App Development

Persona and Data

Overview

Designing an agent persona grounded in a structured database. Students model tables and scoped access keys, write a persona as a job description, and build an AI agent that answers only from a supplied catalogue.

  • 01 By the end, you can Design the tables the product needs, choosing a fitting type for each column and adding example rows.
  • 02 By the end, you can Set up an access key scoped to only the actions needed, and explain why limiting it matters.
  • 03 By the end, you can Write an agent persona as a job description grounded in that data.
  • 04 By the end, you can Combine both so the agent answers only from the catalogue and says "I don’t know" when a fact is absent.
  1. Designing Your Database
    Applied craft
    • Worksheet Design Four Tables Teams sketch four data tables and label what kind of information each column will hold.
  2. Seeding and Connecting Data With AI
    AI concepts
    • Worksheet Predict Then Compare Students write what they expect the AI to change in the database, then check what it really did.
  3. Giving the Agent a Voice and Avatar
    AI concepts
    • Worksheet Persona Sheet On Paper Students fill in a full plan for their AI character on paper before touching any computer.
  4. Critiquing the AI Agents
    Assessment
    • Discussion Critique The Agents Class looks at each team's AI agent and answers three questions about how well it fits and talks.

Applied · App Development

# Knowledge & Skills

ACA-Applied
K Knowledge · what you know
Structured data is made of labelled boxes. Each box has a label and a value. Some boxes hold lists of smaller boxes inside them.
There is a design choice to make. You can save a copy of the catalogue once when the app is built, or fetch fresh data from the database each time it is needed. Both work, but the team must say which one they chose.
A database, a table, a row, and a column each name a different level of a database, and every column has a type that fits the kind of information it holds. point w3d3-s1-p1
An access key can be set to allow only certain actions, such as read, write, or delete. Limiting what the key can do limits what can go wrong. point w3d3-s2-p1
S Skills · what you can do
Design the tables your product needs, and add example rows. point w3d3-s1-p2; evidence w3d3-e4
Give the AI agent a lip-syncing avatar (an animated face that moves with speech), built from named image files. point w3d3-s3-p3; evidence w3d3-e7
Explain what a database, a table, a row, and a column are, and choose the right type for each column. from objectives
Set up an access key that allows only certain actions, use it to read, write, and delete data, and explain why limiting it matters. from objectives
Ask any AI model for its sources, and treat an answer with none as made up. from objectives
T Techniques · named subskills
Every Fact Travels in the Message The model can't see your site or notes; it only knows what's in the prompt.
Example Exchanges as Spec Sample conversations are part of the persona specification, because the model imitates what it's shown.
Persona as Job Description Write the agent's persona as labelled sections and short instructions, not a vibe.
Verify Before Accepting Model-retrieved or model-generated records are checked against the source before use.

AI

Aligned with AI4K12 · Five Big Ideas in AI View framework →
AI4K12 · Five Big Ideas in AI
2-A-II

Symbolic Representations

Using symbols, words, or formal notation to stand for real-world objects or concepts helps computers reason logically.

AI4K12 · Five Big Ideas in AI
2-A-III

Data Structures

Organized ways of storing and arranging information (like lists, trees, or graphs) enable efficient computation.

AI4K12 · Five Big Ideas in AI

Models Only Know What You Type

An AI model cannot see your website, your notes, or your plans. It only knows the facts you type into the message you send it.

AI4K12 · Five Big Ideas in AI

Persona Is a Job Description

A persona is a job description, not a mood or a tone of voice. It has labelled sections and short instructions. Example conversations are part of the plan, because the AI model copies the pattern it is shown.

AI4K12 · Five Big Ideas in AI

Ask for Sources

Any AI model can be asked for its sources, and an answer that gives none is treated as made up.

AI4K12 · Five Big Ideas in AI

Tables, Rows, Columns, Keys

A database holds tables, a table holds rows, and a row holds columns. The type you choose for a column decides what it can store. An access key can be limited to read, write, or delete, and it should be given only the actions it needs.

AI4K12 · Five Big Ideas in AI

Verify Model-Written Data

Any data the AI model writes is checked against a source before it is accepted.

AI4K12 · Five Big Ideas in AI

Set the Shape First

An AI model can be told to answer only in a set shape, such as named fields, a table, or a block of HTML (the code used to build web pages). It keeps that shape for the rest of the conversation once it is set. Ask for the shape before asking for the content.

AI4K12 · Five Big Ideas in AI

Finish the Spec Sheet First

The persona plan is finished on paper before any prompt is written. A prompt is an instruction you type to the AI. If something is not on the sheet, the AI agent does not know it.

Aligned with UNESCO · AI Competency Framework for Students View framework →
Ethics of AI
4.2.3

Application Skills

Gaining practical knowledge of AI techniques and applications, including data handling, algorithm selection, and analytical practices.

Aligned with Derived from course activity
Derived from course activity

Prompt Engineering for AI Agents

Iterating a prompt/system instruction until an AI agent's responses stop sounding generic and actually reflect the specific product or context. Evidence: "Does this information sound generic? Or does it sound actually tailored to the website?" (app-development, w3d3).

Derived from course activity

Deployment & Secret Management

Setting up a dev environment and deploying live while keeping API keys out of source, held as environment variables. Evidence: "SETTING UP A DEVELOPMENT ENVIRONMENT: 1. Make a folder on your laptop..." and "holding the API key as an environment variable" (api/app-development).

# Competency Goals

ACA-AI.Goals
Aligned with UNESCO · AI Competency Framework for Students View framework →
UNESCO · AI Competency Framework for Students
CG4.2.3.3

Analytical and Synthesis Practices

Learning to combine, interpret, and synthesize AI outputs thoughtfully rather than accepting them uncritically.

UNESCO · AI Competency Framework for Students
CG4.2.3.2

Age-Appropriate Technical Skills

Developing hands-on technical abilities in AI work that match one's educational level and learning needs.

UNESCO · AI Competency Framework for Students
CG4.2.3.1

Data and Technique Knowledge

Understanding properties of different datasets and the strengths and limitations of various AI methods and algorithms.

UNESCO · AI Competency Framework for Students

Have Codex Add Rows

Have Codex (an AI tool that writes code and data) add rows to your database, then check every row it wrote.

UNESCO · AI Competency Framework for Students

Pull Stored Data In

Pull that stored data onto your page and into your AI agent.

Deliverable

# App Development

Projects.Briefs

Build an AI agent that uses a structured database, so it only answers using facts from a supplied catalogue. A designed database (four tables with example rows of data and a stated type for each column), an access key that only allows certain actions, rows of data written by Codex and checked one by one, and an AI agent that pulls in this data. A completed persona plan on paper.

Transfers to: Checking every row an AI wrote to your data, before trusting it.

See past cohorts ship this in the Showcase →

# Assessed toward these aims

Projects.Rubrics

This deliverable is where students demonstrate

  1. C Design an agent persona and ground it in a structured database
    Look out for
    • Table design: The tables match what the product needs. Each column has a fitting type of information, and example rows of data are given.
    • Access key limits: An access key with limited permissions is used. Read, write, and delete are only allowed where needed, and the reason for limiting it is explained.
    • Checking AI-written rows: Every row Codex wrote is checked against a source before it is trusted.
    • Persona plan: The persona is written as a job description, with labelled sections and example conversations. It is completed on paper before any prompt is written.
    • Using real data and sources: The AI agent answers using the supplied catalogue. Any answer given with no source is treated as made up.