Switches; Signals; Results!
From standardised rails to classroom change.
… But first, the news:
… But first, the news:
AI is learning to both make and understand video. Video generation is advancing fast: OpenAI’s Sora 2 produces realistic short-form clips , while Google’s Nano Banana faithfully follows instructions. Meanwhile, models are also learning to reason through video: Sora 2 accurately answers benchmark questions via voice-over, Veo 3 demonstrates step-by-step spatial reasoning, and new research shows that GPT-4V can detect human social cues. These are all abilities that will soon enable AI tutors to read classroom dynamics and tailor feedback in real time.
OpenAI wants AI to be a part of your everyday life. OpenAI is rapidly expanding what users can do with ChatGPT. Pulse delivers personalized daily briefings drawn from chat history and app integrations, while the Agentic Commerce Protocol enables secure, step-by-step purchases directly through chat. App integrations with Spotify, Canva, Zillow, and others are transforming ChatGPT into a central hub for managing productivity, creativity, and recreation.
Agents are becoming the defining feature of frontier models. The world’s leading labs are racing to make AI that acts as well as it answers. OpenAI’s new Agent Builder gives anyone a drag-and-drop canvas for designing multi-step, tool-using agents, while Anthropic’s Agent SDK and Microsoft’s Agent Framework offer developers a deeper, code-first framework to embed AI-powered agents directly into their products. Meanwhile, DeepSeek has plans for its own agent platform.
Africa takes a step toward coordinated AI governance. Across Africa, governments are advancing a shared agenda for responsible AI governance. A new UNESCO-supported initiative under the G20’s “AI for Africa” program will help build the capacity of policymakers, civil servants, and judicial officials in ethical governance, data protection, and locally led AI development to bolster institutional readiness for their national AI strategies.
On the rails
Railway systems have taught us a simple but important lesson. If the tracks use the same gauge and rules, trains move, freight moves, and people move. Education data is no different. Shared IDs and a small set of shared pipelines can enable ministries, vendors, and AI tools to move in the same direction. But interoperability is only half the story. Unless governments, school leaders and teachers can act onwhat the data shows them this week, we’ve just built another way to confirm what we already knew - too many children are not learning.
So, what’s the answer?
We do not need perfect learner ID systems to improve instruction. We’ve seen this with TaRL, structured pedagogy, and instructional coaching. We know we can make improvements with data at the classroom and even school level. That being said, we can walk while we chew, right? Let’s dig in. All aboard!
What is interoperability, and why does it matter?
Interoperability is the invisible reason you can board a train at Shinjuku station at 6:49am without worry. It’s the underlying set of rules and standards that lets a complex system run smoothly.
In education, it means:
Shared standards – like one track gauge, ensuring data is shared and flows across EMIS, assessments, and AI tutors without conversion or re-entry.
Aligned identifiers – so each learner record stays consistent across platforms.
Common rails – enabling ministries, vendors, and funders to build upon without dead ends.
When these rails are missing, data splinters. Teachers juggle dashboards that don’t match and ministries can’t link spending to outcomes. AI will just accelerate this fragmentation at rapid pace.
Why it matters:
Budgets and outcomes should ride the same trains. Linking spending, staffing, attendance, and learning outcomes makes impact directly visible.
AI without shared rails == measurement theatre. Integrating AI into governance and classroom practice means we can measure things and give feedback to schools that will result in real change, rather than delivering useless dashboards.
The cost of waiting compounds. Every non-conforming tool added today can become tomorrow’s stranded sidecar.
Equity is at risk. Well-resourced schools can patch systems together whereas under-resourced schools are less able to keep pace.
When the rails are missing, the cost shows up in classrooms
Imagine this:
Melanie, a head teacher in Ghana’s Asempa district, begins her day by checking a stack of tablets on her desk. One tablet records attendance, another manages school lunch payments, a third shows reading scores from a new AI tutoring app, and two more hold data on coaching and exams. Each system was introduced with promise, but none of them recognize the same learner ID. She spends hours trying to reconcile lists of pupils with different IDs in different apps. When the district office requests data for planning, it takes significant effort to pull together a report, and the data still don’t quite match up. District officials can’t easily connect what the AI apps report about her school with spending and exam results at the state and national level.
The potential is obvious: with a single learner ID and open standards, Melanie could turn those five tablets into one connected picture of learning, freeing up teachers’ time, improving data integrity, and linking classroom progress to national outcomes.
Easier said than done, but not impossible by any means.
Where has it worked in national education systems, and what has it enabled?
India’s DPI Stack
Over the last decade, India has built India Stack, the digital public infrastructure(identity, payments, open APIs) that created common rails across sectors. Building on this ecosystem, the Ministry of Education launched DIKSHA, a national platform that connects digital tools with curriculum, assessments, and teacher training.
Over the last couple years in Andhra Pradesh, a personalized adaptive learning program (PAL) was expanded to more than 1,200 schools. By embedding the program in government systems for financing, procurement, and monitoring, the state enabled the innovation to work on existing rails. After 17 months, students in PAL schools gained nearly two years of extra learning, far above expected levels.
Funding choices also mattered. The Government of Andhra Pradesh used funds from Samagra Shiksha, India’s flagship school education program, to support the expansion by providing tablets and training teachers. Because the state financed and supplied the hardware and implementation support, the rollout proceeded through government procurement and management systems. This also meant monitoring was carried out by Samagra Shiksha teams using regular reviews of dashboards at school, district, and state levels, complemented by usage reports to headmasters, reinforcing the program’s integration into government routines.
Support the rails
Interoperability succeeds when funders reinforce system-wide rails instead of stand-alone pilots. To support this shift, funders can:
Anchor solutions in government systems. Direct resources through national programs and budget lines rather than parallel projects, ensuring sustainability and ownership.
Finance the rails, not just the trains. Prioritise investment in open standards, learner IDs, and EMIS upgrades; the infrastructure that enables future tools to run on time and in sync.
Promote alignment across vendors and grantees. Require new solutions to plug into existing identifiers and reporting systems, reducing fragmentation and “shadow registries”.
Invest in decentralized capacity to act on data. Build teacher, school, district officer and ministry skills so that interoperability translates into instructional change, not just more dashboards.
Reward integration over duplication. Use procurement and grant conditions to privilege solutions that connect with the wider system rather than create new silos.
Funders don’t need to drive the trains, but they can help keep the rails aligned so data flows where it’s needed most - from classrooms to ministries back to learners - to support real learning gains.
This edition of the Learning Futures Briefing was led and written by Shabnam Aggarwal, with contributions from Dr. Robin Horn, Sara Cohen, and Faizan Ul Haq.
