Winny AI builds educational AI ready to help learners in Burundi

Five priority programmes in Burundi’s Education Sector Plan

The Education Sector Plan 2022–⁠2030 identifies five priority programmes. These are planned priorities, not achieved coverage or Winny outcomes. The thematic excerpts below are translated and adapted from French.

• Access and equity

  Local learner support and content would be agreed with education partners; the tutor does not remove financial or infrastructure barriers.

  Quote and source

  “improving access and equity”

  Education Sector Plan 2022–⁠2030, printed p. 23 — thematic excerpt translated and adapted from French

• Teaching quality

  Guided questions and teacher-linked progress review can complement teaching, with local evaluation rather than guaranteed national improvement.

  Quote and source

  “quality of teaching”

  Education Sector Plan 2022–⁠2030, printed p. 23 — thematic excerpt translated and adapted from French

• Higher education and research

  Proposed connections with university resources require the operator’s permission, content rights and security and legal review.

  Quote and source

  “scientific research”

  Education Sector Plan 2022–⁠2030, printed p. 23 — thematic excerpt translated and adapted from French

• Technical and vocational pathways

  Guided understanding and exploration of interests can support learning choices, without a job outcome or accredited qualification promise.

  Quote and source

  “technical and vocational training”

  Education Sector Plan 2022–⁠2030, printed p. 23 — thematic excerpt translated and adapted from French

• Educational governance

  Responsible partners can agree project objectives and review authorised learning information; no completed national data platform or API is assumed.

  Quote and source

  “sector governance”

  Education Sector Plan 2022–⁠2030, printed p. 23 — thematic excerpt translated and adapted from French

Learning support for fondamental and post-fondamental education Education Sector Plan 2022–⁠2030, printed pp. 22–⁠24

An AI tutor to support learning

Education research has long known the most effective way to learn: one-to-one work with a tutor. Until now, only a few could have it — AI opens it up to every learner.

Regular classroomWith a personal tutorahead of 98% of peers

98%

On average, a student with a personal tutor outperforms 98% of peers in a regular classroom.

Bloom B. S., The 2 Sigma Problem, Educational Researcher, 1984

1.5 years

Six weeks of sessions with an AI tutor raised results by about 0.3 standard deviations — roughly a year and a half of typical learning.

De Simone et al., World Bank, randomized study, Nigeria, 2025

×2

Students with an AI tutor learned more than twice as much in less time than in active-learning classes.

Kestin et al., Harvard University, Scientific Reports, 2025

+1 pp

A gain of 0.5 standard deviations in learning outcomes is associated with about 1 percentage point faster long-run growth of GDP per capita each year.

Hanushek & Woessmann, OECD, 2010

These are independent studies of tutoring (by people and by AI) and of the economics of education. We propose to measure the impact of Winny AI Tutor together with the country — in a pilot with independent evaluation.

Flagship product of Winny AI

Winny AI Tutor — an international learning ecosystem

A personal AI tutor in a chat format for learners in fondamental and post-fondamental education — across school subjects, foreign languages and hobbies — with connected accounts for parents, teachers and schools.

For the learner

• Leads to understanding

  No finished solutions: guiding questions and step-by-step hints.

• The country’s curriculum and textbooks

  A Burundi setup would agree and validate curricula, textbooks, terminology and assessment requirements with local education partners.

• Beyond lessons

  Answers difficult questions thoughtfully, in a way that suits the learner’s age, and handles sensitive topics with care.

• Encouragement

  Gently supports learners and helps ease stress and anxiety — for example, before exams.

Phone, tablet and computer — in the browser, nothing to install. Voice, photo and text.

Four roles. One ecosystem

• Learner

  A personal tutor around the clock, a learning path, goals and assignments.

  Try it

• Parent

  Their own account and chat: the child’s progress, strengths, gaps and talents.

  Try it

• Teacher

  An AI assistant, class analytics, assignments and groups. Free, with generous limits.

  Try it

• School and government

  Administration, aggregated insights across schools and the whole system, and deployment within the country.

  Request a presentation

Core features are free forever. Free for teachers, with generous limits.

For every question — the best-suited model and real pedagogy, not just an answer

Winny AI’s own technology: a model orchestrator, a pedagogical agent, a safety layer and long-term memory of each learner. The product does not depend on any single AI provider.

• Model orchestrator

  For each request it chooses the best-suited model, including national ones, and tunes its behaviour to the subject, language and task. A model can be replaced without rewriting the product.

• Pedagogical agent

  Plans every reply to lead the learner to understanding: guiding questions and step-by-step hints instead of a finished answer. It holds firm when a learner tries to coax out the answer, and this mode cannot be switched off.

• Safety layer

  Checks every answer for child safety and for correct references to learning materials.

