From Lecturer to Conductor: The Teacher Is Being Promoted

A Globe, a Child, and a Finger

Imagine a classroom in Kenya with more children than desks. This is a composite, imagined for this essay; no real teacher, school or pupil is described. The chalk is nearly gone. The teacher turns a globe on her table so that a girl in the third row can find her own country. The girl's finger lands on it, and for a moment the whole world is the width of a finger.

I keep returning to that scene when people ask whether machines will replace teachers. The question usually arrives as a door: either the teacher stays in the room or she walks out of it. The door is the wrong picture. The globe shows a person whose job is to point, to turn the world toward a child at the right angle, and then to wait while the child finds the place with her own hand.

This essay is the third branch of the founder's Real Time Learning theory, which holds that schooling must keep changing as the models behind it change. It concerns the person at the front of the room. The claim is a judgment, not a fact: the teacher is not being replaced. She is being promoted from lecturer to conductor. What follows weighs the evidence for that claim and sketches what the new job would involve.

This year's World Teachers' Day theme, "Standing with teachers," marks 60 years of "protecting the profession, advancing its status and shaping our futures." The UNESCO teachers page carries the theme and a 2026 fact sheet.

Fifty Million Empty Chairs

The headline figure comes from UNESCO's 2026 World Teachers' Day release. The world, it says, still needs an additional 50 million teachers to reach universal access to pre-primary, primary and secondary education by 2030. The same release counts 107 million teachers worldwide in 2026.

The surprise is in the composition of the gap. The release says that 58% of this demand is driven by teachers leaving the profession. Most of the hole is not new classrooms waiting for new people. It is people walking out of the old ones. The release gives attrition rates of 7.2% in pre-primary, 5.4% in secondary and 4.7% in primary. It also reports that in 27 of the 68 countries analysed, teacher salaries remain below those of similarly qualified professionals.

A profession losing its people cannot be rescued by a better textbook or a cheaper platform alone. I believe it has to be made worth staying in. That is my judgment, not a sourced finding. The UNESCO figure points the same way: most of the demand is replacement, so keeping people in the job matters about as much as training new ones.

The earlier estimate was smaller and counted a narrower group. UNESCO's teachers page cites its 2024 Global Report, which put the need at 44 million additional primary and secondary teachers. The 2026 release counts pre-primary, primary and secondary together and gives 50 million.

UNESCO's figures for sub-Saharan Africa vary by page and year. A 2021 UNESCO brief reported one trained teacher for every 58 primary students on average in the region, and said the region would need to recruit 15 million teachers. The UNESCO teachers page gives 56 pupils per trained primary teacher. The gap itself is not in doubt. The exact figure depends on which UNESCO page and year you read.

The Grid in Kenya

Kenya shows the global pattern in close detail. The figures here are as reported by Kenyan outlets citing the Teachers Service Commission. Sokodirectory reported in late July 2026, citing fresh figures from the Teachers Service Commission, that the public-school teacher shortfall stood at more than 96,000, including 38,054 in primary schools.

In my judgment, the pressure point is the move of the Competency-Based Curriculum into senior school. Imagine a teacher who learns a new subject on a Tuesday night so that she can teach it on Wednesday morning.

The strain has a human ledger too. A KNUT wellness survey, launched in Nairobi on 26 May 2026, found that 60% of teachers experience burnout and 18% are in financial distress. No sample size is given.

So the new conductor is being asked to take the podium while the orchestra is still being hired. The KNUT figures are the stress test the CBC transition has become, and they are the union's survey, not an independent audit. I believe a country that wants teachers to become designers of learning must first make the teacher's ordinary week survivable.

What the Machines Proved, and What They Did Not

One of the more useful studies for the teacher-in-the-loop model is Tutor CoPilot, from Stanford. Its authors, Rose E. Wang and colleagues, ran a randomised trial with 900 tutors and 1,800 K-12 students from historically under-served communities, according to the arXiv paper. Stanford's own summary page gives a smaller sample, more than 700 tutors and 1,000 students, so the counts vary by source. Students whose tutors used the tool were 4 percentage points more likely to master topics (p<0.01). The tool offered suggestions to the tutor; the paper notes that tutors flagged some suggestions as not grade-level appropriate.

The most striking figure concerns the weakest tutors. Students of lower-rated tutors improved mastery by 9 percentage points. In my reading, the machine did most for the person who needed help most, and it helped by prompting her judgment rather than replacing it.

The counterweight deserves equal space. An independent commentary on the study notes that tutors were more likely to ask helpful questions and less likely to give students direct answers. It adds that outside the lowest-rated tutors the results were not especially substantial or statistically significant, and that students in the treatment group did not improve their final assessment scores. In my view, that caveat is the honest headline. The method is promising. The effect is modest, and it depends on who is holding the tool.

