Hands-on Codex training so the people closest to the work can build their own tools with OpenAI’s coding agent, without becoming developers.
10,000+ people trained across 500+ organisations
Codex is OpenAI’s coding agent. It is marketed at engineers, but if your firm already runs on ChatGPT, some of the most useful people to put in front of it are the ones who would never call themselves technical. They know exactly which report, reconciliation or tracker is costing them hours. Codex lets them describe it in plain English and get a working tool back.
This half-day, hands-on class gives non-developers a safe, repeatable way to do that. We build the session around your team’s own files and recurring tasks, so everyone leaves with a small, checked build of their own and the habits to do the next one without us.
What people say after our classes
97%
said the class will help them get more value from AI at work
97%
enjoyed the class
88%
would recommend it to a colleague
5 hrs
average time each person expects to save every week
Post-class surveys across our classes.
Every session is built from scratch around your work, but every session builds the same foundations.
Getting started in the Codex app rather than a terminal, choosing how much Codex may do before it asks for approval, and writing an AGENTS.md file that tells it about your work, your files and your rules.
A plain-English explanation of what Codex does when it writes and runs code, and why a tool built from code gives the same answer every time while a chat reply can change.
Reviewing your recurring work for jobs that suit a small tool, weighing impact against risk, and agreeing what version one deliberately leaves out.
Having Codex ask you questions until the brief is clear, then asking for a plan you approve before it changes any files.
Reading what Codex changed in plain English, asking it to explain anything unclear, undoing what you do not want and choosing the right model and reasoning effort for the job.
Testing the output against a sample where you already know the answers and using Codex’s built-in review to catch problems before anyone relies on the result.
Handing bigger tasks to Codex in the cloud while you get on with your day, then reviewing the result before anything is used.
Keeping sensitive data, passwords and permissions locked down, agreeing which files and systems are in bounds, and giving IT a clear account of what was built.
The capstone
Everyone finishes the day with one small tool built around their own work and checked against known answers, so people leave with something they will use next week, not just notes.
We can usually map your highest-impact use cases in a single conversation.
Get in TouchThese are the kinds of tools non-developers have built in our workshops, from analysis that runs as checked code to dashboards that rebuild themselves. A coding agent lets people take this further, on their own files and data.
Recalculates key totals across statements and schedules and returns an exception log of contradictions and unexplained movements, with the cell or page behind each one.
Compares this period’s dashboard with the last, explains what changed and why it matters, and separates the facts from the guesses.
Turns a recurring customer data export into a short weekly report on movements, anomalies and the actions that need an owner.
Drawn from our anonymised client use case library.
Three signals showing Codex spreading well beyond developers. Each links to its source.
The class runs in six parts, moving from setup to a finished build. By the second half of the session everyone is working on their own task rather than following along with ours.
01
The Codex app, approval settings and an AGENTS.md for your work, with a short explanation of how Codex reads, writes and runs code.
02
A structured look at your recurring tasks, scored on impact, effort and risk, ending with one small build per person and a clear note of what it will not do yet.
03
Codex interviews you until the brief is clear, then sets out a plan that you approve before any file changes.
04
Codex builds while you watch the changes, ask it to explain its choices and send it back when it goes off course.
05
Run the tool against answers you already know, use Codex’s review to look for problems and fix the biggest one first.
06
Guardrails for data and access, a plain-English summary of what was built and checked, and a decision on whether to keep using it, share it or park it.
Delivered in person at your offices.
Going further
When a few people are building with Codex, the next step is making it a shared way of working. Our advanced training helps the keenest builders and the leaders backing them do that safely.
Firms that have rolled out ChatGPT and want their confident users to take the next step without switching tools.
Analysts, finance and operations teams who rebuild the same reports, reconciliations and trackers every week or month.
Anyone who has said “if only we had something that…” and been told it would sit in a backlog for a year.
Heads of team who want their people building small tools safely, with guardrails IT is comfortable with.
Other courses teams pair with this one.
Every Codex session is led by one of our instructors, with a dedicated tutor in the room so nobody gets left behind.
Founder + Chief AI Officer
AI Instructor
AI Instructor
AI Instructor
AI Instructor
AI Instructor
AI Instructor
AI Instructor
Yes. Our Codex training for developers is a separate half-day session for engineering teams, working in their own repositories with AGENTS.md, planning, code review and testing. Many firms run both.
Codex is OpenAI’s coding agent. Instead of replying in a chat, it writes and runs code to get a job done, which means it can build a tool that does the same task the same way every time. This class is about using that safely when you are not a developer.
No. The class is designed for people who have never written code. They describe what they want in plain English, and we teach them to brief Codex well, keep it in check and prove the result works.
Small tools built around their own work: a report that rebuilds itself from this month’s export, a check that compares two spreadsheets and lists the exceptions, a tidy-up of a messy folder or a simple dashboard. Everyone picks one small build so they leave with something that works.
Usually whichever your firm already pays for. If your team lives in ChatGPT, Codex keeps everything in one place. If you use Claude, our Claude Code class teaches the same habits. The skills carry across, so the choice matters less than people think.
In person at your offices by default, with an instructor and a dedicated tutor in the room. Having people on hand makes setup and the first real build much smoother. We can run it remotely if that suits your team better.
Attendees keep a curated prompt library and cheat sheets, and we run a follow-up review call to see what people have built, what got stuck and where to go next.
Yes. We are based in Covent Garden, London, and run most sessions in person at clients’ offices in London and across the UK. We also work with teams worldwide, from Paris and Berlin to New York and Singapore, and can run sessions remotely when that suits better.
From The Distillation, our newsletter on putting AI to work.
GPT-5.4 Thinking is a meaningful accuracy improvement, but the bigger story is what it says about where practical AI adoption is heading: better models, reusable workflows, and more interest in agentic ways of working.
General Purpose ·
AI agents go beyond chat. They complete entire tasks step by step. We put ChatGPT Agent to the test with a real business task to show what it can do.
Tom Hewitson ·
AI sycophancy is not a minor personality issue. When a model becomes too eager to please, it can reinforce bad assumptions, create risky records, and reduce the quality of decision-making.
General Purpose ·
Tell us about the reports, reconciliations and trackers that eat your team’s week, and we will design a day around building the tools that fix them.
Get in Touch