Personalized Learning With AI: Free Tools for CBC Schools

Personalized learning with AI means software that gives each learner practice at their own level, checks every answer instantly and shows the teacher who is stuck on what. A Kenyan school can start this term without buying anything: several capable tools are free at the time of writing, and some run without internet. The hard parts are device time, teacher follow-up and getting parents’ consent before any pupil’s data goes into an app.

Why it matters: in a class of 45 with a 40-minute lesson, 40 x 60 / 45 = about 53 seconds per learner, even if the teacher does no whole-class teaching at all. Software can do the diagnosing; the teacher still does the teaching.

Key Takeaways

• Start with free tools for teachers first, because most AI tutors are either teacher-only outside the US or restricted for under-18s.
• Use personalised practice in Grades 7 and 8 and in senior school, where school-based assessment counts towards the final result.
• Plan device time with arithmetic and a buffer: schedule for 80% attendance, not 100%.
• Get written parental consent and run a data protection impact assessment before pupils log in to any AI tool.
• If your connection is weak, pick an offline-first platform or zero-rated content rather than waiting for better internet.

What does personalised learning with AI actually mean in a CBC classroom?

It means three things happening at once. The learner gets a question pitched at their current level. The system marks it immediately and decides what comes next: a harder item, an easier one, or a short explanation. The teacher gets a summary showing which learners have mastered which competencies.

That is very different from “the children use ChatGPT”. A general chatbot will answer anything, including the homework, and keeps no record a teacher can act on. A proper personalised learning setup has a curriculum behind it, a teacher view and limits on what the learner can do.

There is also a quieter version that needs no fancy software. A teacher uses an AI assistant to produce three versions of the same worksheet (support, core and stretch) and assigns them by group. Our view: for most Kenyan public schools, this teacher-side use is the honest starting point, because it works on one teacher’s phone and touches no pupil data at all. We cover the teacher skills in AI literacy for teachers: a practical curriculum.

How does CBE assessment change the case for personalised practice?

Under the old system, almost everything rode on one national exam at the end. Under Competency Based Education, continuous school-based assessment carries real weight, so gaps left unfixed in Grade 7 cost marks later.

The numbers are worth knowing:

• Junior school (KJSEA). For the 2025 KJSEA, each learner’s final grade combined 20% from KPSEA, 20% from school-based assessment in Grades 7 and 8, and 60% from the Grade 9 summative assessment, as reported by The Star. Scores sit on an eight-point scale per subject, with 90-100% earning 8 points and 75-89% earning 7.
• Senior school. KNEC’s January 2026 statement says Grade 10 school-based assessments are a blend of projects, practicals and written tests, with projects and practicals between May and July 2026 and the written test in October 2026. It adds that formative data from Grades 10, 11 and 12 will form part of the Kenya Certificate of Basic Education results, and that KNEC will pilot the senior school summative assessment in 2027 (KNEC CEO statement, Ministry of Education).
• Secondary under 8-4-4. KNEC says the 2027 KCSE will be the last under the 8-4-4 system (The Standard). Schools still have Form 3 learners to carry through.

What this means in practice: a Grade 8 learner who never really understood fractions is not just weak in maths. That weakness shows up in a school-based score that counts for 20% of the KJSEA result, before Grade 9 even begins. Personalised practice earns its place by catching that gap early, when fixing it is cheap.

Which AI learning tools can Kenyan schools actually use in 2026?

Fewer than the marketing suggests. Several famous tools have restrictions that matter in Kenya. Here is what we verified, at the time of writing:

Two notes on the table. First, the evidence base. Rising Academies reports a 0.3 standard deviation effect on maths learning from school-based studies by University of Oxford and J-PAL researchers (Rising Academies case study). J-PAL’s newer Ghana pilot gives students Rori for one hour a week (J-PAL). That is a useful benchmark for how much time to plan.

Second, general chatbots. OpenAI’s terms for users outside the US require users to be at least 13, and anyone under 18 needs a parent or legal guardian’s permission (OpenAI Terms of Use). A school that tells Grade 7 learners to “use ChatGPT for revision” without consent forms is ignoring the vendor’s own rules.

