In Ghanaian conditions the strongest return is in reducing teacher and administrative workload — lesson planning, marking, School-Based Assessment records, fee reconciliation and family communication. Claims about direct learning gains from AI tutoring depend heavily on device access and supervision, which is exactly what is least reliable in most schools.
Start from the constraint, not the technology
Most international writing on AI in education assumes a device per learner, reliable electricity and home broadband. A Ghanaian school planning around those assumptions will buy something that works in the demonstration and not in the third week of term.
The useful question is narrower: which AI applications produce value under intermittent power, shared devices, uneven connectivity and class sizes of 45 to 60? That question has clear answers, and they are mostly not the ones being marketed.
Where AI genuinely helps in Ghanaian schools today
- Lesson planning. Drafting against a NaCCA indicator, differentiating for a large mixed-ability class, generating retrieval questions. Teacher edits and approves. Runs on one device in the staff room.
- Marking and feedback. First-pass scoring against a rubric with the teacher moderating.
- School-Based Assessment records. The single largest administrative burden the new assessment model creates, and the one where automation helps most — see SBA.
- Terminal reports. Assembling assessment into reports with genuine comments rather than recycled ones, reclaiming days at the end of every term.
- Fee reconciliation. Matching mobile money and bank transfers to learner accounts, which in most schools is manual, error-prone and a source of family disputes.
- Family communication. Drafting, translating and routing across SMS, app and voice.
- Early warning. Flagging attendance and performance patterns that predict a learner dropping out or a family falling into arrears.
Notice what these have in common. They are all adult-facing, and none requires a device per learner.
Where the evidence is thinner than the marketing
- AI tutoring as a substitute for teaching. Results depend enormously on supervision and device access. A product that works in a supervised ICT laboratory session frequently does nothing when pushed to shared phones at home.
- Automated essay scoring at high stakes. Adequate for formative feedback; not reliable enough to certify, and performance degrades on Ghanaian English usage in ways that raise fairness questions.
- Predictive risk models. Useful as a prompt to look, dangerous as a verdict. A model trained on historical outcomes reproduces historical disadvantage.
- AI detection tools. False positive rates are high enough that they should never be the sole basis of a malpractice finding.
Five risks a Ghanaian school must manage
1. Learner data leaving the school
The first failure is almost never the procured system. It is staff pasting learner names, results or health information into a consumer chatbot with no agreement behind it. Under the Data Protection Act 2012 the school remains the data controller regardless. Policy and a sanctioned alternative must arrive together — a ban without an approved tool produces hidden use, not compliance. See Act 843.
2. Assessment integrity
If a project or take-home assignment can be completed by a model in thirty seconds, the mark certifies the model. This is an assessment design problem before it is a policing problem, and it applies particularly to the portfolio and project components of the new senior high model.
3. Equity within the school
Any tool that works well only on a recent smartphone with data will under-serve exactly the learners whose outcomes the school is most anxious about. A homework platform requiring data a family cannot afford widens a gap rather than closing one.
4. Deskilling
If a newly posted teacher never plans a lesson unaided, planning expertise does not develop. Systems should make the pedagogical reasoning visible rather than hiding it behind a generate button.
5. Cost in cedis, benefit in promises
AI features are frequently priced in dollars and sold on outcomes never measured in a Ghanaian school. Ask what it costs in cedis over three years, who carries exchange risk at renewal, and which existing line item goes down.
An adoption sequence that works
| Stage | Focus | Typical duration |
|---|---|---|
| 1 | Acceptable use policy, data protection position, DPC registration check, staff briefing, proprietor sign-off | 2–4 weeks |
| 2 | Administrative load: SBA records, terminal reports, fee reconciliation, family communication | One term |
| 3 | Teacher workload: lesson planning, marking, coverage reporting | One to two terms |
| 4 | Assessment redesign for tasks a model can complete | One academic year |
| 5 | Learner-facing tools, supervised, with age-appropriate limits | Ongoing |
Schools that invert this — starting with learner-facing AI because it demonstrates well to families — typically spend the following year retrofitting governance under pressure.
Questions to ask any AI education vendor in Ghana
- What instructional model does your output follow, and can I see it?
- Is content mapped to NaCCA learning indicators, and at what level?
- Is learner data used to train models? Show me the contract clause, not the marketing page.
- Where is our data stored and processed, and does it leave Ghana? What is your position under Act 843?
- Are you registered with the Data Protection Commission?
- What happens when the power goes and when the network drops?
- What is the total cedi cost over three years, including SMS and data charges?
- What happens to our data if we leave, and what does export cost?
The policy position
Ghana’s Ministry of Education has signalled interest in integrating AI and digital learning into the education sector, and the Data Protection Commission has moved toward enforcement of Act 843. There is not yet a single settled national rulebook for AI use in schools specifically, which leaves proprietors, districts and boards setting their own rules.
What has not changed is that data protection law applies to AI systems exactly as to any other processing, and that a school remains the controller of its learners’ data whatever tool it uses.