, a wide-ability large class is where knowing exactly which sub-strand a specific learner is struggling with matters most, since a single whole-class re-teach can't fix five different gaps for five different reasons.
Why a 50-Plus Learner Roll Makes Rigorous Observation Nearly Impossible Alone
Genuinely observing and recording individual progress across multiple strands for 50 or more learners, on top of actually teaching the lesson, is a workload CBC's design didn't originally scale for. This is where Kenya's large-class problem differs meaningfully from a purely exam-based system: the bottleneck isn't just marking volume, it's the sheer number of individual, ongoing observations a single teacher is expected to make and document.
Where AI Genuinely Helps With CBC's Assessment Load
AI's honest contribution here is structural: turning an unmanageable, freeform observation task into something fast enough to actually sustain across a full term with 50-plus learners.
Structured, Fast-to-Fill Observation Checklists
A checklist tied to a specific sub-strand's learning outcomes is far faster to complete during a lesson than an open-ended anecdotal note, and AI can generate one from a stated outcome in seconds.
- Generate a simple checklist per sub-strand — met, developing, not yet met — rather than free-text notes for every learner
- Ask for the checklist in a format printable on one page per class, so it's physically usable while circulating during a lesson
- Batch-generate a term's checklists at once, covering every sub-strand the term's scheme of work will touch
Competency-Aligned Task Variations at Multiple Entry Points
Say you teach a Grade 4 Science and Technology class of 58 learners working on a sub-strand about simple machines. Rather than one task for the whole class, you could generate two or three entry-point variations of the same hands-on task — all still assessing the identical competency, just at different levels of scaffolding.
- State the exact strand, sub-strand, and learning outcome from KICD's curriculum design in the prompt
- Ask for two or three task variations at different entry points, all assessing the same underlying competency
- Request a matching observation checklist for the specific outcome each variation targets
- Verify the generated task against KICD's actual curriculum design document before teaching it
Batch-Drafting Portfolio and Report Comments
CBC's termly reports typically call for short, individualized comments per strand rather than one overall grade — a genuinely time-consuming task multiplied across 50-plus learners and multiple subjects.
AI can draft a first-pass comment from a learner's recorded checklist results, which a teacher edits and personalizes — not an unsupervised, auto-generated report a parent receives untouched.
Keeping Parent Communication Manageable at Scale
CBC's design leans heavily on parental involvement in a learner's progress, which means regular, individualized parent communication isn't optional — but writing 50-plus personalized updates by hand every term is its own significant workload.
- Draft a first-pass parent note from each learner's checklist data, covering strengths and areas still developing, then personalize the specific details before sending
- Keep a consistent structure across notes — what's going well, what needs support at home, one concrete next step — so drafting stays fast even as content varies per learner
- Reserve fully personalized, handwritten notes for learners with the most significant needs, where a templated structure genuinely isn't sufficient
A Practical Workflow for a Large CBC Classroom
A term-long rhythm turns CBC's assessment demands from a constant background worry into a manageable, scheduled routine.
- Start of term — Generate the term's observation checklists for every sub-strand the scheme of work covers, printed and ready before teaching begins
- Weekly — Fill checklists during or immediately after lessons, rather than trying to reconstruct observations from memory later
- Mid-term — Review checklist patterns across the class, flagging sub-strands where many learners show "developing," not "met," for a targeted re-teach
- End of term — Batch-draft portfolio and report comments from accumulated checklist data, editing each for accuracy and personal detail before sharing with parents
- Between terms — Save which task variations and checklist formats worked well, building a reusable bank for the next cohort covering the same strands
| Assessment Approach | Weekly Time Cost | Individual Detail Captured | Best For |
|---|---|---|---|
| Freeform anecdotal notes | Very high | High, but often inconsistent | Small classes only |
| End-of-term summary grade only | Low | Very low — misses CBC's intent | Not CBC-aligned |
| Structured checklist per sub-strand | Moderate | Consistent, comparable across learners | Large CBC classrooms (40+) |
| AI-drafted comments from checklist data | Low-moderate | High, once checklist data exists | Termly reporting at scale |
Where AI Can't Fix Kenya's Large-Class Problem
Two limits deserve to be named plainly, since overselling AI's role here risks distracting from the staffing and access fixes large CBC classrooms actually need.
TSC's Staffing Gap Is a Policy Problem, Not a Workflow Problem
No checklist template changes how many teachers TSC can recruit, train, and post to a specific school. AI reduces the time cost of observing and recording competency evidence; it does not reduce the number of learners one teacher is responsible for, or add the additional trained staff a genuinely individualized assessment model was designed around.
Connectivity Across Kenya's Counties
Connectivity and device access vary sharply between Nairobi's better-resourced schools and rural counties in the arid and semi-arid regions, which shapes how much of an AI-assisted workflow can run live during the school day. Generating a term's checklists and task variations during a reliable connectivity window, then working from printed copies, fits this variation better than a workflow assuming constant live access.
- Batch-generate a full term's material ahead of time, using whatever connectivity window is available
- Keep printed checklists as the primary classroom tool, updating digital records only periodically
- Share verified checklists and task banks across teachers covering the same strand at a school, so the connectivity-dependent step happens once, not per classroom
A related self-directed practice pattern shows up in exam-prep contexts elsewhere in this series — AI Study Buddy for Matric (SSC) Board Exam Prep and AI for AKM and Asesmen Nasional Prep both cover how a different country's assessment system shapes AI-assisted practice.
