Using AI to Cut CBC/CBE Assessment Paperwork

Say it's the last Friday of term, and you still have rubric scores for 45 learners across a dozen sub-strands waiting to become actual written comments a parent can read. AI can turn a scored rubric level into a clear, individualized comment draft in seconds — the scoring judgment stays yours, but the sentence-writing that eats an entire weekend doesn't have to.

Quick Answer: Kenya's Competency-Based Curriculum (CBC) requires continuous, rubric-based assessment across many learning-outcome strands per subject, which generates far more individual scoring and reporting entries than the old exam-only system did. AI can draft rubric-aligned comments, summarize a term's scores into a parent-facing narrative, and build reusable comment banks by strand — cutting the writing time without changing what a teacher actually scores.

CBC (Competency-Based Curriculum) is Kenya's specific national curriculum reform, rolled out progressively from 2017 to replace the earlier 8-4-4 system; CBE (Competency-Based Education) is the broader pedagogical category CBC belongs to, shared conceptually with similar reforms in Rwanda, Uganda, and elsewhere. This guide uses "CBC" for Kenya-specific detail and "CBE" when the point applies more generally.

For the wider regional picture, see AI in Education Around the World: A 2026 Regional Guide, and for a deeper look at AI-assisted entry-test prep in a neighboring exam system, see AI for ECAT and Engineering Entry Tests.

What CBC Assessment Paperwork Actually Involves

CBC assessment isn't one report card — it's a stack of ongoing, strand-level records that accumulate across a term and then have to be synthesized into something a parent or the next teacher can actually use.

Rubric-Based Continuous Assessment

The Kenya Institute of Curriculum Development (KICD) built CBC around continuous formative assessment using a four-level rubric — commonly Exceeds Expectation, Meets Expectation, Approaches Expectation, and Below Expectation — scored per learning outcome, not just per subject. A single subject can carry several strands, each broken into sub-strands, each requiring its own rubric entry across the term.

Learner Portfolios

CBC asks teachers to maintain a portfolio of learner work samples as evidence supporting rubric scores, rather than relying on a single end-of-term test. Compiling, labeling, and periodically reviewing that evidence is its own ongoing administrative task layered on top of the scoring itself.

Parent-Facing Progress Reports

Rubric scores eventually have to become a report a parent can read and understand — typically a narrative summary per subject, not just a list of levels. Translating a column of "ME, EE, AE" entries into clear written comments, learner by learner, subject by subject, is usually where the paperwork burden is felt most acutely.

Why This Became Such a Heavy Lift for Teachers

CBC's assessment design multiplies the number of individual entries a teacher records compared to the exam-only system it replaced, and everyday classroom realities in Kenya compound that multiplication.

Per-Learner-Per-Strand Scoring at Scale

A teacher with 40-50 learners, several subjects, and multiple sub-strands per subject can easily be managing hundreds of individual rubric entries across a single term — a volume that scales directly with class size, which in many Kenyan public primary schools runs well above what a rubric system of this granularity was designed to comfortably support.

• Kenya National Union of Teachers (KNUT) and Kenya Union of Post Primary Education Teachers (KUPPET) have both repeatedly raised assessment workload as a top implementation concern in public commentary on CBC.
• Usawa Agenda, a Kenyan citizen-led education-monitoring initiative, has documented uneven CBC implementation conditions across different school types and regions.

The Paper-vs-Digital Divide

Kenya's Digital Literacy Programme pushed device access into many schools, but rubric recording in a large share of classrooms — particularly rural and lower-resourced ones — still happens on paper first, meaning a teacher does the scoring twice: once on paper, once when it's transferred into whatever system generates the final report.

Where Generic AI Prompts Fall Short for CBC Specifically

A general AI assistant asked to "write a report comment" has no built-in sense of CBC's rubric levels, strand structure, or competency-based phrasing conventions, and will default to generic praise-or-concern language that doesn't map cleanly onto an Exceeds/Meets/Approaches/Below framework. Naming the exact rubric level and strand explicitly in every prompt closes most of that gap.

Assessment Time Competing With Teaching Time

Every hour spent converting rubric levels into written comments is an hour not spent on lesson preparation or actual instruction — a trade-off that becomes more visible as class sizes and strand counts grow, and one AI-assisted comment drafting is specifically positioned to reduce.

A Step-by-Step Way to Use AI to Lighten It

Building an AI-assisted CBC reporting workflow works best as a repeatable routine tied to how scoring already happens, not a separate system layered on top.

