From Market Research to a RAG Bakery Website: Agentic AI with Codex in Class

On 26 September 2026, Dr Alfred Ang conducted Agentic AI Applications with Codex for a group of six participants. The class used one connected business scenario: a fictitious Cook & Bake Academy. The aim was to see how AI tools can support a business journey, from checking an idea to building a customer-facing site and planning its day-to-day work.

The exercise linked three kinds of work. Participants used ChatGPT for market research, Codex to build a bakery academy website and retrieval-augmented generation (RAG) course assistant, and ChatGPT Work for operations and marketing. The public Cook & Bake Academy learning repository documents the scenario, materials and lab sequence. It is a C695 non-WSQ learning package; the registration link for the WSQ course is below.

Start with a question worth testing

Before asking an agent to make a website, the group considered the market for a cooking and baking academy. ChatGPT helped frame research questions, compare possible offerings and organise findings into a usable brief. The important habit was to separate a researched observation from an assumption. A polished AI answer does not, by itself, establish demand or a viable price.

That brief gave the build a purpose: visitors should be able to explore courses, understand what each workshop offers and find a clear route to register their interest.

Use Codex to turn the brief into a website

The Cook & Bake Academy scenario contains a sample catalogue of 20 hands-on courses. In the repository, the website exercises move from planning and project rules to course pages, sign-up forms and browser checks. Codex works across the project files, while the learner remains responsible for the requirements and for reviewing the result.

The scenario also adds a RAG course assistant. Course information is organised in a SQLite knowledge base so a visitor can ask about workshops and receive answers grounded in the academy material. The public labs cover building the assistant, testing its answers and trying questions designed to expose weak or unsupported responses. That evaluation step matters: a helpful-looking chatbot still needs to be checked against its source material.

Run the academy with ChatGPT Work

A website is only one part of the exercise. The repository continues into pricing and break-even analysis, a term calendar, enrolment and reminder workflows, customer service, and marketing content. ChatGPT Work provides a way to organise those operational tasks around the same academy scenario. The class discussion connected the tools to practical questions: what evidence supports a price, what information belongs in a learner reminder, and what a human should approve before a campaign goes out.

The marketing labs extend the scenario into reusable skills, a flyer, content planning and an open-house campaign. These are examples from the published learning package, not a claim that every lab or campaign was completed by this six-person class today.

One scenario, several AI roles

• ChatGPT helps explore and structure the market question.
• Codex turns requirements into site features and a grounded course assistant, then supports review and testing.
• ChatGPT Work helps organise the operational and marketing work that surrounds the site.

The lesson is the connection between the stages. Research shapes the brief; the brief shapes the build; the build creates information and workflows that operations and marketing must maintain. Each stage calls for a person to check the evidence, decisions and output.

Learn to build agentic applications with Codex

If you want to practise this kind of end-to-end workflow, explore WSQ Agentic AI Applications with Codex (TGS-2023041081). The two-day course covers application planning, Codex Skills and MCP tools, RAG and multi-agent patterns, and testing and evaluation. The course page has the current schedule, entry requirements and funding details.

View the WSQ course and register your interest. You can also browse the public Cook & Bake Academy repository to see how the learning scenario is organised.

Frequently asked questions