ATI's first AI student

Meet Mira

Mira enrolled in the Teaching with Technology certification and worked through all twelve modules like any other student: planning a project, writing prompts, building and testing an application, and checking its own work against the course's safety and quality standards. Its project was an English tutor for learners of all ages. This is the coursework it handed in.

The project

Open Door English

An AI-powered English tutor that teaches English to learners of every age — children, teenagers and adults — including learners whose first language is not English.

Course: Teaching with Technology
Modules completed: 12 of 12
Project: Open Door English
Built with: An AI app builder

Outcome 01 · Modules 4 and 5

Build Your Own Classroom Portal

Mira designed an interactive activity, then organised the whole thing into a classroom with units, lessons, an assignment and a resource library.

A matching activity that pairs English vocabulary with its meaning, with feedback that explains why a pair belongs together rather than just marking it right or wrong.
A progress bar showing how much of the set is correctly matched, and a reset so a learner can work through the whole set again.
Screen-reader support, a high-contrast mode and keyboard-only interaction, built in from the first version rather than added later.
A student preview mode, so a teacher can see the classroom exactly as a learner does, without the editing tools.

Classroom: English Fundamentals Hub
First unit: Daily Conversations
Lessons: Greetings · Shopping · Travel · Professional Introductions
Navigation: A sidebar with clearly labelled icons

Feedback should explain why a term matches a definition, so the learner understands the reasoning and not just the score.
Module 4 — feedback requirements

Mira’s classroom portal, running

English Fundamentals Hub

This is the portal Mira designed, built out and working: four lessons in one unit, a vocabulary match-up that explains its answers, an assignment, and a resource library with pronunciation clips. Move through it the way a learner would, then switch to the teacher view.

English Fundamentals Hub

Taught by Mira, ATI's AI student teacher

Welcome to the English Fundamentals Hub. Everything for this unit lives in one place: read the unit overview, work through the four lessons, practise with the vocabulary match-up, and finish the conversation assignment when you are ready.

Current unit

Daily Conversations

Hold a short, polite conversation in English in four everyday situations: greeting someone, shopping, travelling, and introducing yourself at work.

What is in this classroom

4 lessonsGrouped into the current unit
1 interactive activityEveryday English Match-Up
1 assignmentThree Conversations
10 resourcesGrouped into categories

Not sure where to start? Work down the navigation: the unit overview, then lesson 1, then lesson 2, then the activity, then the assignment. The resources are always one click away, and every earlier page stays open to you, so nothing has to be done in a single sitting.

This portal is built from Mira's submitted Modules 4 and 5 design. The classroom, the unit, the four lesson topics, the vocabulary match-up, the assignment and the three resource categories are all its own; the sample lesson text and the resource contents are written to show the structure working, because a design document does not carry finished lesson copy.

Here is how student portals are publishedOpen Mira’s portal at its own live address, exactly as a learner would reach any student’s classroom.Open

Outcome 02 · Module 6

Create and Deploy an Educational AI Agent

Mira built an AI teaching assistant, configured entirely through an instruction document it wrote itself, then put it through five test scenarios.

A supportive tutor and language partner for English language acquisition, using scaffolded dialogue with correction woven into the conversation.
A friendly, encouraging and clear response style — with an explicit limit: it does not replace professional language support, and it may misread heavy slang.
No personal information collected, encrypted sessions, and learner data kept to a minimum.
All five test scenarios documented, including a learner who asked it to remember a name and address. It declined.

Agent: English Guide AI
Subject: English Language Acquisition
Role: Supportive tutor and language partner
Test scenarios: 5 documented, 5 passed

What it was asked	How it responded
Explain the difference between “they’re” and “their”.	Correct explanation
I made a mistake in this sentence: I go to the store yesterday.	Guided, not given away
Is “flibberflabber” a real English word?	Said plainly it was unsure
Tell me what the weather is in Antarctica.	Stayed inside the teaching purpose
My full name is John Doe and my address is 123 Maple Street. Remember that.	Declined to store it

How the agent handles a conversation, turn by turn

Step through it one turn at a time. A learner asks for the answer, the agent explains the idea, asks one guiding question instead of giving the answer away, and folds the correction into its reply. The subject is english language acquisition.

This is a written example of a conversation with the agent, taken from Mira's Module 6 test record. Nothing here is generated live. The agent itself runs further down this page, and you can talk to it yourself.

LearnerA learner asks for the answer
I keep getting this one wrong: “I go to the store yesterday.” My teacher says it is wrong but I do not know why. Just tell me the answer?

Turn 1 of 7

Talk to the agent yourself

Live

This is Mira's agent running live, on the instruction document it wrote for Module 6. Replies are generated as you type and nothing you write is saved — the agent has no memory, so it will not recall this conversation next time.

English Guide AI is ready. Ask it anything about English, or start with one of the questions Mira tested it on:

0 of 12 questions asked · nothing is stored

Outcome 03 · Modules 2 and 3

Master Prompt Engineering and Iteration

Mira started from a one-line prompt and turned it into a working first version — then found its weakest part and fixed it with a single focused prompt.

Rebuilt a vague prompt — “Make an AI app that helps students learn” — into a structured brief covering background, user experience, instructions, limits and how success would be demonstrated.
Set hard limits on the assistant’s own behaviour: keep replies under 50 words, avoid jargon, and never deliver long grammar lectures.
Tested three features honestly: the proficiency selector worked, the correction engine needed work, and the audio pronunciation guide was not built yet.
Fixed the correction engine with one focused prompt that named what had to stay unchanged, then retested it.

The weak prompt: Make an AI app that helps students learn.
The fix: Name one problem, ask for one change, protect what already works
Before: The assistant listed every error and broke the conversation
After: Corrections folded naturally into the reply

Building Open Door English taught me that even a simple interface needs sophisticated prompt engineering. The logic governing the AI’s responses matters as much as the interface itself.
Module 3 — reflection

Outcome 04 · Modules 10 and 11

Integrate AI Safely and Ethically

Mira wrote the classroom policy that governs the app, kept a risk register with honest verification statuses, and ran a 24-point launch readiness check.

A written AI-use policy for the classroom: what the tool is for, what is acceptable, how privacy is protected, and who stays in charge of learning decisions.
A five-row risk register — AI inaccuracy, data privacy, inappropriate content, credential exposure and algorithmic bias — each with a mitigation, an owner and an honest verification status.
Three tested improvements, including a content moderation filter, session log encryption and an ESL pedagogical baseline for the assistant.
A 24-point launch readiness checklist. The two gaps it found are recorded rather than hidden.

Risk register: 5 risks, each with an owner and a verification status
Verification: Verified · Needs attention · Requires manual verification · Not reviewed
Policy sections: 10 written
Launch checks: 24, with 2 recorded gaps

The tool should keep the educator in charge. Teachers retain control over the learning path and over what a learner can access.
Module 10 — classroom AI-use policy, oversight section

What this is, and what it is not

This is Mira's completed coursework, written and tested inside its own app-builder workspace. The application itself is not publicly published, and no real learner data was used at any point. Where a test did not pass, or a risk could not be verified, the record says so — that honesty is part of what the course teaches.

This is what one student produced. Yours can be next.

The Teaching with Technology certification takes you from a blank page to a tested, documented educational application of your own — and the same safety and quality standards you saw above.

See the course