Imagine a classroom where pupils work in pairs over a problem set while the teacher stands at the front, watching. The tool can answer a question in seconds. Nothing in the study tells us what she was thinking.

The machine is good at answering. The teacher is good at deciding which question a child is ready to carry.

What Teachers Fear, and What They Ask For

Teachers are not waiting politely for permission to be promoted, and the surveys of teacher opinion on AI are too varied in wording and sample to support one headline number here. This essay does not quote one.

The Teachers Service Commission has moved to train its staff. On 29 September 2026 it called for a consultant to retrain teachers in practical AI tools for teaching, including data ethics and AI tools for competency-based education. Expressions of interest were due 12 October. The tender notice, as reported by Kenyans.co.ke, does not say how many teachers would be trained or by when.

In my judgment, there is also a gap in the rules: a teacher needs a national framework from the ministry or the Kenya Institute of Curriculum Development, not guidance that circulates from one school to another. Whatever form it takes, the first rule should be the simple one that a conductor needs before the first note is played: a teacher should not put identifiable details about a learner into an external AI tool.

The fear and the demand look like one thing seen from two sides. A teacher who worries that AI will make children stop thinking is a teacher who cares about thinking. The policy question is whether that care gets a route into the classroom, or is left to fight the tools alone.

The Conductor's Card and a Path to Get There

The new job needs a description specific enough to train for. What follows is a Pawanax proposal, not a sourced international standard, and it is written to be argued with.

The conductor's role has six parts. The curator of attention chooses, from the many offers an agent can make in a minute, what the child should meet next. The designer of projects shapes problems that pupils and agents can work on together, with the teacher setting the stakes and the questions. The mentor is the standing adult relationship that lasts across months, not only across lessons. The assessor of judgment grades the child's reasoning, the path by which an answer was reached, and the moments when the agent was wrong and the child noticed. The witness of growth keeps the record that no model update can erase. The guardian of the learner's data, at minimum, never puts identifiable details into external tools.

The training path runs for twelve months and is modular, so a teacher can complete one module, use it on Monday, and bring the result to a peer group on Friday. Months one to three cover AI literacy: what models do well, what they do badly, and how to check. Months four to six cover project design with agents, anchored in the TSC tender topics. Months seven to nine cover assessment of process and data ethics, and each teacher rewrites her own marking rubric by the end of the module. Months ten to twelve cover mentoring and a capstone, in which each teacher runs one real project with a real class and keeps a portfolio of what happened.

The first module is not a tool. It has to be an unlearning. The lecturer's posture, standing at the front and delivering the answer, is the habit most in need of revision. The training must say plainly that the old skill was real and that the new one is harder.

The path also needs measurement. A pilot in a small number of schools, comparing results over a full year and publishing the failures, should come before any system-wide rollout. The Tutor CoPilot evidence suggests a good design can help the weakest practitioners most. The independent commentary reports that the treatment group's final assessment scores did not improve, so a design can help the process without moving final scores. A pilot is how a country learns which of those two stories it is living in.

The Notebook Across Three Model Updates

Imagine a girl of perhaps twelve who keeps a notebook. It is not an agent's log or a platform export. It is a plain book in which, once a week, her teacher writes three things: a question the child asked, a project she abandoned, and a moment she changed her mind about something she had been certain of. Over one school year the models underneath the classroom change three times. The tools get faster, the explanations smoother, and some answers wrong in new ways. The notebook does not change its mind about what the child has become.

This is the part of the Real Time Learning theory I hold most firmly. If learning is continuous and the tools are updated constantly, what persists cannot be the tool or the lesson. It has to be the relationship between a person who knows the child and the child's own record of thinking. A model update can rewrite a lesson overnight. It cannot remember that this child, in March, stopped trusting her first answer. The teacher can.

The notebook is the teacher's promotion in its simplest form. The lecturer was valued for what she knew. The conductor is valued for what she keeps.

So I come back to the globe. The girl's finger has landed on her country, and the teacher turns the globe slowly so the next child can find his own. The chalk is nearly finished, the bell is about to ring, and nothing in the room has been replaced. The question is no longer whether machines will enter the classroom. It is whether the people in the room will get the training, the time and the respect a conductor needs before the next model update starts the orchestra again.

Editorial note: This essay expresses the founder’s ideas, developed with AI-assisted research and writing (parallel Claude Haiku 5.5 agents) and checked against the linked sources on 30 October 2026. Convictions, proposals and forecasts are identified in the text; scenes framed as imagined are illustrative.

Cover photograph / illustration: Nduati.githae (Wikimedia Commons, own work). CC BY-SA 4.0. Original source. Illustrative image: a teacher and pupils with a globe in a classroom (location not stated in the Commons record). Photograph: Nduati.githae, Wikimedia Commons. CC BY-SA 4.0. This image does not document the events described.