Most ed-tech pitches lean on adjectives: “transformative,” “revolutionary,” “personalised at scale.” AI Ready School (AIRS), the platform built by Hyderabad-and-US-based Intellina AI, has instead started publishing numbers, pilot reports and a multi-year case study from schools it has worked in. That’s a useful thing for a reader to have, because numbers, unlike adjectives, can be checked, questioned, and sat with a little scepticism. Here’s what they show.

What the Company Says It Builds
AIRS is a five-part system rather than a single app. Cypher is a personal AI tutor students can ask questions to. Morpheus is the piece aimed at teachers, letting them build a lesson once and have it adapted for a full class. Zion is a broader toolkit for research and project work. NEO is a physical, on-campus AI and robotics lab. Matrix is the backend infrastructure schools would need to run all of this on their own data, rather than someone else’s servers.
The pitch, in short: don’t replace the teacher, multiply what one teacher can do for thirty different students learning at thirty different speeds.

The Long Case: Four Years at One School in Raipur
The company’s own case study centres on NH Goel World School in Raipur, which it describes as its longest-running partner, renewing the relationship five times since October 2022. According to the company’s account, the relationship began narrowly, teaching computational thinking alongside math and coding, well before AIRS existed as a product, and expanded step by step: creative coding tools, an inter-school AI competition called Data & AI Talks that later grew into a national event called the AI Startup Show, a generative AI course for students, and eventually full adoption of the AIRS platform itself, including a six-day AI training workshop for teachers.

The school’s principal put some of this in writing in an April 2026 appreciation letter, noting that students had become more capable at research, presenting ideas, and independent problem-solving, and that teachers had grown more confident using AI tools for lesson planning and storytelling. That’s a genuine, on-letterhead endorsement, and worth taking at face value as one school’s experience. It’s also worth remembering that a testimonial written by a satisfied customer for the company that served them, not an independent evaluation, and readers should weigh it accordingly.

What’s more interesting than the endorsement itself is the sequencing: the company’s telling is explicit that computational thinking, the unglamorous work of teaching kids to break problems down step by step, came two years before any AI tool showed up. “NH Goel didn’t just adopt what we built,” says Chiranjeevi Maddala, Co-founder and CEO of Intellina AI. “They helped us discover what was worth building in the first place. Every school we work with today benefits from lessons that started in that one campus in Raipur.” It’s a founder’s framing of the relationship, and worth reading as one, but it does match the shape of the case study itself: years of iteration inside one school before anything became a product other schools could buy. That ordering, thinking first, tools second, is a reasonable structural claim independent of who benefits from making it, and it lines up with a broader trend in the field right now: CBSE’s own new computational-thinking mandate for Classes 3-8, rolled out in April 2026, rests on the same foundation, decomposition, pattern recognition, abstraction, algorithmic thinking, that this case study says NH Goel started with years earlier.

The Short Case: A Three-Day Pilot in Dehradun
A newer, smaller pilot at The Indian Public School (TIPS), Dehradun, gives a more granular and more checkable look at what the platform does inside a single lesson. Over three days in late July 2026, 20 Grade 8 students studied a biology unit on health and disease end-to-end through the platform, a baseline test, an AI-built lesson, and a final test on the same material.

The topline numbers: class average moved from 49% to 70%, a 21-point jump. Fifteen of twenty students improved, one stayed flat, and four scored lower on the final test than the baseline, a detail the company’s own report includes rather than omits, which is worth noting. The platform also tracks something it calls “reasoning confidence,” an attempt to separate correct answers backed by real understanding from lucky guesses, and that measure rose from 30 to 55 out of 100.

There’s a smaller, easy-to-miss data point that arguably says more than the headline score: students took nearly three minutes longer on the final test than the baseline, 19 minutes 46 seconds versus 22 minutes 41. A faster wrong answer and a slower right one tell different stories, and taking longer alongside higher accuracy is at least consistent with students working through problems rather than pattern-matching their way to a score.

The pilot also generated something more useful to an actual teacher than a single percentage: a list of specific, named misconceptions. A handful of students believed a disease vector directly causes illness, rather than transmitting a pathogen. A few thought antibiotics work against viruses, or against every kind of pathogen. Two believed malaria is bacterial rather than parasitic. That’s a genuinely different, and more actionable, output than “your class scored 70%”, it tells a teacher exactly what to reteach and to whom.

What the Students and Teachers Actually Said
Eighteen of the twenty students filled out a feedback survey afterwards. The results skew positive, unsurprisingly for a novel, short-duration pilot: 94% wanted to keep using the platform, 88% said it felt more engaging than a normal class, and all the students who’d needed a catch-up mini-lesson said it helped. The two teachers running the pilot rated their overall experience a perfect 5 out of 5 and said they’d recommend it to another teacher without qualification.

According to the report, a few students asked more challenging questions. Some found the video playback speed made content harder to follow. Cypher, the AI tutor, scored slightly lower, 4.5 out of 5, on staying strictly on topic, a small but real gap the company itself flags as “worth continued monitoring as usage scales.”

Reading the Numbers Honestly
A few things are worth holding in mind before treating any of this as proof of anything on a scale.

Twenty students over three days are a pilot, not a study. It’s too small and too short to generalise about, and the company doesn’t claim otherwise; its own report frames this explicitly as evidence to support a decision about expanding the partnership, not as a finished result. A 21-point score jump after a single, unusually intensive lesson delivered as a novelty is a different thing from a 21-point jump sustained over a semester of ordinary use, and the report doesn’t yet have data on the latter.

It’s also self-reported, company-run, and company-published data, collected and written up by the platform being evaluated. That doesn’t make the numbers false, but it does mean they haven’t been through the kind of independent check that would make them fully load-bearing on their own. The 4 students who scored lower than their baseline are a useful reminder that the platform isn’t uniformly effective even within one small pilot group, and it’s worth asking what happened for those specific students before assuming the average tells the whole story.

The NH Goel relationship, running four years and five renewals, is a stronger signal in one way: schools that don’t see value in a tool generally don’t keep paying for it and inviting it deeper into the building. But a single long-running client is still one data point, not a track record across dozens of schools with comparable rigour.

The Bigger Question This Raises
Underneath both case studies is a more interesting structural bet than the AI itself: that curriculum should live with the teacher, not with the platform. Most ed-tech tools hand teachers a finished product to deliver. AIRS’s stated model is closer to giving teachers the tools to build their own lessons, then using AI to personalise how that same lesson reaches each student.

Whether that holds up matters more than any single pilot’s score. A platform where the teacher stays the author of the curriculum is solving a different problem than one where the platform decides what gets taught, and it’s the more defensible position if you’re worried, reasonably, about AI hollowing out a teacher’s role rather than supporting it. Maddala has described the philosophy behind that choice in blunt terms: “We cannot replace teachers. Whatever technology may come, we should not replace teachers. Teachers should become even more powerful.” That’s the kind of line a founder says to every journalist, and it’s fair to treat it with the same scepticism as the rest of the company’s claims, but it does at least match the design choice both case studies point to, tools built for teachers to author, not tools that author for them. Right now, that’s still mostly a design philosophy backed by one detailed long-term relationship and one small short-term pilot. It’s a promising early shape, not yet a proven one, and the company’s own materials, to their credit, mostly say so themselves.

Share on Facebook

Post on X

Follow us

Save