The $11 Phone Call: How Low-Tech Human Contact Beats High-Cost AI Tutors
⏱️ 5 MIN READ
Four stories on innovative ways to bring down costs for education programmes and how students interact with edtech tools
A Nature paper covering five randomised trials that found a basic phone call beats almost everything else in education for eleven dollars a child, ministers adopting a joint statement in Paris, and two arguments about why edtech keeps failing on implementation and on connection
1 | AFRICA Weekly 20-minute phone calls lifted learning three times more than the average education programme across five countries, at $11 a child
-
Nature published five randomised trials run across India, Kenya, Nepal, the Philippines and Uganda, covering 16,936 households. Children received a 20-minute phone call each week for eight weeks, plus SMS, teaching foundational numeracy targeted to the level each child was actually at. The combined arm produced 0.32 standard deviations of learning. SMS alone managed 0.078 and worked in only two of the five countries. At $11 a child, the programme delivers 3.6 learning-adjusted years of schooling per $100, placing it in the top ten of 150 interventions ever costed on that measure.
Two findings break the usual pattern. Government teachers matched NGO instructors, at 0.40 against 0.30 operation levels, where most programmes collapse on handover to government. And effects grew rather than decayed across trials, because targeting accuracy rose from 50.9% in Nepal to 81.5% in Uganda. In Uganda, treated Grade 4 children overtook the Grade 5 children above them. The authors point to a near-identical phone programme in Sierra Leone that delivered untargeted lessons from a mass radio broadcast and produced no effect at all. Karnataka is now scaling the approach as Ganitha Ganaka.
READ MORE AT NATURE
9 SEPTEMBER 2026
2 | GLOBAL More than 25 education ministers signed a joint commitment in Paris to keep education a public good as AI reshapes classrooms
-
Digital Learning Week ran 8 to 11 September at UNESCO headquarters, drawing over 1,500 in person under the theme Facts, Frictions, Frontiers. The statement holds that education must be sustained as a human right and a common good, and notes AI is already changing how students think, how teachers make decisions and how learning is graded. UNESCO released a compilation of policy briefs alongside it covering AI safety governance, sovereignty, procurement, cost models and age-appropriateness. It has now directly supported 58 countries on AI competency frameworks, curricula and quality-assured teacher training. Achille Mbembe of Wits gave the public lecture on the frictions AI is generating across education systems.
READ MORE AT UNESCO
8 SEPTEMBER 2026
3 | USA A two-year trial found students reach for an AI tutor when curious and abandon it when struggling
-
The trial ran across 18 Tennessee middle schools. 96% of students tried the AI tutor at least once, then turned to it in only 17% of the moments they had got something wrong. The tutor was patient, competent and available at eleven at night, and students still did not reach for it when they were stuck. The argument built on that finding is that the sector has consistently underpriced implementation, treating a product that exists as a product that works. The prescriptions are concrete: judge tools by demonstrated learning outcomes rather than usage, build evidence into renewals rather than purchases, and require vendors to publish independent evaluations and disclose what their products cannot do.
READ MORE AT FORTUNE
11 SEPTEMBER 2026
4 | USATen edtech platforms spent nine months testing whether their tools could build student relationships rather than replace them, and found it worked
-
Julia Freeland Fisher at the Clayton Christensen Institute argues 2026 brought two distinct complaints into the open. One is efficacy, whether tools do what they claim. The other is isolation, what they cost students in connection even when they work. Her explanation is market-structural rather than moral: overleveraged, understaffed schools bought technology looking for efficiencies, not relationships, and got what they paid for. Anthropomorphic AI tutors, coaches and mentors now threaten to turn an unsocial past into an antisocial future.
The pilot ran with 10 college and career guidance platforms building what she calls prosocial AI, technology in service of human connection rather than in place of it. Roadtrip Nation built an AI simulation letting students rehearse a career conversation before having a real one, and students reported higher confidence in both booking and conducting interviews. REACH Pathways had students map the networks they already had, and nearly half reached out to a contact during or straight after. Overgrad built an alumni directory connecting current students to recent graduates, after a focus group produced the line that explains the gap in the market: LinkedIn sucks, nobody actually wants to meet anybody there.
READ MORE AT SMART BRIEF
10 SEPTEMBER 2026
Every source in this issue was checked against its published page metadata before inclusion.