1. Keep your rubric scores in whatever format already works — a spreadsheet, a notebook, or your school's existing system — AI drafting works from whatever scores you already have.
2. Build a comment bank per strand first, generating three or four sample comments for each rubric level (Exceeds, Meets, Approaches, Below) so you're not starting from a blank page for every single learner.
3. Personalize the bank's phrasing per learner using specific portfolio evidence — a particular piece of work, a specific moment of progress — rather than leaving comments fully generic.
4. Generate a draft parent-facing summary from a learner's full set of term scores, then review and adjust the tone and specifics before it goes home.
5. Save strong comment drafts back into your bank, so the collection grows richer and more personalized-sounding across terms rather than starting over each time.
6. Keep the actual rubric judgment entirely yours. AI drafts the sentence; you decide the level a learner actually earned, based on the evidence in front of you.

Building a Reusable Rubric-Comment Bank

A comment bank is simply a running set of well-phrased comment templates organized by strand and rubric level, which you adapt rather than rewrite from scratch each time. Building it once per subject at the start of a term — rather than writing every single comment fresh — is where most of the real time savings comes from across an entire term's reporting cycle.

Pro tip: Ask for comments phrased around what a learner can now do, not just their rubric level — CBC's competency framing works best when parent-facing language mirrors it, rather than reading like a converted test score.

A Worked Example: One Strand, Four Levels

Say a Grade 5 teacher is reporting on a numeracy strand — working with fractions — and needs a starting comment for each rubric level before personalizing. A generated bank might look like this:

A teacher then adds one specific detail per learner — a particular piece of portfolio evidence, a specific moment of growth — turning a generic starting point into a comment that actually reflects the individual learner it describes.

The same approach scales across a full subject roster: once the four-level bank exists for one strand, generating the equivalent set for the next strand — decimals, or measurement, or whichever comes next in the term's scheme of work — takes a few minutes rather than starting the drafting process from zero again.

Keeping the Workflow Sustainable Term After Term

A comment bank built once and never revisited eventually feels as generic as no bank at all — the workflow only stays useful if it's maintained deliberately.

Reviewing and Retiring Weak Comments

Not every AI-drafted comment earns a permanent place in the bank. After each term, it's worth a quick pass to retire comments that felt generic or didn't land well with parents, replacing them with better-performing alternatives drafted along the way. A bank that only ever grows, without this occasional pruning, eventually becomes as hard to search as having no bank at all.

Sharing Comment Banks Across a Subject Team

Where a school has more than one teacher covering the same subject and grade band, a shared comment bank — built collaboratively rather than by one teacher alone — spreads the initial drafting effort across more people. It also tends to produce a richer, more varied set of starting phrases than any one teacher would build solo.

A shared spreadsheet or folder works fine for this; the tool matters far less than simply agreeing, as a team, to add to the same bank rather than each person rebuilding an equivalent one privately.

Getting School Leadership on Board

Because CBC's assessment paperwork is a school-wide load, not just an individual teacher's problem, raising an AI-assisted comment-bank workflow with school leadership — as a shared resource rather than a personal workaround — can turn a single teacher's time-saver into a school-wide practice that benefits every subject team.

Framing the conversation around the specific pain point — hours spent converting rubric levels into parent-facing language every term — tends to land better with school leadership than a general pitch about "using AI," since it ties the workflow directly to a workload problem administrators already recognize.

Tools & Technology Comparison

EduGenius can generate targeted practice material — a worksheet or revision set — for a specific strand a learner is scored "Approaches Expectation" on, turning an assessment finding directly into a next instructional step rather than letting the rubric entry sit as a static record.

What This Looks Like Across the CBC Grade Bands

Assessment paperwork volume and type shift across CBC's structure, so an AI-assisted workflow that fits Upper Primary doesn't automatically fit Pre-Primary.

Upper Primary is typically where the paperwork load peaks, simply because the number of subjects and strands per learner is at its highest before Junior Secondary's subject specialization spreads the load across more teachers.

Handling the Volume Without Losing Personalization

The real tension in CBC reporting isn't generating one good comment — it's generating dozens of good ones without the whole set starting to sound identical.

Grouping by Rubric Level First

Rather than working learner by learner, sort your rubric scores by level first — every "Meets Expectation" fractions score together, every "Approaches Expectation" one together. Generating one strong draft per group, then personalizing each copy, is faster than starting fresh for every single learner and produces more consistent quality across the set.

Rotating Sentence Structures Deliberately

A comment bank that only offers one phrasing pattern per level tends to produce reports that read as obviously templated once a parent compares notes with another family. Ask for two or three structurally different draft options per rubric level — one leading with strength, one leading with a growth area, one leading with a specific skill — and rotate between them.

Setting a Realistic Personalization Bar

Not every comment needs a unique, deeply specific sentence — a genuinely useful personalization bar is one concrete detail per comment (a specific piece of work, a specific moment of progress) layered onto a well-chosen template. Chasing full uniqueness for every single entry is where the time savings disappear.