For a wider look at which tutoring tools are worth the effort, read AI tutors for Kenyan classrooms: what works and what is hype.

How much does AI personalised learning cost a Kenyan school?

The software can cost nothing. The real costs are devices, power, data and teacher time, and these are where plans fail.

Worked example (hypothetical): a public junior school in Kakamega with 120 Grade 8 learners and 20 working devices (old laptops plus donated Android tablets). The deputy wants every learner to get one hour a week of personalised maths practice, matching the J-PAL pilot dosage.

• Device-hours needed: 120 learners x 1 hour = 120 device-hours a week.
• Hours the devices must run: 120 / 20 = 6 hours a week.
• Per school day: 6 / 5 = 1.2 hours a day.

Now apply a buffer. Assume one session in five is lost to power cuts, absence or a sports day. To deliver 6 hours, schedule 6 / 0.8 = 7.5 hours a week, or 1.5 hours a day. Plan for 1.5, not 1.2. If the timetable cannot spare that, cut to one stream or one subject rather than spreading thin sessions across everyone.

Senior school has the same squeeze. Learners have 40 lessons a week of 40 minutes each (People Daily), so 40 x 40 = 1,600 minutes, or about 26.7 hours. Two hours of protected practice is 2 / 26.7 = about 7.5% of the week. That is a decision for the timetable committee, not something to squeeze in after games.

Teacher time is the hidden cost. A dashboard nobody reads is worth nothing. Budget at least one 30-minute slot a week per subject teacher to review results and regroup learners.

Can AI personalised learning work without reliable internet?

Yes, but you have to choose for it from the start. Three routes work in Kenyan conditions:

1. Offline-first platform. Kolibri is built for teaching “without the Internet” and runs on low-cost and older devices once installed on a local server, or on Android 6 and newer (Learning Equality). Content is loaded once and learners work on the school network.
2. Zero-rated content. Airtel made the Kenya Education Cloud free of data charges from 5 October 2025, though zero-rating usually covers only the main whitelisted sites, not external links or large downloads (Tech-ish). It only helps where Airtel has coverage.
3. Teacher-side AI only. The teacher prepares differentiated work on their own phone when they have signal, then prints or writes it up. Learners need no device.

Our opinion: route 3 is underrated. It gives most of the benefit of differentiation at almost none of the cost, and it is where we would start any school in Turkana or rural Kitui before buying a single tablet.

If your school serves learners with special needs, the same tools need extra care; see AI for special education and inclusive learning.

What does the Data Protection Act require when pupils use AI tools?

Most learners are children, and Kenyan law treats their data with extra care. The Office of the Data Protection Commissioner’s Guidance Note for Processing Children’s Data spells it out:

• A child is anyone under 18.
• Section 33 of the Data Protection Act requires consent from a parent or guardian, and processing must protect and advance the child’s best interests.
• Before processing children’s data that is likely to be high risk, including profiling, organisations must carry out a Data Protection Impact Assessment under Section 31.
• Regulation 13 of the General Regulations prohibits profiling a child for direct marketing.
• Entities that process children’s data must register with the ODPC as data controllers or processors.
• Parents should be able to view their child’s data and withdraw consent.

The ODPC even gives a worked example of a compliant learning app: it verifies age and gets parental consent before collecting data, encrypts it, keeps it only as long as needed for education, and lets parents request deletion.

A personalised learning tool profiles learners by design. That is the whole point of it. So treat it as high-risk processing: do the impact assessment, read the vendor’s terms on where data is stored and whether it trains their models, and send parents a plain consent letter. Our article on parent engagement through technology in Kenyan schools covers how to communicate this without alarming families.

What can your school do this term?

Here is a plan you can start this week. None of it needs a purchase.