Tools and Technology for CBC Classrooms
Matching the right tool to CBC's specific observation-and-reporting workload keeps it sustainable across a full term, not just a strong first week.
| Tool Type | Best For | Caution |
|---|---|---|
| General AI assistant (Gemini, ChatGPT, Claude) | Drafting a quick checklist or task variation on demand | Requires re-prompting each time; no memory of a class's accumulated checklist history |
| EduGenius | Generating tiered worksheets, quizzes, and answer keys from a class profile, exportable for printing | Strong for building reusable task-variation and practice banks across a term |
| KICD's official curriculum design documents | The authoritative reference for strands, sub-strands, and learning outcomes | Always the starting point before generating anything |
| Printed checklist sheets | Fast, offline-friendly observation recording during lessons | Needs periodic transfer into a digital or portfolio record |
Because it works from class-profile settings, a teacher could use EduGenius to generate task variations and answer keys at multiple entry points for a stated learning outcome, exportable for printing where live connectivity during lessons isn't reliable — a workflow suited to CBC's ongoing assessment demands rather than a single end-of-term test.
For a different curriculum framework entirely, AI Lesson Plans Aligned to IB is worth comparing, alongside AI for Teaching in Sesotho for how a different country's language-and-classroom pressures interact. And Best AI for Math Problems in 2026 (Benchmarked) is a useful companion before trusting any single tool with a term's worth of numeracy-strand content.
Pro Tips for AI-Assisted CBC Classrooms
- Generate checklists at the sub-strand level, not the whole subject. A checklist that's too broad stops being fast to fill during an actual lesson.
- Batch-generate a full term's material at the start, rather than building checklists and tasks week by week under time pressure.
- Review checklist patterns across the whole class, not just individual entries. A sub-strand where most learners show "developing" is worth a re-teach, not fifty individual interventions.
- Draft report comments from checklist data, always edited before sharing. A first-pass draft saves real time; an unedited one risks a generic-sounding report.
- Share a verified task bank with colleagues covering the same strand, since the same CBC assessment demands hit every parallel class, not just one.
What to Avoid
- Don't treat an AI-drafted report comment as final without a teacher's edit. Batch-drafted comments are a fast starting point, not an unsupervised final report.
- Don't build one checklist per subject instead of per sub-strand. Overly broad checklists are too slow to fill during an actual lesson to be useful.
- Don't skip verifying a generated task against KICD's actual curriculum design. A plausible-sounding activity can still miss the specific learning outcome it needs to assess.
- Don't build a workflow assuming constant connectivity. Batch-generate ahead of time and work from printed checklists where access is inconsistent.
Key Takeaways
- CBC's continuous, competency-based assessment model asks far more of a teacher's individual observation and recording than 8-4-4's exam-based approach did.
- Free Primary Education's 2003 enrollment surge and TSC's ongoing teacher shortfall mean CBC's individually demanding model often runs in classrooms larger than it was designed for.
- AI's real value is turning observation and reporting into structured, fast-to-fill checklists and draft comments — not an unsupervised replacement for a teacher's judgment.
- Task variations generated at multiple entry points, all tied to one learning outcome, let a large class practice the same competency at a level that fits each learner.
- AI cannot add the teachers TSC's own data shows Kenya is short of, or reduce the number of learners a single teacher is responsible for.
- A generate-once, work-from-print pattern fits the connectivity gap between Nairobi and many rural counties better than a workflow assuming steady access.
Frequently Asked Questions
How is CBC's assessment approach different from Kenya's old 8-4-4 system?
8-4-4 relied mainly on infrequent, high-stakes exams marked at set points in the year. CBC, rolled out from 2017, assesses learners continuously against specific competencies within each subject's strands and sub-strands, requiring ongoing observation and evidence collection rather than a single end-of-term score.
Can AI actually reduce class sizes in Kenyan schools?
No. AI cannot change TSC's teacher staffing levels or a school's physical capacity. What it can do is generate structured observation checklists, competency-aligned task variations, and draft report comments that make CBC's individually demanding assessment model more manageable within the existing class size.
What's the single highest-leverage change for a teacher overwhelmed by CBC's observation demands?
Generating a printable, sub-strand-level checklist before the term starts, rather than relying on freeform notes during or after lessons, tends to save the most time for the least setup effort — it turns an open-ended task into a fast, consistent one.
Does EduGenius help specifically with CBC-aligned content?
EduGenius isn't a CBC-specific product, but its class-profile system lets a teacher generate tiered task variations and answer keys for a stated grade, subject, and ability range — useful for the multiple-entry-point task design CBC's competency-based model rewards, though matching the exact strand and sub-strand still requires checking against KICD's curriculum documents.
Why does class size matter more under CBC than it did under 8-4-4?
8-4-4's exam-based model let a teacher assess an entire class with one marked test at set points in the year. CBC's continuous, competency-based model expects ongoing observation and evidence collection per learner throughout the term, which scales far less gracefully as class size grows, even though the total amount of "content" covered may be similar.
Sources
- Kenya Institute of Curriculum Development (KICD) — CBC curriculum design, strands and sub-strands.
- Teachers Service Commission (TSC) — national teacher shortage and staffing data.
- UNESCO Institute for Statistics (UIS) — pupil-teacher ratio data for Kenya.
- Ministry of Education, Kenya — Free Primary Education (2003) and CBC rollout (2017) policy background.
- Kenya National Bureau of Statistics (KNBS) — national enrollment and education data.