Practical Applications
AI in education isn't theoretical — thousands of institutions are already applying AI tools to improve learning, teaching, and operations. Here's how.
In K-12 education, AI applications range from intelligent tutoring systems like Khan Academy's Khanmigo to automated essay scoring, personalized reading recommendations, and early intervention systems that identify at-risk students. Schools using AI-powered math platforms report up to 30% improvement in standardized test scores, with the greatest gains among struggling students.
Higher education institutions are deploying AI across the student lifecycle: from AI-powered admissions review and course recommendation engines to virtual teaching assistants that handle routine student queries. Research universities use AI for literature review, data analysis, and even hypothesis generation. Corporate training departments leverage AI for skills gap analysis, personalized learning paths, and performance simulation.
Emerging applications include AI-powered peer review systems, automated curriculum mapping, competency-based progression tracking, and virtual reality training environments enhanced with AI. The most successful implementations share a common thread: they augment human capabilities rather than replace human judgment.
What we have actually built
The most useful thing we can offer here is not a survey of other people's deployments. It is the work we have shipped ourselves, with its constraints and its corrections visible.
- Educator training on the Dominican Penal Code — a full learning experience for a national school system: scenarios, cases, assessment, a help desk that answers only from the legal text, and a public audit report of its own independent legal review. Built in Dominican Spanish, translated afterwards.
- Brevait — a single-purpose French writing coach for a student preparing the brevet. Deliberately narrow: one subject, one level, one job, and no pretence of being a general tutor.
- Prompt Library and Prompt Builder — the unglamorous middle layer, where most classroom value is actually produced.
- Learning Games and the Ethics Simulator — practice environments rather than content delivery.
Applications outside the classroom count too
A school's hardest AI problem is often not instructional. It is filling next year's intake, or funding the building. We have documented both ends of that: a practical stack for executing a school's communications, a year inside a real anonymised enrolment funnel, and AI for fundraising — which is mostly an argument about what AI must not be allowed to touch.
How to tell a good application from a demo
Three questions have served us reliably. Does the learner do more of the work afterwards, or less? Does it survive contact with a real timetable, a real budget and a member of staff who did not volunteer for it? And can the institution keep running it when the pilot enthusiasm and the discount both expire?
A fourth question is specific to this technology: what is it inventing? Anything that generates citations, figures, names or legal specifics must be checked against a source before it reaches a learner — the discipline described in The Infographic I Didn't Publish, and the reason the Penal Code help desk answers only from the statute and links to the article it used.
The curriculum question underneath
Every application eventually raises the same one: if the machine can do this, should we still be teaching it? We have taken it seriously rather than rhetorically, in Should We Still Teach Coding? and The Maps We Copied. Short version: the answer is usually yes, and usually differently.
Read further
Applied work, written up with its numbers and its limits.
- A Year Inside the Funnel: An Anonymized K-12 Enrollment Campaign WalkthroughWhat running the framework looks like over one season — an anonymized Caribbean K-12 school, the actual decisions week by week, the surprises, the final numbers, and what we would do differently.
- How to Use AI to Execute Your School's Social Media Strategy: A Practical Stack for Schools and UniversitiesAI tools mapped section by section onto the playbook from part one — research, audience modeling, content production, multilingual adaptation, analytics.
- Should We Still Teach Coding? A Principal’s Question in the Age of Vibe CodingYes — but teach the foundation, not the syntax. What to keep, what to change, and the trap of letting two vibes replace learning.
- The Maps We Copied — and What They Teach Us About AI in EducationWe all copied the rivers of Africa out of a book, and no teacher called it plagiarism — because the map was never the point.
Try it yourself
- Prompt Librarytested prompts, by subject and level
- Learning Gamespractice environments, not content
- Tool Directorywhat exists, and what it is for
- AI Suitewhat we build with institutions
- Traininggetting a whole staff to competence
The most effective applications I've studied share two traits: strong pedagogical grounding and meaningful teacher involvement in implementation. Technology without pedagogy is just expensive software.
Practical applications must be evaluated not just by efficiency metrics but by their impact on student agency, critical thinking, and equitable access. Not all applications serve all students equally.
The best applications are the ones teachers create themselves. Give educators AI tools and time to experiment, and they'll build applications that no tech company would have thought of.
The pattern repeats across every sector I've worked in: the most powerful applications emerge when domain experts — not technologists — drive the design. Education's practitioners hold the keys.
Comprehensive AI learning designed for educators, by educators. From awareness to mastery.