• Learner’s digital twin

  A competency profile and long-term memory: Winny AI Tutor remembers years of learning — what the learner already understands, where the gaps are, what their talents are and which goal they are working toward.

• Ontological interdisciplinary graphs

  School subjects are connected in a single knowledge graph built from textbooks and approved curricula, so the tutor sees the links between topics across subjects.

• Bloom’s taxonomy

  Every topic is broken down into levels of mastery following Bloom’s taxonomy: learners climb the ladder of thinking instead of collecting ready answers.

Every update goes through a pedagogical quality loop: a test bench with synthetic AI learners and quality scales. The orchestrator and the educational ecosystem are registered as intellectual property.

Global technology. National character

Winny AI Tutor adapts to the country — not the other way around.

• Curriculum and assessments

  A local setup would agree curriculum references, textbooks and assessment requirements with education partners for fondamental and post-fondamental learning and guided school-assessment practice.

• Language and culture

  Explanations and responses in Kirundi, French and English, among more than 90 supported languages. Local terminology and content need review. Tutor language is separate from product interface language; coverage of every teaching language or accessibility need is not assumed.

• Models

  National, regional and open-weight models can be considered with the relevant permissions and agreed educational requirements.

• Infrastructure

  The service and data can stay in the country, in a closed environment.

• Transparency

  A “glass box”, not a “black box”: regulators can check how answers are built and what they contain.

A possible path for learning support in Burundi

Learners and schools in more than 80 countries already use Winny AI Tutor. These are global users, not a claim of deployment in Burundi. A local project would agree its scope, content and pace with responsible education partners.

1. First learners and families

   Learners and parents can start with Winny AI Tutor today: core features are free forever. Teachers join free, with generous limits; training workshops for teachers.

   Try Winny AI Tutor

2. Pilot project

   A possible school project with agreed scope and independent impact evaluation; no existing pilot in Burundi is claimed.

3. National setup

   Curriculum, assessment, language and deployment requirements would be agreed and reviewed for the planned use.

4. Scaling to the whole country

   Expansion can be discussed after local evaluation, permissions and infrastructure requirements have been assessed.

From a school tutor to a lifelong mentor

We are developing an educational AI ecosystem for lifelong learning: from first words, school and career guidance to university, a profession and upskilling.

1. First steps Developing with partners

   A smart speaker for children: learning by voice from the very first words.

2. School Working today

   School subjects in fondamental and post-fondamental education: guided assessment practice, talent discovery and career guidance.

   Try Winny AI Tutor

3. Colleges and universities Developing with partners

   Admission, choosing a programme and support for students throughout their studies.

4. Further education Developing with partners

   Hobbies, new skills, upskilling and reskilling — for children and adults.

5. Lifelong mentor Developing with partners

   Professional growth and career throughout life.

Career guidance that looks ahead

Every learner’s talents and an informed choice of career

The world and the job market are changing fast. Winny AI Tutor sees a learner’s strengths, interests and talents in every subject and helps them choose a direction. The next step is to take labor market trends, new technologies and future professions into account.

• Talents in every subject Working today

  The competency digital twin shows a learner’s strengths, gaps and inclinations in each subject.

• Career guidance Working today

  Recommendations on fields of study based on the learner’s strengths and interests — to support their own decision about the future.

• Atlas of future professions Developing with partners

  Together with experts and universities, we are developing an atlas of future professions so that career guidance reflects labor market trends and new technologies.

• Neuroprofiling Developing with partners

  We work with experts in neuroprofiling to prepare diagnostics of talents and inclinations, including with neural interfaces — voluntary and with parental consent.

For education authorities: learners’ strengths and interests can inform discussion of fields of study and educational planning.

Winny AI · AI-Native Enterprise

A global EdTech company of a new kind

Winny AI builds AI solutions for personalised learning, with an ethical approach to adopting artificial intelligence and respect for national education standards.

A team of experts and hundreds of AI agents

Alongside our team, spread across the world, hundreds of AI agents are at work: they build the company together with us — which is how the team gets so much done.

• R&D
• Product
• Methodology
• Sales
• Customer experience

• Technology

  The founder has more than 14 years in SaaS and product development, more than 8 of them in senior leadership (C-level); he led the technology function of an international IT service with more than 30 million users.

• Education

  Our own methodology department; the methodology was developed with teachers and curriculum specialists.

• AI ethics and regulation

  Experts in the law and regulation of AI in education on the team; our own AI governance charter.

• International work

  We speak at international conferences and round tables on AI in education; nominee for the UNESCO Prize for the Use of ICTs in Education.

• Leadership

  Heads of functions, including holders of doctoral degrees, with experience at big tech and leading EdTech companies: strategy, finance, sales, product, methodology, customer experience.

Around the team — dozens of strategic partners worldwide and academic collaborations.