1. Pick one grade and one subject. Grade 7 or 8 maths is the obvious choice, because school-based scores there count towards KJSEA.
2. Count real devices. Switch on every laptop, tablet and phone the school owns. Count the ones that work for 40 minutes on a charge, not the ones on the asset register.
3. Run the arithmetic. Learners x weekly hours / working devices = hours the devices must run. Divide by 0.8 for the buffer. If the answer will not fit the timetable, shrink the group.
4. Sign up two teachers for a free teacher tool. Khanmigo for Teachers or, if you run Google Workspace for Education, Gemini in Classroom. Have them produce support, core and stretch versions of one week’s maths work.
5. Draft the consent letter and impact assessment. Use the ODPC children’s data checklist. Name the tool, what data it collects, where it is stored and how parents can withdraw.
6. Set up offline content if needed. Install Kolibri on one machine as a trial, or test the Kenya Education Cloud on an Airtel line.
7. Run a two-week pilot and review every Friday. Check three things: who completed sessions, which competencies remain weak, and who missed two or more sessions.
8. Decide at the end of term. Compare pilot learners’ school-based scores with the previous term. Expand, adjust or stop, based on what you saw.

For more ideas on where AI fits across a school, see how AI is used in schools: practical examples for Kenyan schools.

What do Kenyan parents and teachers actually look for online?

Specific, local answers. We built cbcedukenya.com, an education website for Kenya, from a standing start: new domain, no backlinks, no content, Domain Authority 1. We published steadily on real questions Kenyan parents and teachers search. After five months, in the week of 16-22 August 2026, Google Analytics showed 6.1K active users, 10K page views and 6.8K sessions, most of it from Kenya.

The lesson we took: Kenya-specific queries with modest volume beat broad global keywords. The same applies when you explain a pilot to parents. Talk about their child’s CBC learning areas and grades, not “adaptive technology”.

Mistakes to avoid

• Promising parents an AI tutor that is not available here. Khanmigo’s learner tutoring is US-only. Check access for Kenyan learners before announcing anything.
• Letting under-18s use general chatbots without consent. It breaks the vendor’s terms and the Data Protection Act’s consent rule.
• Buying devices before testing the software. Run the tool on one device you already own first.
• Scheduling for perfect attendance. Without the buffer, you will fall short every week and blame the tool.
• Treating the dashboard as the intervention. The data only helps if a teacher regroups learners and reteaches the weak competency.

Frequently asked questions

Is there a free AI tutor for Kenyan students?

There are free tools, but few full AI tutors open to Kenyan learners. Microsoft Reading Coach is free for reading practice, and Rori offers maths tutoring on WhatsApp in Kenya, though its pricing is not published. Khanmigo is free for teachers here, but its learner tutoring is limited to the US.

Can a school let Grade 7 learners use ChatGPT?

Only with care. OpenAI’s terms require users to be at least 13 and to have a parent or guardian’s permission if under 18, and Kenyan law requires parental consent for processing children’s data. Check each learner’s date of birth first, because under-13s cannot use it at all.

Does AI personalised learning replace school-based assessment?

No. KNEC sets and administers school-based assessments, and the scores must be uploaded to the KNEC database. Personalised practice helps learners prepare for those assessments; it does not produce the official score.

What if our school has no internet at all?

Use an offline-first platform such as Kolibri on a local server, or keep AI on the teacher’s side and print differentiated work. Both avoid daily data costs. The trade-off is that content must be updated by hand when the teacher next has a connection.

Do we need to register with the ODPC to use these tools?

The ODPC’s children’s data guidance says entities that process children’s data must register as data controllers or processors. A school collecting learner records is already processing children’s data, so check your registration status before adding new tools. A Data Protection Impact Assessment is also expected for high-risk processing such as profiling.

Which subjects benefit most?

Subjects with clear right and wrong answers, such as maths and early reading, suit automated practice best, which is why Rori focuses on maths and Reading Coach on reading fluency. Essays, projects and practicals still need a teacher’s judgement, and those make up a large part of CBE school-based assessment.

The bottom line

AI will not fix a 45-learner class on its own. What it can do, cheaply and this term, is give teachers a fast way to produce differentiated work and give some learners practice at their real level. Start small, with free tools, honest device arithmetic and proper consent from parents. Expand only when your own school-based scores show it is working.

If you would like a second pair of eyes on your school’s plan, or want to start with our free AI Business Assessment, we are happy to talk it through. WhatsApp us on 0711 344 702.