Choosing an AI learning platform for a school

An AI learning platform for schools processes student work through machine learning models to generate explanations, feedback or practice items. Before choosing one, you need to know where that processing happens, what it costs in infrastructure and licences, and whether it supports your curriculum or requires you to adapt teaching around it.

Where student work is processed and stored

Cloud-based platforms send every question, answer and interaction to remote servers. That means student work, including wrong answers and partial attempts, leaves your network. Check the provider's privacy policy for data retention periods, whether work is used to train models, and which jurisdictions the servers sit in. Some platforms store timestamped transcripts of every tutoring session; others discard interaction data after a set period.

Desktop software that runs models locally keeps student work on the machine or your own file server. Processing happens on-device, so there's no per-query cost and no dependency on uptime of a third-party service. The trade-off is that local models are generally less capable than the largest cloud models, particularly for open-ended essay feedback or nuanced subject explanations. For structured practice—multiple choice, numeric answers, step-by-step working—the gap is narrower.

Internet requirements and network filtering

A browser-based platform needs continuous internet access. If your school filters traffic by domain or category, you'll need to allowlist the platform's domains, and sometimes dozens of third-party CDN or API endpoints. Investigate whether the platform degrades gracefully on a slow or intermittent connection, or whether a brief dropout loses student work.

Desktop applications vary. Some require an active connection for all features; others only check in for licence validation or curriculum updates. If the core models run on-device, students can continue working during network outages. This matters in schools with unreliable connectivity or where IT policy restricts internet access during certain periods.

Device compatibility and minimum specifications

Web platforms work on any device with a modern browser, which simplifies deployment if your school uses a mix of Chromebooks, iPads and Windows laptops. The server does the computing, so device performance matters less. Desktop software imposes minimum CPU, RAM and storage requirements. Running a language model locally needs at least 8 GB of RAM, preferably 16 GB, and several gigabytes of storage. Older or budget hardware may not cope.

Check whether the platform supports the operating systems in your environment. Some run only on Windows; others include macOS and Linux builds. Mobile apps are common for cloud platforms, less so for locally-run software.

Curriculum alignment and syllabus coverage

A platform aligned to your board syllabus—CBSE, CISCE, state boards, Cambridge, IB—reduces teacher workload. Look for topic lists that match the official syllabus codes and learning outcomes. Platforms that generate practice "on the fly" from a general model often miss syllabus-specific wording, difficulty progression or the precise scope of a topic as examined.

Pre-built curriculums should show their structure: how topics are sequenced, what prerequisites exist, and how difficulty ramps within a topic. Mastery-based platforms that unlock concepts only after prerequisites are met can enforce logical progression, but they require that the curriculum graph is well designed. A poorly ordered graph frustrates students by locking essential topics behind unrelated ones.

Impact on teacher workload

An AI platform should reduce workload, not create new dashboards to monitor. Check what reporting is automatic and what requires manual review. Can a teacher see at a glance which students are stuck, falling behind, or haven't logged in this week? How much time does it take to review a class of thirty?

Platforms that let teachers set goals, record check-ins, or commission custom versions of a topic for individual students add flexibility but also responsibility. Decide whether your teachers have time to use those features or whether you need something more automatic. Some platforms require teachers to curate or approve AI-generated content before students see it; others release content directly and log teacher edits after the fact.

Licensing and cost structure

Cloud platforms typically charge per student per year, sometimes with tiers based on feature access. Check whether the count is concurrent users or named seats, and whether licenses are locked to individual students or can be reassigned mid-year. Some providers invoice monthly; others require annual commitment with a penalty for early termination.

Desktop software may charge per device, per named user, or a site licence for unlimited installs. Volume licensing for schools often includes a discount threshold—50 seats, 100 seats, 500 seats. Ask whether the licence covers automatic updates, curriculum additions, and access to new features, or whether those are separately invoiced.

Running a pilot with one class

Start with a single class in one subject for one term. Choose a teacher who will give honest feedback and a group that represents your cohort—not just your highest attainers. Set clear success criteria before you begin: attendance in minutes per week, completion rate of assigned topics, teacher time saved or added, student survey results.

Pilot on your actual school hardware and network. Performance in a vendor demo on a fast symmetric connection and a new laptop may not reflect what students experience. Log technical issues: installation failures, timeouts, compatibility warnings. If the platform requires network allowlisting, test that the allowlist is complete and that no student-facing features break.

Meet the pilot teacher weekly to discuss friction points. Does the platform fit into lesson time, or does it feel like an extra task? Are students able to self-serve, or do they need constant teacher intervention? Does the AI feedback help students correct misconceptions, or does it generate confusion?

Grout as a desktop option

Grout runs on Windows and macOS, with core models processing on-device. One licence covers three applications: GroutApp for study and vision tools, GroutCode for programming practicals, and GroutFilm for video and audio editing. Volume licensing for schools is available, with pricing scaled to institution size.

GroutApp includes Alpha Learning for institutions—a mastery-learning module where students work through a curriculum of 20 to 60 concepts arranged as a prerequisite graph. Each concept unlocks when its prerequisites are mastered. Mastery requires a score of at least 85 per cent, three correct answers in a row, and at least four attempts. Concepts return for spaced review after three days, then at intervals growing by a factor of 2.2, up to 60 days. An AI tutor provides hints without giving away the final answer, with a default limit of 12 hints per session.

A teacher—called a Guide in Alpha Learning—sees class mastery percentages, minutes logged, streaks, and which students are inactive, stuck, or below pace. The Guide can record check-ins and goals, and commission a custom version of a topic for one student. The custom version keeps the same syllabus scope, question count, and rigour, but rephrases explanations or adjusts examples.

Grout also grades scanned handwritten answer sheets, supporting workflows where students write on paper and submit for automated marking. Because the models run locally, deployment does not require continuous cloud connectivity for core study features, though Alpha Learning itself does require an internet connection.

Questions to ask any provider

Where is student data processed, and can you delete it on request? What happens if a student leaves the school mid-year—does the licence return to your pool? Does the platform handle multiple academic years and grade levels under one contract, or do you re-license each year? If the platform generates content, who owns it—the school, the student, or the provider?

Ask for references from schools of similar size and type. Request a breakdown of support: email-only, phone hours, dedicated account manager, on-site training. Clarify what counts as an incident versus a feature request, and what response times are guaranteed. A platform is only as good as the support behind it when your teachers encounter a problem the week before exams.